diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k.yaml b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k.yaml index 8454c619ff31557fa8fbfc389657b9196a11cd5f..f8fca234fbe057dc80d71b6b756a654d9ebb43e5 100644 --- a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k.yaml +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k.yaml @@ -9,6 +9,7 @@ training: num_gpus: 1 batch_size: 128 seed: 42 + report_to: wandb model: rkd_teacher_source: timewarp_action rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46 diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9a13e61b733d3ca3770b1b9a4290f6dfedb745ec --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5000130901875584, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/experiment_cfg/metadata.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..300eb070963337016e16465f50ba8351ea2eb25c --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.3139495849609375, + 0.4876587688922882, + 0.7196521759033203 + ], + "min": [ + -4.94293737411499, + -6.890198230743408, + 0.6996827125549316 + ], + "mean": [ + 2.749288365724937, + -2.0455216239909983, + 0.700532551020533 + ], + "std": [ + 1.6786295897046262, + 1.312897969434244, + 0.00133751758906258 + ], + "q01": [ + -1.4241907881148037, + -5.1716547935950885, + 0.6999670898768614 + ], + "q99": [ + 6.198111545831555, + -0.5873531334104489, + 0.7038833014242587 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.23084999587450247, + 0.6238431088581847 + ], + "std": [ + 0.0, + 0.0, + 0.6556404371644047, + 0.3572252105732042 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 1.7482354593273349e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.9014528393745422, + 0.8003877401351929, + 0.9829942584037781 + ], + "min": [ + -0.37471598386764526, + -0.8472502827644348, + -0.25070279836654663 + ], + "mean": [ + 0.28667332601863704, + -0.038315473170876704, + 0.45974658906777444 + ], + "std": [ + 0.17067058828305154, + 0.23569952037784195, + 0.21950702354181487 + ], + "q01": [ + 0.015172948963784228, + -0.44450502063220615, + 0.23314444103329338 + ], + "q99": [ + 0.5609069689572412, + 0.38454179128009547, + 0.7028102736473852 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999998807907104, + 0.9984455108642578, + 0.9434727430343628, + 0.9062229990959167 + ], + "min": [ + -0.9999930262565613, + -0.9988337755203247, + -0.9618149995803833, + 2.1872274658107926e-07 + ], + "mean": [ + -0.25672297401336, + 0.024425335806687216, + -0.08740977164483847, + 0.16608739644018397 + ], + "std": [ + 0.7932585208392383, + 0.3102260195580416, + 0.3772472576315148, + 0.17451959300158773 + ], + "q01": [ + -0.99379006987882, + -0.5478102050038828, + -0.6101777952971636, + 0.002274085014134748 + ], + "q99": [ + 0.8804856038896288, + 0.5910611092873037, + 0.5216058042993562, + 0.5231965974012255 + ] + }, + "gripper_qpos": { + "max": [ + 0.055869169533252716, + 0.010916369967162609 + ], + "min": [ + -0.011436971835792065, + -0.05664053186774254 + ], + "mean": [ + 0.031651551516052555, + -0.03161481387920421 + ], + "std": [ + 0.013119769222217526, + 0.01306078225192402 + ], + "q01": [ + 0.0060999246446777336, + -0.04062601653964198 + ], + "q99": [ + 0.04054891621517283, + -0.0059163567113456345 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.008144896753718373, + -0.00018893135146078263, + -0.0008739062727221845, + 0.0 + ], + "std": [ + 0.11030045424364411, + 0.10148082570313594, + 0.08975698373467506, + 0.0 + ], + "q01": [ + -0.07287373067581518, + -0.09446994199569948, + -0.07702818259249572, + 0.0 + ], + "q99": [ + 0.04485485414407898, + 0.09626941605965177, + 0.07610665906122742, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9221974313425939 + ], + "std": [ + 0.38670194342612674 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.384693946883099 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.01976129923017491, + -0.020870631344490034, + -0.05993476517349181 + ], + "std": [ + 0.4347024990804053, + 0.4221662976569997, + 0.3845506843044066 + ], + "q01": [ + -0.8835792939767807, + -0.9280919251713778, + -0.783361071470956 + ], + "q99": [ + 0.7622084438449307, + 0.8625125252609077, + 0.7470974753250706 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008089673068544335, + -0.026498254627927348, + 0.0029083551743059005 + ], + "std": [ + 0.11216734503733773, + 0.13206002613798629, + 0.12615870412692864 + ], + "q01": [ + -0.2814627972846672, + -0.4342386516115439, + -0.34489755628340574 + ], + "q99": [ + 0.31879353266939386, + 0.28411797683220563, + 0.3597718671669894 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3824228874769045 + ], + "std": [ + 0.9240015096469786 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.5597272305813592 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/trainer_state.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..dfc5525e5b0f86d5f2b8750082bc55bf222fecaa --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/trainer_state.json @@ -0,0 +1,138034 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 8.728542333430317, + "eval_steps": 500, + "global_step": 30000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.002909514111143439, + "flow_matching_loss": 0.01550564207136631, + "grad_norm": 3.1333067417144775, + "learning_rate": 9e-08, + "loss": 0.3826, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11892975866794586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1291983038187027, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.36705781817436217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999999821186065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10, + "weighted_flow_matching_loss": 0.01550564207136631, + "weighted_rkd_loss": 0.3670578092336655 + }, + { + "epoch": 0.005819028222286878, + "flow_matching_loss": 0.016602713521569966, + "grad_norm": 3.0757880210876465, + "learning_rate": 1.9e-07, + "loss": 0.3791, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11873007193207741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1250869430601597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.36254708766937255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999998509883881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20, + "weighted_flow_matching_loss": 0.016602713521569966, + "weighted_rkd_loss": 0.36254703104496 + }, + { + "epoch": 0.008728542333430317, + "flow_matching_loss": 0.01471330774948001, + "grad_norm": 2.5311384201049805, + "learning_rate": 2.9000000000000003e-07, + "loss": 0.3863, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.120485969632864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.13058788776397706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3715598344802856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999995768070221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 30, + "weighted_flow_matching_loss": 0.01471330774948001, + "weighted_rkd_loss": 0.371559676527977 + }, + { + "epoch": 0.011638056444573755, + "flow_matching_loss": 0.01742495195940137, + "grad_norm": 2.9743077754974365, + "learning_rate": 3.8999999999999997e-07, + "loss": 0.3838, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11861473023891449, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.12916370555758477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3663931548595428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999999177455902, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 40, + "weighted_flow_matching_loss": 0.01742495195940137, + "weighted_rkd_loss": 0.3663928508758545 + }, + { + "epoch": 0.014547570555717195, + "flow_matching_loss": 0.016293933894485236, + "grad_norm": 3.059931755065918, + "learning_rate": 4.9e-07, + "loss": 0.3763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11814559176564217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.12369979098439217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.35999098122119905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999986529350281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 50, + "weighted_flow_matching_loss": 0.016293933894485236, + "weighted_rkd_loss": 0.35999049842357633 + }, + { + "epoch": 0.017457084666860634, + "flow_matching_loss": 0.015452074538916349, + "grad_norm": 2.875544548034668, + "learning_rate": 5.9e-07, + "loss": 0.3618, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11515828296542167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11603189185261727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3463484555482864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999979555606842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 60, + "weighted_flow_matching_loss": 0.015452074538916349, + "weighted_rkd_loss": 0.3463477462530136 + }, + { + "epoch": 0.020366598778004074, + "flow_matching_loss": 0.016762651037424803, + "grad_norm": 3.106590509414673, + "learning_rate": 6.9e-07, + "loss": 0.362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11491406336426735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11540707796812058, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3452351987361908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999971389770508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 70, + "weighted_flow_matching_loss": 0.016762651037424803, + "weighted_rkd_loss": 0.345234215259552 + }, + { + "epoch": 0.02327611288914751, + "flow_matching_loss": 0.016631669737398623, + "grad_norm": 2.956284284591675, + "learning_rate": 7.900000000000001e-07, + "loss": 0.349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1118672177195549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1086497075855732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33238414227962493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999961912631988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 80, + "weighted_flow_matching_loss": 0.016631669737398623, + "weighted_rkd_loss": 0.33238287568092345 + }, + { + "epoch": 0.02618562700029095, + "flow_matching_loss": 0.016641059145331382, + "grad_norm": 2.5246362686157227, + "learning_rate": 8.900000000000001e-07, + "loss": 0.3577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11394575089216233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11313787326216698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.34102937281131745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999995106458664, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 90, + "weighted_flow_matching_loss": 0.016641059145331382, + "weighted_rkd_loss": 0.34102770388126374 + }, + { + "epoch": 0.02909514111143439, + "flow_matching_loss": 0.015576249454170465, + "grad_norm": 2.452319860458374, + "learning_rate": 9.9e-07, + "loss": 0.3413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11022084951400757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10530913323163986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3257508367300034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999938786029816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 100, + "weighted_flow_matching_loss": 0.015576249454170465, + "weighted_rkd_loss": 0.32574884295463563 + }, + { + "epoch": 0.03200465522257783, + "flow_matching_loss": 0.015184195153415203, + "grad_norm": 2.52805233001709, + "learning_rate": 1.0900000000000002e-06, + "loss": 0.3431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11051653325557709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10683545619249343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3278685241937637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999925076961518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 110, + "weighted_flow_matching_loss": 0.015184195153415203, + "weighted_rkd_loss": 0.32786606550216674 + }, + { + "epoch": 0.03491416933372127, + "flow_matching_loss": 0.017162218131124972, + "grad_norm": 2.1398568153381348, + "learning_rate": 1.19e-06, + "loss": 0.3285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10658194124698639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09821665287017822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.311380535364151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999910116195678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 120, + "weighted_flow_matching_loss": 0.017162218131124972, + "weighted_rkd_loss": 0.31137773394584656 + }, + { + "epoch": 0.037823683444864704, + "flow_matching_loss": 0.014759884029626847, + "grad_norm": 2.4938411712646484, + "learning_rate": 1.29e-06, + "loss": 0.3218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10611278638243675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09480291008949279, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3070284813642502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999893724918365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 130, + "weighted_flow_matching_loss": 0.014759884029626847, + "weighted_rkd_loss": 0.307025220990181 + }, + { + "epoch": 0.04073319755600815, + "flow_matching_loss": 0.017189308535307646, + "grad_norm": 2.2937464714050293, + "learning_rate": 1.39e-06, + "loss": 0.33, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10619326010346412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10044751316308975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.31283403635025026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999876022338867, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 140, + "weighted_flow_matching_loss": 0.017189308535307646, + "weighted_rkd_loss": 0.3128301650285721 + }, + { + "epoch": 0.043642711667151585, + "flow_matching_loss": 0.016698854975402355, + "grad_norm": 2.460662603378296, + "learning_rate": 1.49e-06, + "loss": 0.3139, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10227289497852325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09261290282011032, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2971586942672729, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999856889247895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 150, + "weighted_flow_matching_loss": 0.016698854975402355, + "weighted_rkd_loss": 0.2971544474363327 + }, + { + "epoch": 0.04655222577829502, + "flow_matching_loss": 0.01476333076134324, + "grad_norm": 2.256152391433716, + "learning_rate": 1.59e-06, + "loss": 0.3045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10208632946014404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08557594567537308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2897486060857773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999836325645447, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 160, + "weighted_flow_matching_loss": 0.01476333076134324, + "weighted_rkd_loss": 0.2897438704967499 + }, + { + "epoch": 0.049461739889438465, + "flow_matching_loss": 0.017648557014763356, + "grad_norm": 1.8371410369873047, + "learning_rate": 1.69e-06, + "loss": 0.2945, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09761943444609641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08161071985960007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2768495887517929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999814450740814, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 170, + "weighted_flow_matching_loss": 0.017648557014763356, + "weighted_rkd_loss": 0.2768444448709488 + }, + { + "epoch": 0.0523712540005819, + "flow_matching_loss": 0.016323222778737545, + "grad_norm": 1.7470184564590454, + "learning_rate": 1.79e-06, + "loss": 0.297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09842248931527138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08384960815310478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2806945890188217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999791383743286, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 180, + "weighted_flow_matching_loss": 0.016323222778737545, + "weighted_rkd_loss": 0.2806887298822403 + }, + { + "epoch": 0.055280768111725345, + "flow_matching_loss": 0.015566583629697562, + "grad_norm": 1.8406331539154053, + "learning_rate": 1.8900000000000001e-06, + "loss": 0.2876, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09661478996276855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07885545119643211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.27208503186702726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999976658821106, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 190, + "weighted_flow_matching_loss": 0.015566583629697562, + "weighted_rkd_loss": 0.2720786780118942 + }, + { + "epoch": 0.05819028222286878, + "flow_matching_loss": 0.0139276253990829, + "grad_norm": 1.7983605861663818, + "learning_rate": 1.99e-06, + "loss": 0.2933, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09830287471413612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08276783600449562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.27937358915805816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999740660190582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 200, + "weighted_flow_matching_loss": 0.0139276253990829, + "weighted_rkd_loss": 0.2793663501739502 + }, + { + "epoch": 0.06109979633401222, + "flow_matching_loss": 0.014639203902333975, + "grad_norm": 1.5982331037521362, + "learning_rate": 2.09e-06, + "loss": 0.2867, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0962134137749672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07960830554366112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.27203513085842135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999713242053986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 210, + "weighted_flow_matching_loss": 0.014639203902333975, + "weighted_rkd_loss": 0.272027325630188 + }, + { + "epoch": 0.06400931044515566, + "flow_matching_loss": 0.014623281452804804, + "grad_norm": 1.8078464269638062, + "learning_rate": 2.1899999999999998e-06, + "loss": 0.2799, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09471267461776733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07590387910604476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2653292268514633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999684512615203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 220, + "weighted_flow_matching_loss": 0.014623281452804804, + "weighted_rkd_loss": 0.2653208523988724 + }, + { + "epoch": 0.06691882455629909, + "flow_matching_loss": 0.015173925179988145, + "grad_norm": 1.5022159814834595, + "learning_rate": 2.29e-06, + "loss": 0.2723, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0923381894826889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0724766343832016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.25715301334857943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999654412269592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 230, + "weighted_flow_matching_loss": 0.015173925179988145, + "weighted_rkd_loss": 0.2571441322565079 + }, + { + "epoch": 0.06982833866744254, + "flow_matching_loss": 0.014600659161806107, + "grad_norm": 1.9008384943008423, + "learning_rate": 2.39e-06, + "loss": 0.2687, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09150076135993004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07110384702682496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.25410536527633665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999962306022644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 240, + "weighted_flow_matching_loss": 0.014600659161806107, + "weighted_rkd_loss": 0.254095783829689 + }, + { + "epoch": 0.07273785277858598, + "flow_matching_loss": 0.01673982935026288, + "grad_norm": 1.637431263923645, + "learning_rate": 2.4900000000000003e-06, + "loss": 0.2649, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08931345045566559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06949383355677127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2481207326054573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999590218067169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 250, + "weighted_flow_matching_loss": 0.01673982935026288, + "weighted_rkd_loss": 0.24811056554317473 + }, + { + "epoch": 0.07564736688972941, + "flow_matching_loss": 0.014957763813436031, + "grad_norm": 1.3640989065170288, + "learning_rate": 2.59e-06, + "loss": 0.2668, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09011384695768357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07157910726964474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.25180679857730864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999556005001068, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 260, + "weighted_flow_matching_loss": 0.014957763813436031, + "weighted_rkd_loss": 0.2517956241965294 + }, + { + "epoch": 0.07855688100087285, + "flow_matching_loss": 0.016356484964489937, + "grad_norm": 1.8053337335586548, + "learning_rate": 2.69e-06, + "loss": 0.2579, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08830772563815117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06498716361820697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24160261303186417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999520480632782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 270, + "weighted_flow_matching_loss": 0.016356484964489937, + "weighted_rkd_loss": 0.2415910318493843 + }, + { + "epoch": 0.0814663951120163, + "flow_matching_loss": 0.01568873142823577, + "grad_norm": 1.3989553451538086, + "learning_rate": 2.79e-06, + "loss": 0.2596, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08800508305430413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0678689256310463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24387909173965455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999483466148377, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 280, + "weighted_flow_matching_loss": 0.01568873142823577, + "weighted_rkd_loss": 0.2438664987683296 + }, + { + "epoch": 0.08437590922315973, + "flow_matching_loss": 0.01595317805185914, + "grad_norm": 1.3708187341690063, + "learning_rate": 2.8900000000000003e-06, + "loss": 0.2619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08899093046784401, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06795177534222603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.245933635532856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999944519996643, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 290, + "weighted_flow_matching_loss": 0.01595317805185914, + "weighted_rkd_loss": 0.24591998606920243 + }, + { + "epoch": 0.08728542333430317, + "flow_matching_loss": 0.015479940362274647, + "grad_norm": 1.258388876914978, + "learning_rate": 2.99e-06, + "loss": 0.2487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08562967330217361, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.062015485018491745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23327483087778092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999940550327301, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 300, + "weighted_flow_matching_loss": 0.015479940362274647, + "weighted_rkd_loss": 0.23326096832752227 + }, + { + "epoch": 0.09019493744544661, + "flow_matching_loss": 0.013275783509016037, + "grad_norm": 1.2859525680541992, + "learning_rate": 3.09e-06, + "loss": 0.2549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08794007524847984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06575872078537941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2416388735175133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999936431646347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 310, + "weighted_flow_matching_loss": 0.013275783509016037, + "weighted_rkd_loss": 0.24162351340055466 + }, + { + "epoch": 0.09310445155659004, + "flow_matching_loss": 0.014993091579526663, + "grad_norm": 1.0723984241485596, + "learning_rate": 3.19e-06, + "loss": 0.247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08494328707456589, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06216979101300239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23205636590719222, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999932199716568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 320, + "weighted_flow_matching_loss": 0.014993091579526663, + "weighted_rkd_loss": 0.2320406347513199 + }, + { + "epoch": 0.09601396566773349, + "flow_matching_loss": 0.014991955738514662, + "grad_norm": 1.1878397464752197, + "learning_rate": 3.29e-06, + "loss": 0.2518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08629920333623886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06418352499604225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23678192794322966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999278128147125, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 330, + "weighted_flow_matching_loss": 0.014991955738514662, + "weighted_rkd_loss": 0.2367648348212242 + }, + { + "epoch": 0.09892347977887693, + "flow_matching_loss": 0.016863089613616468, + "grad_norm": 1.381775975227356, + "learning_rate": 3.39e-06, + "loss": 0.2497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08491246923804283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06303159594535827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23285653442144394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999923312664032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 340, + "weighted_flow_matching_loss": 0.016863089613616468, + "weighted_rkd_loss": 0.2328386887907982 + }, + { + "epoch": 0.10183299389002037, + "flow_matching_loss": 0.01664912709966302, + "grad_norm": 1.1944044828414917, + "learning_rate": 3.49e-06, + "loss": 0.2459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08440772816538811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06048487983644009, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22930033653974533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999186396598816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 350, + "weighted_flow_matching_loss": 0.01664912709966302, + "weighted_rkd_loss": 0.22928168773651122 + }, + { + "epoch": 0.1047425080011638, + "flow_matching_loss": 0.015240008011460304, + "grad_norm": 1.3408432006835938, + "learning_rate": 3.5900000000000004e-06, + "loss": 0.2422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08358708173036575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0598407331854105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.227014896273613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999138534069061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 360, + "weighted_flow_matching_loss": 0.015240008011460304, + "weighted_rkd_loss": 0.2269953429698944 + }, + { + "epoch": 0.10765202211230725, + "flow_matching_loss": 0.014214762207120656, + "grad_norm": 1.0159770250320435, + "learning_rate": 3.6900000000000002e-06, + "loss": 0.234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0812440499663353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05733156315982342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21981966197490693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999089300632477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 370, + "weighted_flow_matching_loss": 0.014214762207120656, + "weighted_rkd_loss": 0.219799642264843 + }, + { + "epoch": 0.11056153622345069, + "flow_matching_loss": 0.014771767985075712, + "grad_norm": 1.1171494722366333, + "learning_rate": 3.7899999999999997e-06, + "loss": 0.2385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08258533999323844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.058532482013106346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22370315939188004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999038696289062, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 380, + "weighted_flow_matching_loss": 0.014771767985075712, + "weighted_rkd_loss": 0.22368165701627732 + }, + { + "epoch": 0.11347105033459412, + "flow_matching_loss": 0.015692740492522717, + "grad_norm": 1.0914907455444336, + "learning_rate": 3.890000000000001e-06, + "loss": 0.2334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08070294633507728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05635931976139545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2177652135491371, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998986721038818, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 390, + "weighted_flow_matching_loss": 0.015692740492522717, + "weighted_rkd_loss": 0.21774315983057022 + }, + { + "epoch": 0.11638056444573756, + "flow_matching_loss": 0.01563012823462486, + "grad_norm": 0.9959664344787598, + "learning_rate": 3.99e-06, + "loss": 0.2301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07985119670629501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.054760342836380003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21446273773908614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99989333152771, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 400, + "weighted_flow_matching_loss": 0.01563012823462486, + "weighted_rkd_loss": 0.21443986445665358 + }, + { + "epoch": 0.11929007855688101, + "flow_matching_loss": 0.01667370181530714, + "grad_norm": 0.9493701457977295, + "learning_rate": 4.09e-06, + "loss": 0.2291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07920584604144096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0539976492524147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21240934282541274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998878479003906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 410, + "weighted_flow_matching_loss": 0.01667370181530714, + "weighted_rkd_loss": 0.21238551884889603 + }, + { + "epoch": 0.12219959266802444, + "flow_matching_loss": 0.016663321852684022, + "grad_norm": 1.090376615524292, + "learning_rate": 4.19e-06, + "loss": 0.2292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07916617915034294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0542093962430954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21254175454378127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998822391033173, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 420, + "weighted_flow_matching_loss": 0.016663321852684022, + "weighted_rkd_loss": 0.2125167265534401 + }, + { + "epoch": 0.12510910677916787, + "flow_matching_loss": 0.015671449527144433, + "grad_norm": 0.8521004319190979, + "learning_rate": 4.29e-06, + "loss": 0.2311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08002924695611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.055418449267745015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21547694206237794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998764932155609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 430, + "weighted_flow_matching_loss": 0.015671449527144433, + "weighted_rkd_loss": 0.21545032858848573 + }, + { + "epoch": 0.12801862089031132, + "flow_matching_loss": 0.01636008694767952, + "grad_norm": 0.982880175113678, + "learning_rate": 4.39e-06, + "loss": 0.234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08097354918718339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05570739023387432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.217654487490654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998705983161926, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 440, + "weighted_flow_matching_loss": 0.01636008694767952, + "weighted_rkd_loss": 0.2176263302564621 + }, + { + "epoch": 0.13092813500145475, + "flow_matching_loss": 0.015555653907358646, + "grad_norm": 0.9242200255393982, + "learning_rate": 4.49e-06, + "loss": 0.2275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07865319848060608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.054678288847208024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.211984683573246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998645782470703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 450, + "weighted_flow_matching_loss": 0.015555653907358646, + "weighted_rkd_loss": 0.2119559794664383 + }, + { + "epoch": 0.13383764911259818, + "flow_matching_loss": 0.017018662579357625, + "grad_norm": 1.1206122636795044, + "learning_rate": 4.59e-06, + "loss": 0.2262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0779559150338173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05325199924409389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20916383117437362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998584032058716, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 460, + "weighted_flow_matching_loss": 0.017018662579357625, + "weighted_rkd_loss": 0.20913420766592025 + }, + { + "epoch": 0.13674716322374164, + "flow_matching_loss": 0.0143358931876719, + "grad_norm": 0.9243751168251038, + "learning_rate": 4.69e-06, + "loss": 0.2195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07668408080935478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.051865966990590096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20523413121700287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998521208763123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 470, + "weighted_flow_matching_loss": 0.0143358931876719, + "weighted_rkd_loss": 0.20520378202199935 + }, + { + "epoch": 0.13965667733488507, + "flow_matching_loss": 0.015302463807165623, + "grad_norm": 0.8226845264434814, + "learning_rate": 4.790000000000001e-06, + "loss": 0.2161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07553937435150146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04974716007709503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20082591027021407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999845689535141, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 480, + "weighted_flow_matching_loss": 0.015302463807165623, + "weighted_rkd_loss": 0.2007949247956276 + }, + { + "epoch": 0.1425661914460285, + "flow_matching_loss": 0.01320085171610117, + "grad_norm": 0.989552915096283, + "learning_rate": 4.890000000000001e-06, + "loss": 0.2135, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07562590464949608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.049048150703310964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20029996186494828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998391091823577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 490, + "weighted_flow_matching_loss": 0.01320085171610117, + "weighted_rkd_loss": 0.20026773810386658 + }, + { + "epoch": 0.14547570555717196, + "flow_matching_loss": 0.014644200913608074, + "grad_norm": 1.1507350206375122, + "learning_rate": 4.9900000000000005e-06, + "loss": 0.215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0753984197974205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04957800731062889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20037484765052796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998324036598205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 500, + "weighted_flow_matching_loss": 0.014644200913608074, + "weighted_rkd_loss": 0.20034126043319703 + }, + { + "epoch": 0.1483852196683154, + "flow_matching_loss": 0.015617415960878133, + "grad_norm": 0.8233609199523926, + "learning_rate": 5.0899999999999995e-06, + "loss": 0.2144, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0747145175933838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.049428869411349295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1988579049706459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998255550861359, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 510, + "weighted_flow_matching_loss": 0.015617415960878133, + "weighted_rkd_loss": 0.19882321506738662 + }, + { + "epoch": 0.15129473377945882, + "flow_matching_loss": 0.013979686237871648, + "grad_norm": 0.6986652612686157, + "learning_rate": 5.1899999999999994e-06, + "loss": 0.2106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0740484707057476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.048524245247244836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19662118703126907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998185694217682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 520, + "weighted_flow_matching_loss": 0.013979686237871648, + "weighted_rkd_loss": 0.1965855211019516 + }, + { + "epoch": 0.15420424789060228, + "flow_matching_loss": 0.014558969624340534, + "grad_norm": 0.8354929685592651, + "learning_rate": 5.29e-06, + "loss": 0.2089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07348118275403977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04746444597840309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19442681074142457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998114585876465, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 530, + "weighted_flow_matching_loss": 0.014558969624340534, + "weighted_rkd_loss": 0.194390144944191 + }, + { + "epoch": 0.1571137620017457, + "flow_matching_loss": 0.01430925251916051, + "grad_norm": 0.8181521892547607, + "learning_rate": 5.39e-06, + "loss": 0.2023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07138690799474716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04521216079592705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18798597753047944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998041927814484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 540, + "weighted_flow_matching_loss": 0.01430925251916051, + "weighted_rkd_loss": 0.18794917911291123 + }, + { + "epoch": 0.16002327611288913, + "flow_matching_loss": 0.016528716217726468, + "grad_norm": 0.5677958130836487, + "learning_rate": 5.49e-06, + "loss": 0.2009, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07015014141798019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0441164381802082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18441672176122664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997968018054962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 550, + "weighted_flow_matching_loss": 0.016528716217726468, + "weighted_rkd_loss": 0.18437924981117249 + }, + { + "epoch": 0.1629327902240326, + "flow_matching_loss": 0.015504703670740128, + "grad_norm": 0.8407217860221863, + "learning_rate": 5.59e-06, + "loss": 0.199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06993300393223763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.043634640797972676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18350064903497695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997892677783966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 560, + "weighted_flow_matching_loss": 0.015504703670740128, + "weighted_rkd_loss": 0.18346198201179503 + }, + { + "epoch": 0.16584230433517602, + "flow_matching_loss": 0.015375380776822567, + "grad_norm": 0.597751259803772, + "learning_rate": 5.690000000000001e-06, + "loss": 0.1969, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06903594061732292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04348538145422935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18155726194381713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997815787792206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 570, + "weighted_flow_matching_loss": 0.015375380776822567, + "weighted_rkd_loss": 0.18151761144399642 + }, + { + "epoch": 0.16875181844631945, + "flow_matching_loss": 0.015282897744327784, + "grad_norm": 0.8267744183540344, + "learning_rate": 5.7900000000000005e-06, + "loss": 0.1939, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06742648854851722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04376812018454075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1786210983991623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997737944126129, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 580, + "weighted_flow_matching_loss": 0.015282897744327784, + "weighted_rkd_loss": 0.17858068346977235 + }, + { + "epoch": 0.1716613325574629, + "flow_matching_loss": 0.015070965979248285, + "grad_norm": 0.926098644733429, + "learning_rate": 5.89e-06, + "loss": 0.1905, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06685581281781197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04179429188370705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17550591826438905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997658431529999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 590, + "weighted_flow_matching_loss": 0.015070965979248285, + "weighted_rkd_loss": 0.17546481788158416 + }, + { + "epoch": 0.17457084666860634, + "flow_matching_loss": 0.0156327398493886, + "grad_norm": 0.5555564165115356, + "learning_rate": 5.99e-06, + "loss": 0.1868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06575434878468514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03972295857965946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1712316557765007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997577667236328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 600, + "weighted_flow_matching_loss": 0.0156327398493886, + "weighted_rkd_loss": 0.17119017839431763 + }, + { + "epoch": 0.17748036077974977, + "flow_matching_loss": 0.016743500623852013, + "grad_norm": 0.7697194814682007, + "learning_rate": 6.090000000000001e-06, + "loss": 0.1863, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06464217826724053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0402839370071888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1695682942867279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997495412826538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 610, + "weighted_flow_matching_loss": 0.016743500623852013, + "weighted_rkd_loss": 0.16952582597732543 + }, + { + "epoch": 0.18038987489089323, + "flow_matching_loss": 0.01587068410590291, + "grad_norm": 0.7104418277740479, + "learning_rate": 6.19e-06, + "loss": 0.1861, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06500126905739308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04029692448675633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1702994614839554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997411906719208, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 620, + "weighted_flow_matching_loss": 0.01587068410590291, + "weighted_rkd_loss": 0.170255383849144 + }, + { + "epoch": 0.18329938900203666, + "flow_matching_loss": 0.015684692095965146, + "grad_norm": 0.7447714805603027, + "learning_rate": 6.29e-06, + "loss": 0.1789, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0631203930824995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03703546114265919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1632762461900711, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997327029705048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 630, + "weighted_flow_matching_loss": 0.015684692095965146, + "weighted_rkd_loss": 0.16323259770870208 + }, + { + "epoch": 0.18620890311318009, + "flow_matching_loss": 0.015971421729773283, + "grad_norm": 0.6658583879470825, + "learning_rate": 6.39e-06, + "loss": 0.1783, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06258972398936749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.037160228937864304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16233967393636703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997240781784058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 640, + "weighted_flow_matching_loss": 0.015971421729773283, + "weighted_rkd_loss": 0.16229488104581832 + }, + { + "epoch": 0.18911841722432354, + "flow_matching_loss": 0.015593070164322853, + "grad_norm": 0.6570332646369934, + "learning_rate": 6.49e-06, + "loss": 0.1798, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06315903179347515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03794240839779377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16426047384738923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997153162956238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 650, + "weighted_flow_matching_loss": 0.015593070164322853, + "weighted_rkd_loss": 0.16421371102333068 + }, + { + "epoch": 0.19202793133546697, + "flow_matching_loss": 0.014295803662389518, + "grad_norm": 0.6184343695640564, + "learning_rate": 6.5900000000000004e-06, + "loss": 0.1758, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06194727942347526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.037611284479498866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16150584518909455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997064173221588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 660, + "weighted_flow_matching_loss": 0.014295803662389518, + "weighted_rkd_loss": 0.16145842969417573 + }, + { + "epoch": 0.1949374454466104, + "flow_matching_loss": 0.014863279275596141, + "grad_norm": 0.5587443113327026, + "learning_rate": 6.69e-06, + "loss": 0.1691, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0595221396535635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03524497672915459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15428925603628157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996973752975464, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 670, + "weighted_flow_matching_loss": 0.014863279275596141, + "weighted_rkd_loss": 0.15424256324768065 + }, + { + "epoch": 0.19784695955775386, + "flow_matching_loss": 0.015302929934114218, + "grad_norm": 0.6179419159889221, + "learning_rate": 6.79e-06, + "loss": 0.1656, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.058148641139268875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034094827994704244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15039211064577102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996882021427155, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 680, + "weighted_flow_matching_loss": 0.015302929934114218, + "weighted_rkd_loss": 0.15034520626068115 + }, + { + "epoch": 0.2007564736688973, + "flow_matching_loss": 0.015332216583192348, + "grad_norm": 0.45495113730430603, + "learning_rate": 6.89e-06, + "loss": 0.1672, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05860772915184498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03471335731446743, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15192881524562835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999678885936737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 690, + "weighted_flow_matching_loss": 0.015332216583192348, + "weighted_rkd_loss": 0.15188003480434417 + }, + { + "epoch": 0.20366598778004075, + "flow_matching_loss": 0.016358512360602616, + "grad_norm": 0.6826573610305786, + "learning_rate": 6.990000000000001e-06, + "loss": 0.1605, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.056005945429205894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0322153490036726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14422723799943923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996694326400757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 700, + "weighted_flow_matching_loss": 0.016358512360602616, + "weighted_rkd_loss": 0.14417956918478012 + }, + { + "epoch": 0.20657550189118418, + "flow_matching_loss": 0.01659400323405862, + "grad_norm": 0.47655436396598816, + "learning_rate": 7.090000000000001e-06, + "loss": 0.1622, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.056083903089165686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03349186442792416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1456596702337265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996598482131958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 710, + "weighted_flow_matching_loss": 0.01659400323405862, + "weighted_rkd_loss": 0.14561012536287307 + }, + { + "epoch": 0.2094850160023276, + "flow_matching_loss": 0.014927447587251664, + "grad_norm": 0.48825377225875854, + "learning_rate": 7.19e-06, + "loss": 0.1583, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.055617480725049975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03214142732322216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14337638765573502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996501326560974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 720, + "weighted_flow_matching_loss": 0.014927447587251664, + "weighted_rkd_loss": 0.14332622140645981 + }, + { + "epoch": 0.21239453011347106, + "flow_matching_loss": 0.01564619494602084, + "grad_norm": 0.47077274322509766, + "learning_rate": 7.29e-06, + "loss": 0.1547, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.053855713829398154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.031382813677191736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13909424245357513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996402621269226, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 730, + "weighted_flow_matching_loss": 0.01564619494602084, + "weighted_rkd_loss": 0.1390442132949829 + }, + { + "epoch": 0.2153040442246145, + "flow_matching_loss": 0.015656139794737102, + "grad_norm": 0.5632472038269043, + "learning_rate": 7.3899999999999995e-06, + "loss": 0.1565, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.054376039281487466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03219075966626406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14094283878803254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996302723884583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 740, + "weighted_flow_matching_loss": 0.015656139794737102, + "weighted_rkd_loss": 0.14089073240756989 + }, + { + "epoch": 0.21821355833575792, + "flow_matching_loss": 0.017486758809536696, + "grad_norm": 0.6317457556724548, + "learning_rate": 7.49e-06, + "loss": 0.1538, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05249287150800228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03142751175910234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13641325533390045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996201395988464, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 750, + "weighted_flow_matching_loss": 0.017486758809536696, + "weighted_rkd_loss": 0.13636143505573273 + }, + { + "epoch": 0.22112307244690138, + "flow_matching_loss": 0.01615606565028429, + "grad_norm": 0.5441702604293823, + "learning_rate": 7.59e-06, + "loss": 0.1485, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05128881558775902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02980778906494379, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13238541781902313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996098577976227, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 760, + "weighted_flow_matching_loss": 0.01615606565028429, + "weighted_rkd_loss": 0.13233377337455748 + }, + { + "epoch": 0.2240325865580448, + "flow_matching_loss": 0.016218704264611007, + "grad_norm": 0.5214950442314148, + "learning_rate": 7.690000000000001e-06, + "loss": 0.1463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05009442158043385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02991833332926035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13010717630386354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995994508266449, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 770, + "weighted_flow_matching_loss": 0.016218704264611007, + "weighted_rkd_loss": 0.1300550617277622 + }, + { + "epoch": 0.22694210066918824, + "flow_matching_loss": 0.015413323044776916, + "grad_norm": 0.5139783620834351, + "learning_rate": 7.79e-06, + "loss": 0.1456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05028948374092579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02970222905278206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13028119504451752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995889067649841, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 780, + "weighted_flow_matching_loss": 0.015413323044776916, + "weighted_rkd_loss": 0.13022763580083846 + }, + { + "epoch": 0.2298516147803317, + "flow_matching_loss": 0.015303638111799956, + "grad_norm": 0.5124013423919678, + "learning_rate": 7.89e-06, + "loss": 0.137, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04713186100125313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027437731437385083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1217014528810978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995782315731049, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 790, + "weighted_flow_matching_loss": 0.015303638111799956, + "weighted_rkd_loss": 0.12165012434124947 + }, + { + "epoch": 0.23276112889147513, + "flow_matching_loss": 0.015144110936671496, + "grad_norm": 0.5285835266113281, + "learning_rate": 7.99e-06, + "loss": 0.1413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0488121535629034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028633879870176314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1262581877410412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995674014091491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 800, + "weighted_flow_matching_loss": 0.015144110936671496, + "weighted_rkd_loss": 0.12620357275009156 + }, + { + "epoch": 0.23567064300261856, + "flow_matching_loss": 0.017987354192882776, + "grad_norm": 0.5882594585418701, + "learning_rate": 8.09e-06, + "loss": 0.1452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.049063516408205034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029149488359689713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1272765204310417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995564520359039, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 810, + "weighted_flow_matching_loss": 0.017987354192882776, + "weighted_rkd_loss": 0.12722006291151047 + }, + { + "epoch": 0.23858015711376201, + "flow_matching_loss": 0.016999167297035454, + "grad_norm": 0.497800350189209, + "learning_rate": 8.190000000000001e-06, + "loss": 0.1381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04716922417283058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026820527203381062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12115897685289383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995453536510468, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 820, + "weighted_flow_matching_loss": 0.016999167297035454, + "weighted_rkd_loss": 0.12110389471054077 + }, + { + "epoch": 0.24148967122490544, + "flow_matching_loss": 0.01558567238971591, + "grad_norm": 0.5204104781150818, + "learning_rate": 8.29e-06, + "loss": 0.1368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046873098611831664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02754077911376953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12128697484731674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995341181755066, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 830, + "weighted_flow_matching_loss": 0.01558567238971591, + "weighted_rkd_loss": 0.1212304711341858 + }, + { + "epoch": 0.24439918533604887, + "flow_matching_loss": 0.016919424664229153, + "grad_norm": 0.49191349744796753, + "learning_rate": 8.390000000000001e-06, + "loss": 0.1358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04627951644361019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02634277231991291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11890180483460426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995227515697479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 840, + "weighted_flow_matching_loss": 0.016919424664229153, + "weighted_rkd_loss": 0.11884506121277809 + }, + { + "epoch": 0.24730869944719233, + "flow_matching_loss": 0.01605758387595415, + "grad_norm": 0.4036836326122284, + "learning_rate": 8.49e-06, + "loss": 0.13, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04406052641570568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02592048328369856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11404153630137444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995112538337707, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 850, + "weighted_flow_matching_loss": 0.01605758387595415, + "weighted_rkd_loss": 0.11398580223321915 + }, + { + "epoch": 0.25021821355833573, + "flow_matching_loss": 0.018025321420282125, + "grad_norm": 0.46462222933769226, + "learning_rate": 8.59e-06, + "loss": 0.1328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04448857828974724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02581450715661049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1147916629910469, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994996070861817, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 860, + "weighted_flow_matching_loss": 0.018025321420282125, + "weighted_rkd_loss": 0.11473423540592194 + }, + { + "epoch": 0.2531277276694792, + "flow_matching_loss": 0.014978806115686893, + "grad_norm": 0.5570517182350159, + "learning_rate": 8.690000000000002e-06, + "loss": 0.1276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043908930569887164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02489650472998619, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11271436661481857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999487841129303, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 870, + "weighted_flow_matching_loss": 0.014978806115686893, + "weighted_rkd_loss": 0.11265662387013435 + }, + { + "epoch": 0.25603724178062265, + "flow_matching_loss": 0.016052319016307593, + "grad_norm": 0.44528305530548096, + "learning_rate": 8.79e-06, + "loss": 0.1244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0421783335506916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0240774292498827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10843409597873688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994759202003479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 880, + "weighted_flow_matching_loss": 0.016052319016307593, + "weighted_rkd_loss": 0.10837726294994354 + }, + { + "epoch": 0.2589467558917661, + "flow_matching_loss": 0.017423460725694895, + "grad_norm": 0.46206334233283997, + "learning_rate": 8.89e-06, + "loss": 0.1225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04085315689444542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023390434309840204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10509674847126008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994638621807098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 890, + "weighted_flow_matching_loss": 0.017423460725694895, + "weighted_rkd_loss": 0.10504041016101837 + }, + { + "epoch": 0.2618562700029095, + "flow_matching_loss": 0.01734229288995266, + "grad_norm": 0.52879399061203, + "learning_rate": 8.99e-06, + "loss": 0.1243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04163628853857517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023732551001012327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10700512826442718, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994516789913177, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 900, + "weighted_flow_matching_loss": 0.01734229288995266, + "weighted_rkd_loss": 0.10694645047187805 + }, + { + "epoch": 0.26476578411405294, + "flow_matching_loss": 0.017090780287981035, + "grad_norm": 0.4703419804573059, + "learning_rate": 9.09e-06, + "loss": 0.1215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04065173380076885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023155871964991092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10445933938026428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994393527507782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 910, + "weighted_flow_matching_loss": 0.017090780287981035, + "weighted_rkd_loss": 0.10440076887607574 + }, + { + "epoch": 0.26767529822519637, + "flow_matching_loss": 0.016106484550982715, + "grad_norm": 0.4477415680885315, + "learning_rate": 9.19e-06, + "loss": 0.1209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04076178632676601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02332416605204344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10484773889183999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994268715381622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 920, + "weighted_flow_matching_loss": 0.016106484550982715, + "weighted_rkd_loss": 0.10478764921426773 + }, + { + "epoch": 0.27058481233633985, + "flow_matching_loss": 0.01691915197297931, + "grad_norm": 0.4876635670661926, + "learning_rate": 9.289999999999999e-06, + "loss": 0.1208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04010870270431042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023736794665455817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10395420119166374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994142949581146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 930, + "weighted_flow_matching_loss": 0.01691915197297931, + "weighted_rkd_loss": 0.1038933128118515 + }, + { + "epoch": 0.2734943264474833, + "flow_matching_loss": 0.016947032231837512, + "grad_norm": 0.5491437911987305, + "learning_rate": 9.39e-06, + "loss": 0.1143, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03804816119372845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021268735453486443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09736505895853043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994015514850616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 940, + "weighted_flow_matching_loss": 0.016947032231837512, + "weighted_rkd_loss": 0.09730678647756577 + }, + { + "epoch": 0.2764038405586267, + "flow_matching_loss": 0.017421757988631725, + "grad_norm": 0.4656124413013458, + "learning_rate": 9.49e-06, + "loss": 0.1178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.038932641968131065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02257927618920803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10044455975294113, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993886768817901, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 950, + "weighted_flow_matching_loss": 0.017421757988631725, + "weighted_rkd_loss": 0.10038315057754517 + }, + { + "epoch": 0.27931335466977014, + "flow_matching_loss": 0.017892222665250302, + "grad_norm": 0.5980035066604614, + "learning_rate": 9.59e-06, + "loss": 0.1192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03925154060125351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022913327254354954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1014164075255394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993756711483002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 960, + "weighted_flow_matching_loss": 0.017892222665250302, + "weighted_rkd_loss": 0.10135308280587196 + }, + { + "epoch": 0.28222286878091357, + "flow_matching_loss": 0.015580792725086213, + "grad_norm": 0.42320093512535095, + "learning_rate": 9.69e-06, + "loss": 0.112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03752100765705109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021445353887975215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09648736864328385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993625104427337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 970, + "weighted_flow_matching_loss": 0.015580792725086213, + "weighted_rkd_loss": 0.09642586559057235 + }, + { + "epoch": 0.285132382892057, + "flow_matching_loss": 0.01739452127367258, + "grad_norm": 0.4320252537727356, + "learning_rate": 9.79e-06, + "loss": 0.1097, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03602619655430317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020282964035868645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09233535677194596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993492364883423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 980, + "weighted_flow_matching_loss": 0.01739452127367258, + "weighted_rkd_loss": 0.09227527678012848 + }, + { + "epoch": 0.2880418970032005, + "flow_matching_loss": 0.016440205741673708, + "grad_norm": 0.4890933334827423, + "learning_rate": 9.89e-06, + "loss": 0.1129, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03740277215838432, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02173846196383238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09654400646686553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993358135223389, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 990, + "weighted_flow_matching_loss": 0.016440205741673708, + "weighted_rkd_loss": 0.09647987559437751 + }, + { + "epoch": 0.2909514111143439, + "flow_matching_loss": 0.017298280727118254, + "grad_norm": 0.5970342755317688, + "learning_rate": 9.990000000000001e-06, + "loss": 0.1115, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03657592087984085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021107819490134717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09425966143608093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993222475051879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1000, + "weighted_flow_matching_loss": 0.017298280727118254, + "weighted_rkd_loss": 0.09419575855135917 + }, + { + "epoch": 0.29386092522548735, + "flow_matching_loss": 0.016320373211055995, + "grad_norm": 0.4389134347438812, + "learning_rate": 1.009e-05, + "loss": 0.1102, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0366119921207428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020740056037902833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09396403953433037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993085503578186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1010, + "weighted_flow_matching_loss": 0.016320373211055995, + "weighted_rkd_loss": 0.09389907494187355 + }, + { + "epoch": 0.2967704393366308, + "flow_matching_loss": 0.015554591175168752, + "grad_norm": 0.8931312561035156, + "learning_rate": 1.019e-05, + "loss": 0.1103, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036759023740887645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021259809844195843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09477785751223564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992947280406952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1020, + "weighted_flow_matching_loss": 0.015554591175168752, + "weighted_rkd_loss": 0.09471100121736527 + }, + { + "epoch": 0.2996799534477742, + "flow_matching_loss": 0.016102798748761415, + "grad_norm": 0.45124495029449463, + "learning_rate": 1.0290000000000001e-05, + "loss": 0.1055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03498197048902511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019538758508861064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0895026981830597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992807626724243, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1030, + "weighted_flow_matching_loss": 0.016102798748761415, + "weighted_rkd_loss": 0.08943832889199257 + }, + { + "epoch": 0.30258946755891764, + "flow_matching_loss": 0.015958820190280675, + "grad_norm": 0.5179141163825989, + "learning_rate": 1.039e-05, + "loss": 0.1049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034519029408693315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019967910088598727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08900596722960472, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992666482925415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1040, + "weighted_flow_matching_loss": 0.015958820190280675, + "weighted_rkd_loss": 0.08894068598747254 + }, + { + "epoch": 0.3054989816700611, + "flow_matching_loss": 0.014143578335642814, + "grad_norm": 0.4228968918323517, + "learning_rate": 1.0490000000000001e-05, + "loss": 0.1036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03473798036575317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02002559993416071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0895015612244606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992524206638336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1050, + "weighted_flow_matching_loss": 0.014143578335642814, + "weighted_rkd_loss": 0.08943464681506157 + }, + { + "epoch": 0.30840849578120455, + "flow_matching_loss": 0.015324947703629733, + "grad_norm": 0.3882687985897064, + "learning_rate": 1.059e-05, + "loss": 0.1032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03423340637236834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019473423063755036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08794023618102073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992380261421203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1060, + "weighted_flow_matching_loss": 0.015324947703629733, + "weighted_rkd_loss": 0.087873225659132 + }, + { + "epoch": 0.311318009892348, + "flow_matching_loss": 0.017320666182786226, + "grad_norm": 0.41547322273254395, + "learning_rate": 1.0690000000000001e-05, + "loss": 0.1042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03394942879676819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019060219824314117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0869590774178505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992235124111175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1070, + "weighted_flow_matching_loss": 0.017320666182786226, + "weighted_rkd_loss": 0.0868915632367134 + }, + { + "epoch": 0.3142275240034914, + "flow_matching_loss": 0.017014538962394, + "grad_norm": 0.4521830677986145, + "learning_rate": 1.0790000000000002e-05, + "loss": 0.1042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03394660819321871, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019362064264714717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08725528120994568, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992088615894318, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1080, + "weighted_flow_matching_loss": 0.017014538962394, + "weighted_rkd_loss": 0.08718624264001847 + }, + { + "epoch": 0.31713703811463484, + "flow_matching_loss": 0.015395821537822485, + "grad_norm": 0.4296054244041443, + "learning_rate": 1.089e-05, + "loss": 0.097, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032095359079539774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0174601498991251, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08165086805820465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991940557956696, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1090, + "weighted_flow_matching_loss": 0.015395821537822485, + "weighted_rkd_loss": 0.08158506751060486 + }, + { + "epoch": 0.32004655222577827, + "flow_matching_loss": 0.016578020993620157, + "grad_norm": 0.4908936321735382, + "learning_rate": 1.099e-05, + "loss": 0.0982, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0317534239962697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01815151609480381, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08165836483240127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991791367530822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1100, + "weighted_flow_matching_loss": 0.016578020993620157, + "weighted_rkd_loss": 0.08159132972359658 + }, + { + "epoch": 0.32295606633692175, + "flow_matching_loss": 0.015285632107406855, + "grad_norm": 0.4644777178764343, + "learning_rate": 1.1089999999999999e-05, + "loss": 0.0973, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031901773624122144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01832310575991869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08212665393948555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999164080619812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1110, + "weighted_flow_matching_loss": 0.015285632107406855, + "weighted_rkd_loss": 0.08205800727009774 + }, + { + "epoch": 0.3258655804480652, + "flow_matching_loss": 0.015305731073021888, + "grad_norm": 0.4871329069137573, + "learning_rate": 1.119e-05, + "loss": 0.0988, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032422063313424586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018691304698586465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08353543058037757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991488814353943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1120, + "weighted_flow_matching_loss": 0.015305731073021888, + "weighted_rkd_loss": 0.08346433863043785 + }, + { + "epoch": 0.3287750945592086, + "flow_matching_loss": 0.017702855914831162, + "grad_norm": 0.4126362204551697, + "learning_rate": 1.129e-05, + "loss": 0.0982, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03139788862317801, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017759868409484625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08055564612150193, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991335451602936, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1130, + "weighted_flow_matching_loss": 0.017702855914831162, + "weighted_rkd_loss": 0.08048585429787636 + }, + { + "epoch": 0.33168460867035204, + "flow_matching_loss": 0.016021419595927, + "grad_norm": 0.41553357243537903, + "learning_rate": 1.139e-05, + "loss": 0.0976, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031856043823063375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017890631593763827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08160271942615509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991180717945098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1140, + "weighted_flow_matching_loss": 0.016021419595927, + "weighted_rkd_loss": 0.08153074607253075 + }, + { + "epoch": 0.3345941227814955, + "flow_matching_loss": 0.015615587588399649, + "grad_norm": 0.364560604095459, + "learning_rate": 1.149e-05, + "loss": 0.0975, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03176059126853943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01848536916077137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08200655207037925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991024553775787, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1150, + "weighted_flow_matching_loss": 0.015615587588399649, + "weighted_rkd_loss": 0.08193295300006867 + }, + { + "epoch": 0.3375036368926389, + "flow_matching_loss": 0.01663275035098195, + "grad_norm": 0.40495726466178894, + "learning_rate": 1.1589999999999999e-05, + "loss": 0.0953, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030630207248032094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017503476608544587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07876389026641846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990867078304291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1160, + "weighted_flow_matching_loss": 0.01663275035098195, + "weighted_rkd_loss": 0.07869197055697441 + }, + { + "epoch": 0.3404131510037824, + "flow_matching_loss": 0.01569695193320513, + "grad_norm": 0.42846280336380005, + "learning_rate": 1.169e-05, + "loss": 0.0932, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03025749232620001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017012717388570308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07752770259976387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999070817232132, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1170, + "weighted_flow_matching_loss": 0.01569695193320513, + "weighted_rkd_loss": 0.0774556614458561 + }, + { + "epoch": 0.3433226651149258, + "flow_matching_loss": 0.019037979282438756, + "grad_norm": 0.413841187953949, + "learning_rate": 1.179e-05, + "loss": 0.0958, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02996022757142782, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016912759467959403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07683321461081505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990548193454742, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1180, + "weighted_flow_matching_loss": 0.019037979282438756, + "weighted_rkd_loss": 0.07676060795783997 + }, + { + "epoch": 0.34623217922606925, + "flow_matching_loss": 0.016569431964308025, + "grad_norm": 0.4611040949821472, + "learning_rate": 1.189e-05, + "loss": 0.0956, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03088572882115841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017346873879432678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07911833077669143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990386366844177, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1190, + "weighted_flow_matching_loss": 0.016569431964308025, + "weighted_rkd_loss": 0.07904228419065476 + }, + { + "epoch": 0.3491416933372127, + "flow_matching_loss": 0.015800755377858876, + "grad_norm": 0.3633919358253479, + "learning_rate": 1.199e-05, + "loss": 0.0904, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029256513342261314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01618056232109666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07469358965754509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990223526954651, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1200, + "weighted_flow_matching_loss": 0.015800755377858876, + "weighted_rkd_loss": 0.07462057545781135 + }, + { + "epoch": 0.3520512074483561, + "flow_matching_loss": 0.01640492444857955, + "grad_norm": 0.3518412113189697, + "learning_rate": 1.2090000000000001e-05, + "loss": 0.0896, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028518204391002656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016198486369103194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07323489636182785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999005913734436, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1210, + "weighted_flow_matching_loss": 0.01640492444857955, + "weighted_rkd_loss": 0.07316209487617016 + }, + { + "epoch": 0.35496072155949954, + "flow_matching_loss": 0.016688000410795212, + "grad_norm": 0.3728281557559967, + "learning_rate": 1.219e-05, + "loss": 0.0912, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029159228317439556, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016227651853114367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07454610764980316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998989337682724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1220, + "weighted_flow_matching_loss": 0.016688000410795212, + "weighted_rkd_loss": 0.07447076290845871 + }, + { + "epoch": 0.357870235670643, + "flow_matching_loss": 0.015873149130493404, + "grad_norm": 0.3857617676258087, + "learning_rate": 1.2290000000000001e-05, + "loss": 0.088, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02824133150279522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01567573444917798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07215839847922326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989726364612579, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1230, + "weighted_flow_matching_loss": 0.015873149130493404, + "weighted_rkd_loss": 0.07208427935838699 + }, + { + "epoch": 0.36077974978178645, + "flow_matching_loss": 0.015527943056076765, + "grad_norm": 0.3957199156284332, + "learning_rate": 1.239e-05, + "loss": 0.0892, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028957341238856314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015856235194951294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07377091720700264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989557921886444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1240, + "weighted_flow_matching_loss": 0.015527943056076765, + "weighted_rkd_loss": 0.07369389235973359 + }, + { + "epoch": 0.3636892638929299, + "flow_matching_loss": 0.015271489322185517, + "grad_norm": 0.39577335119247437, + "learning_rate": 1.249e-05, + "loss": 0.0869, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02814809437841177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015400848165154458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.071697036921978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989388108253479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1250, + "weighted_flow_matching_loss": 0.015271489322185517, + "weighted_rkd_loss": 0.07162095680832863 + }, + { + "epoch": 0.3665987780040733, + "flow_matching_loss": 0.014960747957229615, + "grad_norm": 0.34822478890419006, + "learning_rate": 1.2590000000000001e-05, + "loss": 0.0824, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026483613811433316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01458760593086481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06755483336746693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989216983318329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1260, + "weighted_flow_matching_loss": 0.014960747957229615, + "weighted_rkd_loss": 0.06748199798166751 + }, + { + "epoch": 0.36950829211521674, + "flow_matching_loss": 0.017399974074214695, + "grad_norm": 0.41794389486312866, + "learning_rate": 1.269e-05, + "loss": 0.0888, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028006809204816817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015472510457038879, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07148612961173058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989044368267059, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1270, + "weighted_flow_matching_loss": 0.017399974074214695, + "weighted_rkd_loss": 0.07140781134366989 + }, + { + "epoch": 0.37241780622636017, + "flow_matching_loss": 0.016034725215286018, + "grad_norm": 0.4286845922470093, + "learning_rate": 1.2790000000000001e-05, + "loss": 0.0848, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02684537637978792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015141901094466448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06883265413343906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988870561122895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1280, + "weighted_flow_matching_loss": 0.016034725215286018, + "weighted_rkd_loss": 0.06875602640211582 + }, + { + "epoch": 0.37532732033750366, + "flow_matching_loss": 0.016771531105041503, + "grad_norm": 0.38728028535842896, + "learning_rate": 1.289e-05, + "loss": 0.0846, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0267026886343956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014459406770765782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0678647831082344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988695383071899, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1290, + "weighted_flow_matching_loss": 0.016771531105041503, + "weighted_rkd_loss": 0.06778806000947953 + }, + { + "epoch": 0.3782368344486471, + "flow_matching_loss": 0.015718570444732904, + "grad_norm": 0.39787113666534424, + "learning_rate": 1.2990000000000001e-05, + "loss": 0.0839, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026716973260045053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014845553599298001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06827950067818164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988518714904785, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1300, + "weighted_flow_matching_loss": 0.015718570444732904, + "weighted_rkd_loss": 0.06820110939443111 + }, + { + "epoch": 0.3811463485597905, + "flow_matching_loss": 0.016414857748895884, + "grad_norm": 0.4158303737640381, + "learning_rate": 1.309e-05, + "loss": 0.0858, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027164106257259846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015182842966169119, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06951105520129204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988340735435486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1310, + "weighted_flow_matching_loss": 0.016414857748895884, + "weighted_rkd_loss": 0.06943001970648766 + }, + { + "epoch": 0.38405586267093395, + "flow_matching_loss": 0.0147714177146554, + "grad_norm": 0.465976357460022, + "learning_rate": 1.3189999999999999e-05, + "loss": 0.0847, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027212218008935453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015568569675087929, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06999300569295883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988161325454712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1320, + "weighted_flow_matching_loss": 0.0147714177146554, + "weighted_rkd_loss": 0.06991014182567597 + }, + { + "epoch": 0.3869653767820774, + "flow_matching_loss": 0.01703065391629934, + "grad_norm": 0.3892403542995453, + "learning_rate": 1.329e-05, + "loss": 0.0844, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026501698419451714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014480963256210089, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06748436056077481, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987980604171753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1330, + "weighted_flow_matching_loss": 0.01703065391629934, + "weighted_rkd_loss": 0.06740325130522251 + }, + { + "epoch": 0.3898748908932208, + "flow_matching_loss": 0.01366970855742693, + "grad_norm": 0.43484216928482056, + "learning_rate": 1.339e-05, + "loss": 0.0838, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027352802641689777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015533743612468242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07023934945464134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987798511981965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1340, + "weighted_flow_matching_loss": 0.01366970855742693, + "weighted_rkd_loss": 0.0701536439359188 + }, + { + "epoch": 0.3927844050043643, + "flow_matching_loss": 0.016617665346711875, + "grad_norm": 0.4298429489135742, + "learning_rate": 1.349e-05, + "loss": 0.0878, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027697544172406198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015859872475266455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07125496156513692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987614989280701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1350, + "weighted_flow_matching_loss": 0.016617665346711875, + "weighted_rkd_loss": 0.0711667187511921 + }, + { + "epoch": 0.3956939191155077, + "flow_matching_loss": 0.015051871817559005, + "grad_norm": 0.39757540822029114, + "learning_rate": 1.359e-05, + "loss": 0.0823, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026318621635437012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014655143208801746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06729238666594028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987430155277253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1360, + "weighted_flow_matching_loss": 0.015051871817559005, + "weighted_rkd_loss": 0.06720780730247497 + }, + { + "epoch": 0.39860343322665115, + "flow_matching_loss": 0.016774856392294167, + "grad_norm": 0.40159061551094055, + "learning_rate": 1.369e-05, + "loss": 0.0799, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024804387614130972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013588076271116734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06319685205817223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987244009971619, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1370, + "weighted_flow_matching_loss": 0.016774856392294167, + "weighted_rkd_loss": 0.06311622709035873 + }, + { + "epoch": 0.4015129473377946, + "flow_matching_loss": 0.015098648704588413, + "grad_norm": 0.40766724944114685, + "learning_rate": 1.379e-05, + "loss": 0.0801, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025731538794934748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013616334088146687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06507941223680973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987056434154511, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1380, + "weighted_flow_matching_loss": 0.015098648704588413, + "weighted_rkd_loss": 0.06499517261981964 + }, + { + "epoch": 0.404422461448938, + "flow_matching_loss": 0.01550383698195219, + "grad_norm": 0.4111372232437134, + "learning_rate": 1.389e-05, + "loss": 0.0807, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02531243786215782, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014640749059617519, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06526562571525574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986867487430573, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1390, + "weighted_flow_matching_loss": 0.01550383698195219, + "weighted_rkd_loss": 0.06517991945147514 + }, + { + "epoch": 0.4073319755600815, + "flow_matching_loss": 0.01761183189228177, + "grad_norm": 0.38645532727241516, + "learning_rate": 1.399e-05, + "loss": 0.0816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025249662436544897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013618274591863155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06411759853363037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986677169799805, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1400, + "weighted_flow_matching_loss": 0.01761183189228177, + "weighted_rkd_loss": 0.06403216943144799 + }, + { + "epoch": 0.4102414896712249, + "flow_matching_loss": 0.016414745803922416, + "grad_norm": 0.3625677227973938, + "learning_rate": 1.409e-05, + "loss": 0.0823, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025737934187054633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01446183081716299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06593769900500775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986485421657563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1410, + "weighted_flow_matching_loss": 0.016414745803922416, + "weighted_rkd_loss": 0.06584857925772666 + }, + { + "epoch": 0.41315100378236835, + "flow_matching_loss": 0.014745130855590104, + "grad_norm": 0.41453802585601807, + "learning_rate": 1.419e-05, + "loss": 0.0784, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025217971950769424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013322088774293661, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06375803239643574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998629242181778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1420, + "weighted_flow_matching_loss": 0.014745130855590104, + "weighted_rkd_loss": 0.06367063820362091 + }, + { + "epoch": 0.4160605178935118, + "flow_matching_loss": 0.017717666551470758, + "grad_norm": 0.35513371229171753, + "learning_rate": 1.429e-05, + "loss": 0.0789, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024061903730034827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01314023919403553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061264047399163246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986097991466523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1430, + "weighted_flow_matching_loss": 0.017717666551470758, + "weighted_rkd_loss": 0.061178889498114584 + }, + { + "epoch": 0.4189700320046552, + "flow_matching_loss": 0.016720334999263287, + "grad_norm": 0.39700791239738464, + "learning_rate": 1.4390000000000001e-05, + "loss": 0.0799, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024956375174224377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013340453431010247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06325320415198803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985902011394501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1440, + "weighted_flow_matching_loss": 0.016720334999263287, + "weighted_rkd_loss": 0.06316402480006218 + }, + { + "epoch": 0.42187954611579864, + "flow_matching_loss": 0.01570017458871007, + "grad_norm": 0.3554668426513672, + "learning_rate": 1.449e-05, + "loss": 0.0754, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02367484327405691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012436124309897423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05978581085801125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985705077648163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1450, + "weighted_flow_matching_loss": 0.01570017458871007, + "weighted_rkd_loss": 0.0597003597766161 + }, + { + "epoch": 0.42478906022694213, + "flow_matching_loss": 0.014634272456169129, + "grad_norm": 0.36682265996932983, + "learning_rate": 1.4590000000000001e-05, + "loss": 0.0782, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024942419677972793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013724571838974953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06360941082239151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985506594181061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1460, + "weighted_flow_matching_loss": 0.014634272456169129, + "weighted_rkd_loss": 0.06351722292602062 + }, + { + "epoch": 0.42769857433808556, + "flow_matching_loss": 0.015417718049138784, + "grad_norm": 0.35425472259521484, + "learning_rate": 1.469e-05, + "loss": 0.0756, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0239930285140872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01233079405501485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06031685173511505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985306680202484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1470, + "weighted_flow_matching_loss": 0.015417718049138784, + "weighted_rkd_loss": 0.060228240117430686 + }, + { + "epoch": 0.430608088449229, + "flow_matching_loss": 0.014390330575406552, + "grad_norm": 0.3597845435142517, + "learning_rate": 1.479e-05, + "loss": 0.0755, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024256918765604495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012651440687477589, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061165278777480125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985105395317078, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1480, + "weighted_flow_matching_loss": 0.014390330575406552, + "weighted_rkd_loss": 0.06107417270541191 + }, + { + "epoch": 0.4335176025603724, + "flow_matching_loss": 0.015459821559488773, + "grad_norm": 0.3741139769554138, + "learning_rate": 1.4890000000000001e-05, + "loss": 0.075, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023552429117262363, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012499928940087557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05960478782653809, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998490285873413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1490, + "weighted_flow_matching_loss": 0.015459821559488773, + "weighted_rkd_loss": 0.05951480753719807 + }, + { + "epoch": 0.43642711667151585, + "flow_matching_loss": 0.015037015732377767, + "grad_norm": 0.38964352011680603, + "learning_rate": 1.499e-05, + "loss": 0.0752, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0237407261505723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0127963250502944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06027777791023255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998469889163971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1500, + "weighted_flow_matching_loss": 0.015037015732377767, + "weighted_rkd_loss": 0.06018553487956524 + }, + { + "epoch": 0.4393366307826593, + "flow_matching_loss": 0.016153406724333762, + "grad_norm": 0.37791576981544495, + "learning_rate": 1.5090000000000001e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023653713054955006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012885043676942588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060192469507455826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984493494033814, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1510, + "weighted_flow_matching_loss": 0.016153406724333762, + "weighted_rkd_loss": 0.0600991390645504 + }, + { + "epoch": 0.44224614489380276, + "flow_matching_loss": 0.01727958181872964, + "grad_norm": 0.4603315591812134, + "learning_rate": 1.519e-05, + "loss": 0.0764, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02320409957319498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012773923948407174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05918212346732617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984286844730377, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1520, + "weighted_flow_matching_loss": 0.01727958181872964, + "weighted_rkd_loss": 0.05908914655447006 + }, + { + "epoch": 0.4451556590049462, + "flow_matching_loss": 0.016051847580820323, + "grad_norm": 0.3737494647502899, + "learning_rate": 1.529e-05, + "loss": 0.0762, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023545168340206146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013163582980632782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06025392040610313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984078705310822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1530, + "weighted_flow_matching_loss": 0.016051847580820323, + "weighted_rkd_loss": 0.06015798486769199 + }, + { + "epoch": 0.4480651731160896, + "flow_matching_loss": 0.016402185987681152, + "grad_norm": 0.4282234311103821, + "learning_rate": 1.539e-05, + "loss": 0.0762, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023596237972378732, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01272859973832965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05992107577621937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983869314193725, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1540, + "weighted_flow_matching_loss": 0.016402185987681152, + "weighted_rkd_loss": 0.05982442423701286 + }, + { + "epoch": 0.45097468722723305, + "flow_matching_loss": 0.015393358189612627, + "grad_norm": 0.385597825050354, + "learning_rate": 1.549e-05, + "loss": 0.0732, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022880104556679725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01210059430450201, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057860803604125974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983658492565155, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1550, + "weighted_flow_matching_loss": 0.015393358189612627, + "weighted_rkd_loss": 0.057766251266002655 + }, + { + "epoch": 0.4538842013383765, + "flow_matching_loss": 0.014598778914660215, + "grad_norm": 0.3853124678134918, + "learning_rate": 1.559e-05, + "loss": 0.0711, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02241024523973465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011748540122061968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05656903050839901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998344624042511, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1560, + "weighted_flow_matching_loss": 0.014598778914660215, + "weighted_rkd_loss": 0.0564753957092762 + }, + { + "epoch": 0.4567937154495199, + "flow_matching_loss": 0.015267981775105, + "grad_norm": 0.42268839478492737, + "learning_rate": 1.569e-05, + "loss": 0.0709, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022006320394575597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011740200873464345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05575284138321877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983232617378235, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1570, + "weighted_flow_matching_loss": 0.015267981775105, + "weighted_rkd_loss": 0.05565935745835304 + }, + { + "epoch": 0.4597032295606634, + "flow_matching_loss": 0.013712641224265099, + "grad_norm": 0.44339194893836975, + "learning_rate": 1.579e-05, + "loss": 0.072, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022834908217191696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012692888081073762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05836270414292812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983017861843109, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1580, + "weighted_flow_matching_loss": 0.013712641224265099, + "weighted_rkd_loss": 0.058263587206602095 + }, + { + "epoch": 0.4626127436718068, + "flow_matching_loss": 0.015493222884833813, + "grad_norm": 0.3640845715999603, + "learning_rate": 1.589e-05, + "loss": 0.0724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022424258664250372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012186278775334358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057034796103835106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982801616191864, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1590, + "weighted_flow_matching_loss": 0.015493222884833813, + "weighted_rkd_loss": 0.05693670436739921 + }, + { + "epoch": 0.46552225778295026, + "flow_matching_loss": 0.016484576836228372, + "grad_norm": 0.3803039789199829, + "learning_rate": 1.599e-05, + "loss": 0.0767, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023553106375038622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013187975995242596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06029418855905533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982583880424499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1600, + "weighted_flow_matching_loss": 0.016484576836228372, + "weighted_rkd_loss": 0.06018917560577393 + }, + { + "epoch": 0.4684317718940937, + "flow_matching_loss": 0.01703231716528535, + "grad_norm": 0.4118577241897583, + "learning_rate": 1.609e-05, + "loss": 0.0762, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02348604016005993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012347964104264975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05932004451751709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982364892959594, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1610, + "weighted_flow_matching_loss": 0.01703231716528535, + "weighted_rkd_loss": 0.05921544432640076 + }, + { + "epoch": 0.4713412860052371, + "flow_matching_loss": 0.01509592728689313, + "grad_norm": 0.41369467973709106, + "learning_rate": 1.619e-05, + "loss": 0.0708, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021989327669143677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011806926131248474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05578558146953583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998214453458786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1620, + "weighted_flow_matching_loss": 0.01509592728689313, + "weighted_rkd_loss": 0.055685976147651674 + }, + { + "epoch": 0.47425080011638054, + "flow_matching_loss": 0.015673425700515508, + "grad_norm": 0.37758761644363403, + "learning_rate": 1.629e-05, + "loss": 0.0685, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02079226616770029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011367014236748218, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052951546385884285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981922805309296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1630, + "weighted_flow_matching_loss": 0.015673425700515508, + "weighted_rkd_loss": 0.05285583138465881 + }, + { + "epoch": 0.47716031422752403, + "flow_matching_loss": 0.016205679439008237, + "grad_norm": 0.412687212228775, + "learning_rate": 1.639e-05, + "loss": 0.0727, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022457952983677386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011643229983747005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0565591350197792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981699764728547, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1640, + "weighted_flow_matching_loss": 0.016205679439008237, + "weighted_rkd_loss": 0.05645563453435898 + }, + { + "epoch": 0.48006982833866746, + "flow_matching_loss": 0.01637302925810218, + "grad_norm": 0.35259392857551575, + "learning_rate": 1.649e-05, + "loss": 0.0696, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02107181362807751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011201562359929085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0533451896160841, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981475174427032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1650, + "weighted_flow_matching_loss": 0.01637302925810218, + "weighted_rkd_loss": 0.053246365115046504 + }, + { + "epoch": 0.4829793424498109, + "flow_matching_loss": 0.01372239412739873, + "grad_norm": 0.382150799036026, + "learning_rate": 1.6590000000000002e-05, + "loss": 0.0664, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02098140213638544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010813282802700996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05277608707547188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981249570846558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1660, + "weighted_flow_matching_loss": 0.01372239412739873, + "weighted_rkd_loss": 0.05267713107168674 + }, + { + "epoch": 0.4858888565609543, + "flow_matching_loss": 0.01567007349804044, + "grad_norm": 0.43044430017471313, + "learning_rate": 1.669e-05, + "loss": 0.0696, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021399904601275922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011199195589870214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053999004885554314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998102217912674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1670, + "weighted_flow_matching_loss": 0.01567007349804044, + "weighted_rkd_loss": 0.05389650873839855 + }, + { + "epoch": 0.48879837067209775, + "flow_matching_loss": 0.014369886741042137, + "grad_norm": 0.34207186102867126, + "learning_rate": 1.679e-05, + "loss": 0.0671, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020891633257269858, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011062660161405802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052845926210284236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980793595314026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1680, + "weighted_flow_matching_loss": 0.014369886741042137, + "weighted_rkd_loss": 0.05274443328380585 + }, + { + "epoch": 0.4917078847832412, + "flow_matching_loss": 0.017558909580111505, + "grad_norm": 0.3587521016597748, + "learning_rate": 1.689e-05, + "loss": 0.0727, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022035041265189646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011192205641418696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05526228845119476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980563700199128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1690, + "weighted_flow_matching_loss": 0.017558909580111505, + "weighted_rkd_loss": 0.05515486449003219 + }, + { + "epoch": 0.49461739889438466, + "flow_matching_loss": 0.01654331823810935, + "grad_norm": 0.38730770349502563, + "learning_rate": 1.699e-05, + "loss": 0.0707, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021504408307373525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01121688038110733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054225696995854375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980332434177399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1700, + "weighted_flow_matching_loss": 0.01654331823810935, + "weighted_rkd_loss": 0.05411904715001583 + }, + { + "epoch": 0.4975269130055281, + "flow_matching_loss": 0.015763928834348916, + "grad_norm": 0.3271329402923584, + "learning_rate": 1.709e-05, + "loss": 0.0646, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019421930238604546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01010506395250559, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04894892424345017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998009979724884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1710, + "weighted_flow_matching_loss": 0.015763928834348916, + "weighted_rkd_loss": 0.04885152503848076 + }, + { + "epoch": 0.5004364271166715, + "flow_matching_loss": 0.015003425627946853, + "grad_norm": 0.36657002568244934, + "learning_rate": 1.719e-05, + "loss": 0.0682, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021312795206904413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0106994547881186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0533250454813242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979865670204162, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1720, + "weighted_flow_matching_loss": 0.015003425627946853, + "weighted_rkd_loss": 0.05321766808629036 + }, + { + "epoch": 0.503345941227815, + "flow_matching_loss": 0.016155014187097548, + "grad_norm": 0.36162859201431274, + "learning_rate": 1.7290000000000002e-05, + "loss": 0.066, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019844240881502627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010288487188518047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04997696839272976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979630291461945, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1730, + "weighted_flow_matching_loss": 0.016155014187097548, + "weighted_rkd_loss": 0.049875174462795255 + }, + { + "epoch": 0.5062554553389584, + "flow_matching_loss": 0.01431829184293747, + "grad_norm": 0.37035590410232544, + "learning_rate": 1.739e-05, + "loss": 0.0667, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020662140659987928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011209748405963182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052534029632806775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979393601417541, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1740, + "weighted_flow_matching_loss": 0.01431829184293747, + "weighted_rkd_loss": 0.052425767108798024 + }, + { + "epoch": 0.5091649694501018, + "flow_matching_loss": 0.016259791050106286, + "grad_norm": 0.4470047056674957, + "learning_rate": 1.749e-05, + "loss": 0.0687, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0207964476197958, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01095120357349515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05254409834742546, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979155480861663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1750, + "weighted_flow_matching_loss": 0.016259791050106286, + "weighted_rkd_loss": 0.052434572204947474 + }, + { + "epoch": 0.5120744835612453, + "flow_matching_loss": 0.01568042328581214, + "grad_norm": 0.3467431664466858, + "learning_rate": 1.7590000000000003e-05, + "loss": 0.0662, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020112792775034905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010435351915657521, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050660938024520874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978915929794312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1760, + "weighted_flow_matching_loss": 0.01568042328581214, + "weighted_rkd_loss": 0.0505541305989027 + }, + { + "epoch": 0.5149839976723887, + "flow_matching_loss": 0.01707179294899106, + "grad_norm": 0.37078797817230225, + "learning_rate": 1.7690000000000002e-05, + "loss": 0.0692, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02085592206567526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010519121401011943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052230966091156, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978675127029419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1770, + "weighted_flow_matching_loss": 0.01707179294899106, + "weighted_rkd_loss": 0.05211957842111588 + }, + { + "epoch": 0.5178935117835322, + "flow_matching_loss": 0.014705556165426969, + "grad_norm": 0.3904695510864258, + "learning_rate": 1.779e-05, + "loss": 0.0655, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02010110691189766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010665483027696609, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05086769685149193, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978432893753052, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1780, + "weighted_flow_matching_loss": 0.014705556165426969, + "weighted_rkd_loss": 0.050757965072989464 + }, + { + "epoch": 0.5208030258946756, + "flow_matching_loss": 0.014815268572419882, + "grad_norm": 0.37523970007896423, + "learning_rate": 1.7890000000000003e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02015482299029827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010363227128982544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050672873854637146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978189170360565, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1790, + "weighted_flow_matching_loss": 0.014815268572419882, + "weighted_rkd_loss": 0.05056234709918499 + }, + { + "epoch": 0.523712540005819, + "flow_matching_loss": 0.016376248840242625, + "grad_norm": 0.42351609468460083, + "learning_rate": 1.7990000000000002e-05, + "loss": 0.065, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019367977604269983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00996522679924965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04870118163526058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977944314479827, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1800, + "weighted_flow_matching_loss": 0.016376248840242625, + "weighted_rkd_loss": 0.04859377630054951 + }, + { + "epoch": 0.5266220541169625, + "flow_matching_loss": 0.015321314707398415, + "grad_norm": 0.37691155076026917, + "learning_rate": 1.809e-05, + "loss": 0.0646, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019585336558520794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010245303343981504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04941597692668438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977697968482971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1810, + "weighted_flow_matching_loss": 0.015321314707398415, + "weighted_rkd_loss": 0.0493057556450367 + }, + { + "epoch": 0.5295315682281059, + "flow_matching_loss": 0.015324690192937852, + "grad_norm": 0.3702077567577362, + "learning_rate": 1.819e-05, + "loss": 0.0647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019683145359158515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01013564458116889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04950193539261818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977450370788574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1820, + "weighted_flow_matching_loss": 0.015324690192937852, + "weighted_rkd_loss": 0.04939030148088932 + }, + { + "epoch": 0.5324410823392494, + "flow_matching_loss": 0.016452391073107718, + "grad_norm": 0.34870827198028564, + "learning_rate": 1.8290000000000003e-05, + "loss": 0.0642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019083899073302747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009736311901360751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04790411032736301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977201223373413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1830, + "weighted_flow_matching_loss": 0.016452391073107718, + "weighted_rkd_loss": 0.04779488742351532 + }, + { + "epoch": 0.5353505964503927, + "flow_matching_loss": 0.016141571383923293, + "grad_norm": 0.3354225754737854, + "learning_rate": 1.8390000000000002e-05, + "loss": 0.0658, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019701102934777736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01038189372047782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04978409968316555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976950824260712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1840, + "weighted_flow_matching_loss": 0.016141571383923293, + "weighted_rkd_loss": 0.04966934509575367 + }, + { + "epoch": 0.5382601105615362, + "flow_matching_loss": 0.01530659394338727, + "grad_norm": 0.3667350113391876, + "learning_rate": 1.8489999999999997e-05, + "loss": 0.065, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01967826336622238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01040984895080328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049766375496983525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997669917345047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1850, + "weighted_flow_matching_loss": 0.01530659394338727, + "weighted_rkd_loss": 0.04965040758252144 + }, + { + "epoch": 0.5411696246726797, + "flow_matching_loss": 0.01584753505885601, + "grad_norm": 0.34890857338905334, + "learning_rate": 1.859e-05, + "loss": 0.0641, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019158812053501607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010007586795836687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04832521080970764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997644591331482, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1860, + "weighted_flow_matching_loss": 0.01584753505885601, + "weighted_rkd_loss": 0.04821139425039291 + }, + { + "epoch": 0.5440791387838231, + "flow_matching_loss": 0.014614031556993723, + "grad_norm": 0.35384854674339294, + "learning_rate": 1.869e-05, + "loss": 0.0652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020146389305591584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010434519127011299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0507272981107235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976191461086273, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1870, + "weighted_flow_matching_loss": 0.014614031556993723, + "weighted_rkd_loss": 0.050606519728899 + }, + { + "epoch": 0.5469886528949666, + "flow_matching_loss": 0.015785959921777247, + "grad_norm": 0.3518941104412079, + "learning_rate": 1.8789999999999998e-05, + "loss": 0.0617, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01830156408250332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009398241620510816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046001370251178744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975935518741608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1880, + "weighted_flow_matching_loss": 0.015785959921777247, + "weighted_rkd_loss": 0.045890673995018005 + }, + { + "epoch": 0.5498981670061099, + "flow_matching_loss": 0.014509180188179016, + "grad_norm": 0.39430829882621765, + "learning_rate": 1.889e-05, + "loss": 0.0621, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018822086788713933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010087048448622227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04773122221231461, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975678384304046, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1890, + "weighted_flow_matching_loss": 0.014509180188179016, + "weighted_rkd_loss": 0.04761511720716953 + }, + { + "epoch": 0.5528076811172534, + "flow_matching_loss": 0.01590345399454236, + "grad_norm": 0.38754162192344666, + "learning_rate": 1.899e-05, + "loss": 0.0611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017892777733504773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009562286268919706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04534784108400345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975419819355011, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1900, + "weighted_flow_matching_loss": 0.01590345399454236, + "weighted_rkd_loss": 0.045236366987228396 + }, + { + "epoch": 0.5557171952283969, + "flow_matching_loss": 0.014926782716065646, + "grad_norm": 0.3634357750415802, + "learning_rate": 1.909e-05, + "loss": 0.0587, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01743811648339033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009037833288311958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04391406662762165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99751598238945, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1910, + "weighted_flow_matching_loss": 0.014926782716065646, + "weighted_rkd_loss": 0.04380496144294739 + }, + { + "epoch": 0.5586267093395403, + "flow_matching_loss": 0.016189378593116997, + "grad_norm": 0.34900227189064026, + "learning_rate": 1.919e-05, + "loss": 0.0601, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017493308614939453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009017580840736627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04400419816374779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974898576736451, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1920, + "weighted_flow_matching_loss": 0.016189378593116997, + "weighted_rkd_loss": 0.04389376230537891 + }, + { + "epoch": 0.5615362234506838, + "flow_matching_loss": 0.01563925500959158, + "grad_norm": 0.3301798105239868, + "learning_rate": 1.929e-05, + "loss": 0.0609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018123341724276543, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009143259795382618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045389942824840546, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997463583946228, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1930, + "weighted_flow_matching_loss": 0.01563925500959158, + "weighted_rkd_loss": 0.04527482464909553 + }, + { + "epoch": 0.5644457375618271, + "flow_matching_loss": 0.015264737699180842, + "grad_norm": 0.3147616684436798, + "learning_rate": 1.939e-05, + "loss": 0.0604, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018162711709737777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008882400859147311, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045207824930548667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974371731281281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1940, + "weighted_flow_matching_loss": 0.015264737699180842, + "weighted_rkd_loss": 0.0450919684022665 + }, + { + "epoch": 0.5673552516729706, + "flow_matching_loss": 0.016963245533406734, + "grad_norm": 0.34298962354660034, + "learning_rate": 1.9489999999999998e-05, + "loss": 0.0635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01848691198974848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009715326782315969, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04668915160000324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974106311798095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1950, + "weighted_flow_matching_loss": 0.016963245533406734, + "weighted_rkd_loss": 0.046568240597844125 + }, + { + "epoch": 0.570264765784114, + "flow_matching_loss": 0.019057933986186982, + "grad_norm": 0.3809748888015747, + "learning_rate": 1.959e-05, + "loss": 0.0626, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017320067249238492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008978266129270196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04361840039491653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973839581012726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1960, + "weighted_flow_matching_loss": 0.019057933986186982, + "weighted_rkd_loss": 0.043504276499152186 + }, + { + "epoch": 0.5731742798952575, + "flow_matching_loss": 0.0165269591845572, + "grad_norm": 0.3433297574520111, + "learning_rate": 1.969e-05, + "loss": 0.0597, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017304834723472596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008689400181174279, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0432990700006485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973571419715881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1970, + "weighted_flow_matching_loss": 0.0165269591845572, + "weighted_rkd_loss": 0.04318464510142803 + }, + { + "epoch": 0.576083794006401, + "flow_matching_loss": 0.014499860722571612, + "grad_norm": 0.37725961208343506, + "learning_rate": 1.979e-05, + "loss": 0.06, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01824506651610136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009124229522421956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04561436213552952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973301887512207, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1980, + "weighted_flow_matching_loss": 0.014499860722571612, + "weighted_rkd_loss": 0.04549257308244705 + }, + { + "epoch": 0.5789933081175443, + "flow_matching_loss": 0.013517383113503456, + "grad_norm": 0.3413808345794678, + "learning_rate": 1.989e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017394158709794282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008714408660307527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04350272603332996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973030984401703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1990, + "weighted_flow_matching_loss": 0.013517383113503456, + "weighted_rkd_loss": 0.04338539354503155 + }, + { + "epoch": 0.5819028222286878, + "flow_matching_loss": 0.015042653679847718, + "grad_norm": 0.34982067346572876, + "learning_rate": 1.999e-05, + "loss": 0.0608, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018328010570257902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00926920836791396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04592522941529751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9972758829593659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2000, + "weighted_flow_matching_loss": 0.015042653679847718, + "weighted_rkd_loss": 0.0458001222461462 + }, + { + "epoch": 0.5848123363398312, + "flow_matching_loss": 0.01437370441854, + "grad_norm": 0.32186874747276306, + "learning_rate": 2.009e-05, + "loss": 0.0596, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018016928154975177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009357173647731542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045391029492020604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9972485303878784, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2010, + "weighted_flow_matching_loss": 0.01437370441854, + "weighted_rkd_loss": 0.04526614584028721 + }, + { + "epoch": 0.5877218504509747, + "flow_matching_loss": 0.01540403189137578, + "grad_norm": 0.3549467921257019, + "learning_rate": 2.019e-05, + "loss": 0.0608, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018205059692263603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00912020429968834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04553032405674458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997221028804779, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2020, + "weighted_flow_matching_loss": 0.01540403189137578, + "weighted_rkd_loss": 0.045403793826699256 + }, + { + "epoch": 0.5906313645621182, + "flow_matching_loss": 0.014903429057449102, + "grad_norm": 0.34401488304138184, + "learning_rate": 2.029e-05, + "loss": 0.0584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017468627728521823, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008679836103692651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04361709095537662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971933960914612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2030, + "weighted_flow_matching_loss": 0.014903429057449102, + "weighted_rkd_loss": 0.04349470064043999 + }, + { + "epoch": 0.5935408786732616, + "flow_matching_loss": 0.0164266855455935, + "grad_norm": 0.339686781167984, + "learning_rate": 2.039e-05, + "loss": 0.0617, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01809462606906891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009250270994380116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045439523085951804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971656262874603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2040, + "weighted_flow_matching_loss": 0.0164266855455935, + "weighted_rkd_loss": 0.04531071782112121 + }, + { + "epoch": 0.596450392784405, + "flow_matching_loss": 0.01591276554390788, + "grad_norm": 0.3391713798046112, + "learning_rate": 2.0490000000000002e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01861813124269247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009465453308075666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04670171551406384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997137725353241, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2050, + "weighted_flow_matching_loss": 0.01591276554390788, + "weighted_rkd_loss": 0.04656803347170353 + }, + { + "epoch": 0.5993599068955484, + "flow_matching_loss": 0.014564005751162767, + "grad_norm": 0.34811875224113464, + "learning_rate": 2.059e-05, + "loss": 0.059, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017704728990793228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009161585848778487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04457104280591011, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971096754074097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2060, + "weighted_flow_matching_loss": 0.014564005751162767, + "weighted_rkd_loss": 0.04444222599267959 + }, + { + "epoch": 0.6022694210066919, + "flow_matching_loss": 0.01433127224445343, + "grad_norm": 0.37648460268974304, + "learning_rate": 2.069e-05, + "loss": 0.0598, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018054882995784283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009459910169243812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04556967690587044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970815062522889, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2070, + "weighted_flow_matching_loss": 0.01433127224445343, + "weighted_rkd_loss": 0.04543667733669281 + }, + { + "epoch": 0.6051789351178353, + "flow_matching_loss": 0.013801776431500911, + "grad_norm": 0.3476460576057434, + "learning_rate": 2.079e-05, + "loss": 0.0575, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017538138944655658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00877137533389032, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04384765289723873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997053188085556, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2080, + "weighted_flow_matching_loss": 0.013801776431500911, + "weighted_rkd_loss": 0.043718422576785085 + }, + { + "epoch": 0.6080884492289788, + "flow_matching_loss": 0.01607928639277816, + "grad_norm": 0.3365230858325958, + "learning_rate": 2.089e-05, + "loss": 0.0612, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017915977258235218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00937380949035287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04520576409995556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970247447490692, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2090, + "weighted_flow_matching_loss": 0.01607928639277816, + "weighted_rkd_loss": 0.0450712576508522 + }, + { + "epoch": 0.6109979633401222, + "flow_matching_loss": 0.015171894803643227, + "grad_norm": 0.3684842884540558, + "learning_rate": 2.099e-05, + "loss": 0.0613, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018215122167021036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009808867936953902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046239112690091136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969961524009705, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2100, + "weighted_flow_matching_loss": 0.015171894803643227, + "weighted_rkd_loss": 0.04610023461282253 + }, + { + "epoch": 0.6139074774512656, + "flow_matching_loss": 0.015657520107924938, + "grad_norm": 0.4272646903991699, + "learning_rate": 2.109e-05, + "loss": 0.0616, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018266001157462596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009515086468309165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04604708887636662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969674289226532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2110, + "weighted_flow_matching_loss": 0.015657520107924938, + "weighted_rkd_loss": 0.04590743854641914 + }, + { + "epoch": 0.6168169915624091, + "flow_matching_loss": 0.015717709064483644, + "grad_norm": 0.29024621844291687, + "learning_rate": 2.1190000000000002e-05, + "loss": 0.0581, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01678026942536235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008932371251285076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042492910847067834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969385743141175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2120, + "weighted_flow_matching_loss": 0.015717709064483644, + "weighted_rkd_loss": 0.04236283712089062 + }, + { + "epoch": 0.6197265056735525, + "flow_matching_loss": 0.014762227423489094, + "grad_norm": 0.3228585124015808, + "learning_rate": 2.129e-05, + "loss": 0.0584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017465665843337773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008853450883179903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04378478266298771, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969095766544342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2130, + "weighted_flow_matching_loss": 0.014762227423489094, + "weighted_rkd_loss": 0.04364947527647019 + }, + { + "epoch": 0.622636019784696, + "flow_matching_loss": 0.015444719977676869, + "grad_norm": 0.4058065116405487, + "learning_rate": 2.139e-05, + "loss": 0.0585, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017181070055812597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008822827553376555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04318496696650982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996880441904068, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2140, + "weighted_flow_matching_loss": 0.015444719977676869, + "weighted_rkd_loss": 0.04305023849010468 + }, + { + "epoch": 0.6255455338958394, + "flow_matching_loss": 0.015314830653369426, + "grad_norm": 0.39167824387550354, + "learning_rate": 2.1490000000000003e-05, + "loss": 0.0583, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017028451524674894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00903561869636178, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043092521652579305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9968511760234833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2150, + "weighted_flow_matching_loss": 0.015314830653369426, + "weighted_rkd_loss": 0.042956819757819176 + }, + { + "epoch": 0.6284550480069828, + "flow_matching_loss": 0.01679164469242096, + "grad_norm": 0.35311049222946167, + "learning_rate": 2.159e-05, + "loss": 0.0578, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016281194426119326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00854224651120603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04110463485121727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9968217790126801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2160, + "weighted_flow_matching_loss": 0.01679164469242096, + "weighted_rkd_loss": 0.040973993763327596 + }, + { + "epoch": 0.6313645621181263, + "flow_matching_loss": 0.016490891762077808, + "grad_norm": 0.31128835678100586, + "learning_rate": 2.169e-05, + "loss": 0.0574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01623095227405429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008583107637241483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041045012697577474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967922329902649, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2170, + "weighted_flow_matching_loss": 0.016490891762077808, + "weighted_rkd_loss": 0.04091335721313953 + }, + { + "epoch": 0.6342740762292697, + "flow_matching_loss": 0.015830063633620738, + "grad_norm": 0.3804088830947876, + "learning_rate": 2.1790000000000003e-05, + "loss": 0.061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018053152970969676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009180188132449984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045286494120955464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967625617980957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2180, + "weighted_flow_matching_loss": 0.015830063633620738, + "weighted_rkd_loss": 0.045139885321259496 + }, + { + "epoch": 0.6371835903404132, + "flow_matching_loss": 0.01512301629409194, + "grad_norm": 0.3489106595516205, + "learning_rate": 2.1890000000000002e-05, + "loss": 0.0568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01669208109378815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008452936448156834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041837099194526675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996732747554779, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2190, + "weighted_flow_matching_loss": 0.01512301629409194, + "weighted_rkd_loss": 0.04170044101774693 + }, + { + "epoch": 0.6400931044515565, + "flow_matching_loss": 0.01335493689402938, + "grad_norm": 0.34609687328338623, + "learning_rate": 2.199e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015897339675575495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00787006001919508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03966473937034607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967027962207794, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2200, + "weighted_flow_matching_loss": 0.01335493689402938, + "weighted_rkd_loss": 0.039533966407179835 + }, + { + "epoch": 0.6430026185627, + "flow_matching_loss": 0.01582076232880354, + "grad_norm": 0.3690781593322754, + "learning_rate": 2.209e-05, + "loss": 0.0579, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016767273284494878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008698679134249687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042233226075768474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966727137565613, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2210, + "weighted_flow_matching_loss": 0.01582076232880354, + "weighted_rkd_loss": 0.04209269657731056 + }, + { + "epoch": 0.6459121326738435, + "flow_matching_loss": 0.01505490392446518, + "grad_norm": 0.3424728810787201, + "learning_rate": 2.2190000000000003e-05, + "loss": 0.0567, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01671166904270649, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00832042470574379, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04174376241862774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966424882411957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2220, + "weighted_flow_matching_loss": 0.01505490392446518, + "weighted_rkd_loss": 0.04160362109541893 + }, + { + "epoch": 0.6488216467849869, + "flow_matching_loss": 0.013389377575367688, + "grad_norm": 0.31703492999076843, + "learning_rate": 2.2290000000000002e-05, + "loss": 0.0568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017402235977351667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008768392913043498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04357286468148232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966121435165405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2230, + "weighted_flow_matching_loss": 0.013389377575367688, + "weighted_rkd_loss": 0.04342524074018002 + }, + { + "epoch": 0.6517311608961304, + "flow_matching_loss": 0.015105330757796765, + "grad_norm": 0.34054893255233765, + "learning_rate": 2.239e-05, + "loss": 0.0567, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016695270966738462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008363069407641888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04175361096858978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996581643819809, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2240, + "weighted_flow_matching_loss": 0.015105330757796765, + "weighted_rkd_loss": 0.04161087535321713 + }, + { + "epoch": 0.6546406750072737, + "flow_matching_loss": 0.014983541890978813, + "grad_norm": 0.3823171555995941, + "learning_rate": 2.2490000000000003e-05, + "loss": 0.0583, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017291141487658025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00890432121232152, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04348660446703434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9965510070323944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2250, + "weighted_flow_matching_loss": 0.014983541890978813, + "weighted_rkd_loss": 0.04333660416305065 + }, + { + "epoch": 0.6575501891184172, + "flow_matching_loss": 0.016892097890377045, + "grad_norm": 0.3458651006221771, + "learning_rate": 2.2590000000000002e-05, + "loss": 0.0604, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01722193732857704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009199184831231833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04364305958151817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9965202391147614, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2260, + "weighted_flow_matching_loss": 0.016892097890377045, + "weighted_rkd_loss": 0.04349120520055294 + }, + { + "epoch": 0.6604597032295607, + "flow_matching_loss": 0.01623787796124816, + "grad_norm": 0.37273189425468445, + "learning_rate": 2.269e-05, + "loss": 0.0587, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01695653423666954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008686094591394068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0425991628319025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964893460273743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2270, + "weighted_flow_matching_loss": 0.01623787796124816, + "weighted_rkd_loss": 0.042449602484703065 + }, + { + "epoch": 0.6633692173407041, + "flow_matching_loss": 0.015588636603206396, + "grad_norm": 0.3290416896343231, + "learning_rate": 2.279e-05, + "loss": 0.0553, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01605245964601636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007701475033536553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03980639427900314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964583098888398, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2280, + "weighted_flow_matching_loss": 0.015588636603206396, + "weighted_rkd_loss": 0.039665408059954645 + }, + { + "epoch": 0.6662787314518476, + "flow_matching_loss": 0.014758547209203244, + "grad_norm": 0.3098236620426178, + "learning_rate": 2.289e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015852730441838502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008083392539992929, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03978885300457478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964271366596222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2290, + "weighted_flow_matching_loss": 0.014758547209203244, + "weighted_rkd_loss": 0.03964669518172741 + }, + { + "epoch": 0.669188245562991, + "flow_matching_loss": 0.016344337444752454, + "grad_norm": 0.3222261369228363, + "learning_rate": 2.299e-05, + "loss": 0.0575, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016395217273384333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008510356256738304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04130079112946987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963958263397217, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2300, + "weighted_flow_matching_loss": 0.016344337444752454, + "weighted_rkd_loss": 0.04115194268524647 + }, + { + "epoch": 0.6720977596741344, + "flow_matching_loss": 0.016511339973658324, + "grad_norm": 0.32285553216934204, + "learning_rate": 2.309e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01563662486150861, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007938203494995832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039211453124880793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963643729686738, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2310, + "weighted_flow_matching_loss": 0.016511339973658324, + "weighted_rkd_loss": 0.039068923145532605 + }, + { + "epoch": 0.6750072737852778, + "flow_matching_loss": 0.01479148892685771, + "grad_norm": 0.4231078326702118, + "learning_rate": 2.319e-05, + "loss": 0.055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016008960828185082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008345703314989805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04036362543702125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963327884674072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2320, + "weighted_flow_matching_loss": 0.01479148892685771, + "weighted_rkd_loss": 0.04021560475230217 + }, + { + "epoch": 0.6779167878964213, + "flow_matching_loss": 0.015578951966017485, + "grad_norm": 0.3070625066757202, + "learning_rate": 2.329e-05, + "loss": 0.0552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015986025147140027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0077914300840348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03976348079741001, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963010728359223, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2330, + "weighted_flow_matching_loss": 0.015578951966017485, + "weighted_rkd_loss": 0.03961640223860741 + }, + { + "epoch": 0.6808263020075648, + "flow_matching_loss": 0.015248175151646138, + "grad_norm": 0.32599547505378723, + "learning_rate": 2.3389999999999998e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015042029973119497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007448763586580753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037532823719084264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962692201137543, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2340, + "weighted_flow_matching_loss": 0.015248175151646138, + "weighted_rkd_loss": 0.03739278316497803 + }, + { + "epoch": 0.6837358161187082, + "flow_matching_loss": 0.01674783546477556, + "grad_norm": 0.38390064239501953, + "learning_rate": 2.349e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016032136138528587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008191214874386787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04025548696517944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962372303009033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2350, + "weighted_flow_matching_loss": 0.01674783546477556, + "weighted_rkd_loss": 0.04010402634739876 + }, + { + "epoch": 0.6866453302298516, + "flow_matching_loss": 0.01728355335071683, + "grad_norm": 0.37142297625541687, + "learning_rate": 2.359e-05, + "loss": 0.0616, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017701013013720512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009082433301955462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044484459608793256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962050914764404, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2360, + "weighted_flow_matching_loss": 0.01728355335071683, + "weighted_rkd_loss": 0.044315633550286296 + }, + { + "epoch": 0.689554844340995, + "flow_matching_loss": 0.016242070030421017, + "grad_norm": 0.4527985453605652, + "learning_rate": 2.369e-05, + "loss": 0.0539, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014947023335844278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007944193668663502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03783824034035206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961728453636169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2370, + "weighted_flow_matching_loss": 0.016242070030421017, + "weighted_rkd_loss": 0.037693418189883234 + }, + { + "epoch": 0.6924643584521385, + "flow_matching_loss": 0.016078397165983915, + "grad_norm": 0.3445203900337219, + "learning_rate": 2.379e-05, + "loss": 0.0553, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01576598584651947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00782151403836906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03935348615050316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961404442787171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2380, + "weighted_flow_matching_loss": 0.016078397165983915, + "weighted_rkd_loss": 0.03920160382986069 + }, + { + "epoch": 0.695373872563282, + "flow_matching_loss": 0.015121183451265096, + "grad_norm": 0.35277751088142395, + "learning_rate": 2.389e-05, + "loss": 0.0563, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016601926926523448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008111320529133081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041315175220370294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961079120635986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2390, + "weighted_flow_matching_loss": 0.015121183451265096, + "weighted_rkd_loss": 0.04115435481071472 + }, + { + "epoch": 0.6982833866744254, + "flow_matching_loss": 0.015879561193287373, + "grad_norm": 0.3108234107494354, + "learning_rate": 2.399e-05, + "loss": 0.0554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015892994683235885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007926209038123489, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03971219845116138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960752367973328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2400, + "weighted_flow_matching_loss": 0.015879561193287373, + "weighted_rkd_loss": 0.03955634571611881 + }, + { + "epoch": 0.7011929007855688, + "flow_matching_loss": 0.014977172948420048, + "grad_norm": 0.37587350606918335, + "learning_rate": 2.409e-05, + "loss": 0.0548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01591930529102683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008097504684701561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0399361152201891, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960424363613128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2410, + "weighted_flow_matching_loss": 0.014977172948420048, + "weighted_rkd_loss": 0.03977806530892849 + }, + { + "epoch": 0.7041024148967122, + "flow_matching_loss": 0.015259722713381052, + "grad_norm": 0.33943790197372437, + "learning_rate": 2.419e-05, + "loss": 0.0533, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015286954492330551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007591468561440706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03816537782549858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960094869136811, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2420, + "weighted_flow_matching_loss": 0.015259722713381052, + "weighted_rkd_loss": 0.03801308758556843 + }, + { + "epoch": 0.7070119290078557, + "flow_matching_loss": 0.017972799111157657, + "grad_norm": 0.3424619138240814, + "learning_rate": 2.429e-05, + "loss": 0.0578, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01591960173100233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008146555908024312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039985759556293486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959764122962952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2430, + "weighted_flow_matching_loss": 0.017972799111157657, + "weighted_rkd_loss": 0.03982486724853516 + }, + { + "epoch": 0.7099214431189991, + "flow_matching_loss": 0.01666923752054572, + "grad_norm": 0.30686748027801514, + "learning_rate": 2.439e-05, + "loss": 0.0563, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01591216940432787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007983538368716837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03980787731707096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959432005882263, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2440, + "weighted_flow_matching_loss": 0.01666923752054572, + "weighted_rkd_loss": 0.0396464079618454 + }, + { + "epoch": 0.7128309572301426, + "flow_matching_loss": 0.016680462379008533, + "grad_norm": 0.3651812970638275, + "learning_rate": 2.449e-05, + "loss": 0.056, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015836496837437153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007782263634726405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03945525735616684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959098517894744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2450, + "weighted_flow_matching_loss": 0.016680462379008533, + "weighted_rkd_loss": 0.03929388672113419 + }, + { + "epoch": 0.715740471341286, + "flow_matching_loss": 0.017509388737380505, + "grad_norm": 0.32039061188697815, + "learning_rate": 2.459e-05, + "loss": 0.0576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016139380168169737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008018943713977933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04029770381748676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958763599395752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2460, + "weighted_flow_matching_loss": 0.017509388737380505, + "weighted_rkd_loss": 0.04013151489198208 + }, + { + "epoch": 0.7186499854524294, + "flow_matching_loss": 0.01630025142803788, + "grad_norm": 0.32619962096214294, + "learning_rate": 2.469e-05, + "loss": 0.055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015544740948826075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00781084643676877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03890032842755318, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958427429199219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2470, + "weighted_flow_matching_loss": 0.01630025142803788, + "weighted_rkd_loss": 0.0387386042624712 + }, + { + "epoch": 0.7215594995635729, + "flow_matching_loss": 0.016267802752554416, + "grad_norm": 0.3505100905895233, + "learning_rate": 2.479e-05, + "loss": 0.0558, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015917046647518873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007873925706371665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03970801904797554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958089888095856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2480, + "weighted_flow_matching_loss": 0.016267802752554416, + "weighted_rkd_loss": 0.0395416259765625 + }, + { + "epoch": 0.7244690136747163, + "flow_matching_loss": 0.01491163270547986, + "grad_norm": 0.3098830580711365, + "learning_rate": 2.489e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014778550341725349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007403499213978648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03696059975773096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957750737667084, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2490, + "weighted_flow_matching_loss": 0.01491163270547986, + "weighted_rkd_loss": 0.036804436892271045 + }, + { + "epoch": 0.7273785277858598, + "flow_matching_loss": 0.015133097395300865, + "grad_norm": 0.30527520179748535, + "learning_rate": 2.499e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015415399800986052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0075528342742472885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038383633643388745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957410573959351, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2500, + "weighted_flow_matching_loss": 0.015133097395300865, + "weighted_rkd_loss": 0.03822019211947918 + }, + { + "epoch": 0.7302880418970032, + "flow_matching_loss": 0.01521907290443778, + "grad_norm": 0.436470091342926, + "learning_rate": 2.5090000000000002e-05, + "loss": 0.0525, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015004417952150106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007416924322023987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03742576017975807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957068979740142, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2510, + "weighted_flow_matching_loss": 0.01521907290443778, + "weighted_rkd_loss": 0.03726507611572742 + }, + { + "epoch": 0.7331975560081466, + "flow_matching_loss": 0.01540538826957345, + "grad_norm": 0.3270501494407654, + "learning_rate": 2.519e-05, + "loss": 0.054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015520140063017607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007710387976840138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03875066824257374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956725835800171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2520, + "weighted_flow_matching_loss": 0.01540538826957345, + "weighted_rkd_loss": 0.03858298249542713 + }, + { + "epoch": 0.7361070701192901, + "flow_matching_loss": 0.014663149137049913, + "grad_norm": 0.3371429443359375, + "learning_rate": 2.529e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014763431809842586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072201905772089955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03674705438315869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956381559371948, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2530, + "weighted_flow_matching_loss": 0.014663149137049913, + "weighted_rkd_loss": 0.03658678643405437 + }, + { + "epoch": 0.7390165842304335, + "flow_matching_loss": 0.016747122909873725, + "grad_norm": 0.34914639592170715, + "learning_rate": 2.5390000000000003e-05, + "loss": 0.054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014902212470769883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007622887752950192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037427313067018984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956035852432251, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2540, + "weighted_flow_matching_loss": 0.016747122909873725, + "weighted_rkd_loss": 0.037262767925858496 + }, + { + "epoch": 0.741926098341577, + "flow_matching_loss": 0.015583668276667595, + "grad_norm": 0.36614519357681274, + "learning_rate": 2.549e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015172607079148293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007845922838896513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03819113653153181, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9955688655376435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2550, + "weighted_flow_matching_loss": 0.015583668276667595, + "weighted_rkd_loss": 0.03802191000431776 + }, + { + "epoch": 0.7448356124527203, + "flow_matching_loss": 0.016033740900456905, + "grad_norm": 0.3771423399448395, + "learning_rate": 2.559e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015036659222096205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007657338865101337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03773065656423569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9955340266227722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2560, + "weighted_flow_matching_loss": 0.016033740900456905, + "weighted_rkd_loss": 0.03756213784217834 + }, + { + "epoch": 0.7477451265638638, + "flow_matching_loss": 0.014672737289220094, + "grad_norm": 0.3145878314971924, + "learning_rate": 2.569e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015382291749119759, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007729298947378993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03849388249218464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954990386962891, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2570, + "weighted_flow_matching_loss": 0.014672737289220094, + "weighted_rkd_loss": 0.03832061737775803 + }, + { + "epoch": 0.7506546406750073, + "flow_matching_loss": 0.015352742653340102, + "grad_norm": 0.276364803314209, + "learning_rate": 2.5790000000000002e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015247670002281666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007955308118835092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03845064789056778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954639375209808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2580, + "weighted_flow_matching_loss": 0.015352742653340102, + "weighted_rkd_loss": 0.0382762536406517 + }, + { + "epoch": 0.7535641547861507, + "flow_matching_loss": 0.01539937946945429, + "grad_norm": 0.30865100026130676, + "learning_rate": 2.589e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014539351779967546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007140807947143913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03621951155364513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954286694526673, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2590, + "weighted_flow_matching_loss": 0.01539937946945429, + "weighted_rkd_loss": 0.036053945496678355 + }, + { + "epoch": 0.7564736688972942, + "flow_matching_loss": 0.017900564521551133, + "grad_norm": 0.33750706911087036, + "learning_rate": 2.599e-05, + "loss": 0.0538, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014424343965947628, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007176004769280553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03602469265460968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9953932762145996, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2600, + "weighted_flow_matching_loss": 0.017900564521551133, + "weighted_rkd_loss": 0.03585873655974865 + }, + { + "epoch": 0.7593831830084375, + "flow_matching_loss": 0.014442676771432161, + "grad_norm": 0.3685067594051361, + "learning_rate": 2.6090000000000003e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014159718994051218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00724478904157877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035564227029681206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9953577578067779, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2610, + "weighted_flow_matching_loss": 0.014442676771432161, + "weighted_rkd_loss": 0.0353991225361824 + }, + { + "epoch": 0.762292697119581, + "flow_matching_loss": 0.016177260689437388, + "grad_norm": 0.3264629542827606, + "learning_rate": 2.619e-05, + "loss": 0.0532, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014767614658921957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007676657754927874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03721188679337502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9953221023082733, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2620, + "weighted_flow_matching_loss": 0.016177260689437388, + "weighted_rkd_loss": 0.03703780248761177 + }, + { + "epoch": 0.7652022112307245, + "flow_matching_loss": 0.015008798707276583, + "grad_norm": 0.36046260595321655, + "learning_rate": 2.629e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014612004347145557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0073498435318470005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036573851853609084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9952863037586213, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2630, + "weighted_flow_matching_loss": 0.015008798707276583, + "weighted_rkd_loss": 0.03640144243836403 + }, + { + "epoch": 0.7681117253418679, + "flow_matching_loss": 0.014606331288814545, + "grad_norm": 0.2963907718658447, + "learning_rate": 2.6390000000000003e-05, + "loss": 0.0521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015081328060477971, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00751155181787908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03767420761287212, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9952503621578217, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2640, + "weighted_flow_matching_loss": 0.014606331288814545, + "weighted_rkd_loss": 0.03749528639018536 + }, + { + "epoch": 0.7710212394530114, + "flow_matching_loss": 0.015497339144349098, + "grad_norm": 0.30377551913261414, + "learning_rate": 2.6490000000000002e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01483020382001996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007731536496430635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03739194460213184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.995214295387268, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2650, + "weighted_flow_matching_loss": 0.015497339144349098, + "weighted_rkd_loss": 0.03721303381025791 + }, + { + "epoch": 0.7739307535641547, + "flow_matching_loss": 0.015802678000181914, + "grad_norm": 0.3319263756275177, + "learning_rate": 2.659e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014910346735268832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007219915371388197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03704060837626457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951781034469604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2660, + "weighted_flow_matching_loss": 0.015802678000181914, + "weighted_rkd_loss": 0.03686200045049191 + }, + { + "epoch": 0.7768402676752982, + "flow_matching_loss": 0.015405059047043323, + "grad_norm": 0.30923551321029663, + "learning_rate": 2.6690000000000004e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015134790074080228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007239931495860219, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037509511597454546, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951417624950409, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2670, + "weighted_flow_matching_loss": 0.015405059047043323, + "weighted_rkd_loss": 0.037327281944453714 + }, + { + "epoch": 0.7797497817864416, + "flow_matching_loss": 0.016378914844244717, + "grad_norm": 0.32510754466056824, + "learning_rate": 2.6790000000000003e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014478760492056608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007142682792618871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03610020391643047, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951052844524384, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2680, + "weighted_flow_matching_loss": 0.016378914844244717, + "weighted_rkd_loss": 0.035923524759709835 + }, + { + "epoch": 0.7826592958975851, + "flow_matching_loss": 0.013964326214045286, + "grad_norm": 0.3278246819972992, + "learning_rate": 2.6890000000000002e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014478574506938457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00749409212730825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03645124211907387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9950686693191528, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2690, + "weighted_flow_matching_loss": 0.013964326214045286, + "weighted_rkd_loss": 0.0362714871764183 + }, + { + "epoch": 0.7855688100087286, + "flow_matching_loss": 0.015597615763545037, + "grad_norm": 0.2889263331890106, + "learning_rate": 2.699e-05, + "loss": 0.052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014640600327402354, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007307762606069445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036588963493704794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9950319230556488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2700, + "weighted_flow_matching_loss": 0.015597615763545037, + "weighted_rkd_loss": 0.03640717789530754 + }, + { + "epoch": 0.788478324119872, + "flow_matching_loss": 0.014907386992126704, + "grad_norm": 0.290340393781662, + "learning_rate": 2.709e-05, + "loss": 0.0547, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0159353107213974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008098050532862545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03996867183595896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9949950575828552, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2710, + "weighted_flow_matching_loss": 0.014907386992126704, + "weighted_rkd_loss": 0.039768640883266926 + }, + { + "epoch": 0.7913878382310154, + "flow_matching_loss": 0.016928056441247463, + "grad_norm": 0.299978643655777, + "learning_rate": 2.719e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014783071726560593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007256605289876461, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0368227481842041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9949580252170562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2720, + "weighted_flow_matching_loss": 0.016928056441247463, + "weighted_rkd_loss": 0.036637074500322345 + }, + { + "epoch": 0.7942973523421588, + "flow_matching_loss": 0.01472033690661192, + "grad_norm": 0.3139698803424835, + "learning_rate": 2.7289999999999998e-05, + "loss": 0.0525, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01529951198026538, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007358191255480051, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037957214564085004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9949208736419678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2730, + "weighted_flow_matching_loss": 0.01472033690661192, + "weighted_rkd_loss": 0.037764431536197664 + }, + { + "epoch": 0.7972068664533023, + "flow_matching_loss": 0.01687188632786274, + "grad_norm": 0.3111835718154907, + "learning_rate": 2.739e-05, + "loss": 0.0548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015377077925950288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007396018225699663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0381501741707325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948835849761963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2740, + "weighted_flow_matching_loss": 0.01687188632786274, + "weighted_rkd_loss": 0.03795496821403503 + }, + { + "epoch": 0.8001163805644458, + "flow_matching_loss": 0.01745677748695016, + "grad_norm": 0.3036944270133972, + "learning_rate": 2.749e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014079061057418584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007241903292015195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03540002536028623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948461651802063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2750, + "weighted_flow_matching_loss": 0.01745677748695016, + "weighted_rkd_loss": 0.03521757256239653 + }, + { + "epoch": 0.8030258946755892, + "flow_matching_loss": 0.01741602420806885, + "grad_norm": 0.35732501745224, + "learning_rate": 2.759e-05, + "loss": 0.0546, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014986283518373967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007440864061936736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03741343170404434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948085963726043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2760, + "weighted_flow_matching_loss": 0.01741602420806885, + "weighted_rkd_loss": 0.037219198793172835 + }, + { + "epoch": 0.8059354087867326, + "flow_matching_loss": 0.018091184552758932, + "grad_norm": 0.33368656039237976, + "learning_rate": 2.769e-05, + "loss": 0.0548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014730813913047314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007412001118063926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036873629130423066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9947708964347839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2770, + "weighted_flow_matching_loss": 0.018091184552758932, + "weighted_rkd_loss": 0.03668082095682621 + }, + { + "epoch": 0.808844922897876, + "flow_matching_loss": 0.016734916437417268, + "grad_norm": 0.3211786150932312, + "learning_rate": 2.779e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014161328226327896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007444669678807259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03576732631772757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9947330713272095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2780, + "weighted_flow_matching_loss": 0.016734916437417268, + "weighted_rkd_loss": 0.035578937456011775 + }, + { + "epoch": 0.8117544370090195, + "flow_matching_loss": 0.015111323539167642, + "grad_norm": 0.30808237195014954, + "learning_rate": 2.789e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015115795377641916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0076826723292469975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03791426327079535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946951031684875, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2790, + "weighted_flow_matching_loss": 0.015111323539167642, + "weighted_rkd_loss": 0.037713102623820306 + }, + { + "epoch": 0.814663951120163, + "flow_matching_loss": 0.016267956886440515, + "grad_norm": 0.3324669301509857, + "learning_rate": 2.799e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014733114745467901, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007337960181757808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03680419027805328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946569800376892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2800, + "weighted_flow_matching_loss": 0.016267956886440515, + "weighted_rkd_loss": 0.0366075437515974 + }, + { + "epoch": 0.8175734652313064, + "flow_matching_loss": 0.014789675455540418, + "grad_norm": 0.296677827835083, + "learning_rate": 2.809e-05, + "loss": 0.049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013724477682262658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00691966968588531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034368624538183214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946187555789947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2810, + "weighted_flow_matching_loss": 0.014789675455540418, + "weighted_rkd_loss": 0.034183675423264503 + }, + { + "epoch": 0.8204829793424498, + "flow_matching_loss": 0.015572454128414393, + "grad_norm": 0.3080546259880066, + "learning_rate": 2.819e-05, + "loss": 0.0521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014657732471823693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007453169533982873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03676863461732864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9945803880691528, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2820, + "weighted_flow_matching_loss": 0.015572454128414393, + "weighted_rkd_loss": 0.036569371446967126 + }, + { + "epoch": 0.8233924934535932, + "flow_matching_loss": 0.016775162890553474, + "grad_norm": 0.2727871835231781, + "learning_rate": 2.829e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014393807761371136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007337388396263122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036125003919005395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994541871547699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2830, + "weighted_flow_matching_loss": 0.016775162890553474, + "weighted_rkd_loss": 0.03592783696949482 + }, + { + "epoch": 0.8263020075647367, + "flow_matching_loss": 0.016247391514480115, + "grad_norm": 0.3328983783721924, + "learning_rate": 2.839e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01447893436998129, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007247776910662651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03620564565062523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9945032298564911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2840, + "weighted_flow_matching_loss": 0.016247391514480115, + "weighted_rkd_loss": 0.03600666895508766 + }, + { + "epoch": 0.8292115216758801, + "flow_matching_loss": 0.015727230161428452, + "grad_norm": 0.2920207381248474, + "learning_rate": 2.849e-05, + "loss": 0.0507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014073178451508284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006995284091681242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03514164127409458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9944644510746002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2850, + "weighted_flow_matching_loss": 0.015727230161428452, + "weighted_rkd_loss": 0.0349471103399992 + }, + { + "epoch": 0.8321210357870236, + "flow_matching_loss": 0.01529264273121953, + "grad_norm": 0.4001481235027313, + "learning_rate": 2.859e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015011704433709383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0076805319171398875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037703941017389296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9944255232810975, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2860, + "weighted_flow_matching_loss": 0.01529264273121953, + "weighted_rkd_loss": 0.03749377634376287 + }, + { + "epoch": 0.835030549898167, + "flow_matching_loss": 0.01611219048500061, + "grad_norm": 0.3332190215587616, + "learning_rate": 2.869e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014287740923464298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0073320422321558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03590752333402634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994386476278305, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2870, + "weighted_flow_matching_loss": 0.01611219048500061, + "weighted_rkd_loss": 0.035705959796905516 + }, + { + "epoch": 0.8379400640093104, + "flow_matching_loss": 0.014700972568243742, + "grad_norm": 0.3827730119228363, + "learning_rate": 2.879e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013838628306984901, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00692046289332211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03459771946072578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9943472921848298, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2880, + "weighted_flow_matching_loss": 0.014700972568243742, + "weighted_rkd_loss": 0.03440216928720474 + }, + { + "epoch": 0.8408495781204539, + "flow_matching_loss": 0.015028050728142261, + "grad_norm": 0.29881325364112854, + "learning_rate": 2.889e-05, + "loss": 0.0508, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01442569401115179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071014455985277895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03595283403992653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9943079710006714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2890, + "weighted_flow_matching_loss": 0.015028050728142261, + "weighted_rkd_loss": 0.035748188942670824 + }, + { + "epoch": 0.8437590922315973, + "flow_matching_loss": 0.015434383507817983, + "grad_norm": 0.30381155014038086, + "learning_rate": 2.8990000000000002e-05, + "loss": 0.0508, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014170419797301293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007211080426350236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03555191997438669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9942685067653656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2900, + "weighted_flow_matching_loss": 0.015434383507817983, + "weighted_rkd_loss": 0.03534811586141586 + }, + { + "epoch": 0.8466686063427408, + "flow_matching_loss": 0.014478703215718269, + "grad_norm": 0.3953309655189514, + "learning_rate": 2.909e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01542584365233779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007690225588157773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03854191284626722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9942289173603058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2910, + "weighted_flow_matching_loss": 0.014478703215718269, + "weighted_rkd_loss": 0.03831947017461061 + }, + { + "epoch": 0.8495781204538843, + "flow_matching_loss": 0.01585750374943018, + "grad_norm": 0.3602532148361206, + "learning_rate": 2.919e-05, + "loss": 0.0521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014569425582885742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007305801473557949, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03644465245306492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9941891849040985, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2920, + "weighted_flow_matching_loss": 0.01585750374943018, + "weighted_rkd_loss": 0.036232877895236014 + }, + { + "epoch": 0.8524876345650276, + "flow_matching_loss": 0.016147565096616745, + "grad_norm": 0.40261751413345337, + "learning_rate": 2.9290000000000002e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014094305597245694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007376518798992038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035565130040049554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9941492974758148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2930, + "weighted_flow_matching_loss": 0.016147565096616745, + "weighted_rkd_loss": 0.035357053950428964 + }, + { + "epoch": 0.8553971486761711, + "flow_matching_loss": 0.017102697864174843, + "grad_norm": 0.37816962599754333, + "learning_rate": 2.939e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014869253803044557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007823720620945097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03756222799420357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9941093146800994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2940, + "weighted_flow_matching_loss": 0.017102697864174843, + "weighted_rkd_loss": 0.037340952455997466 + }, + { + "epoch": 0.8583066627873145, + "flow_matching_loss": 0.015777156595140694, + "grad_norm": 0.2836054861545563, + "learning_rate": 2.949e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013940590340644122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0066762899979949, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0345574701204896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9940691828727722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2950, + "weighted_flow_matching_loss": 0.015777156595140694, + "weighted_rkd_loss": 0.034352504462003705 + }, + { + "epoch": 0.861216176898458, + "flow_matching_loss": 0.01588933253660798, + "grad_norm": 0.37992754578590393, + "learning_rate": 2.959e-05, + "loss": 0.0521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014528977125883103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007325655175372958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036383610032498834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994028913974762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2960, + "weighted_flow_matching_loss": 0.01588933253660798, + "weighted_rkd_loss": 0.036166335083544256 + }, + { + "epoch": 0.8641256910096013, + "flow_matching_loss": 0.016699770279228688, + "grad_norm": 0.3525565564632416, + "learning_rate": 2.9690000000000002e-05, + "loss": 0.0537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014873004518449306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007464209804311395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037210218235850336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9939885020256043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2970, + "weighted_flow_matching_loss": 0.016699770279228688, + "weighted_rkd_loss": 0.036986560001969336 + }, + { + "epoch": 0.8670352051207448, + "flow_matching_loss": 0.01615284979343414, + "grad_norm": 0.3315102458000183, + "learning_rate": 2.979e-05, + "loss": 0.0524, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01461018631234765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007279913639649749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036500286497175696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993947958946228, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2980, + "weighted_flow_matching_loss": 0.01615284979343414, + "weighted_rkd_loss": 0.03627936691045761 + }, + { + "epoch": 0.8699447192318883, + "flow_matching_loss": 0.014951193798333406, + "grad_norm": 0.3244689404964447, + "learning_rate": 2.989e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013732380140572787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006978637026622892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03444339763373137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9939072787761688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2990, + "weighted_flow_matching_loss": 0.014951193798333406, + "weighted_rkd_loss": 0.03423354141414166 + }, + { + "epoch": 0.8728542333430317, + "flow_matching_loss": 0.015651259291917085, + "grad_norm": 0.30256569385528564, + "learning_rate": 2.9990000000000003e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013765433523803949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006705554062500596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03423642106354237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9938664674758911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3000, + "weighted_flow_matching_loss": 0.015651259291917085, + "weighted_rkd_loss": 0.03402642831206322 + }, + { + "epoch": 0.8757637474541752, + "flow_matching_loss": 0.018271446321159603, + "grad_norm": 0.2870028614997864, + "learning_rate": 2.9999998154575393e-05, + "loss": 0.0532, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014064775966107845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007009228272363543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035138780064880845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993825513124466, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3010, + "weighted_flow_matching_loss": 0.018271446321159603, + "weighted_rkd_loss": 0.03492181040346622 + }, + { + "epoch": 0.8786732615653186, + "flow_matching_loss": 0.01595303537324071, + "grad_norm": 0.3205980658531189, + "learning_rate": 2.999999177533042e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014206629898399114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006888303859159351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03530156314373016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9937844395637512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3020, + "weighted_flow_matching_loss": 0.01595303537324071, + "weighted_rkd_loss": 0.035082141123712066 + }, + { + "epoch": 0.881582775676462, + "flow_matching_loss": 0.01689468724653125, + "grad_norm": 0.3011910915374756, + "learning_rate": 2.9999980839483992e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014066769927740096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006764378445222974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03489791825413704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9937432110309601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3030, + "weighted_flow_matching_loss": 0.01689468724653125, + "weighted_rkd_loss": 0.034679586067795753 + }, + { + "epoch": 0.8844922897876055, + "flow_matching_loss": 0.015593399200588465, + "grad_norm": 0.2877105474472046, + "learning_rate": 2.999996534703944e-05, + "loss": 0.0533, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0150573811493814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007806125376373529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037920888140797616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9937018573284149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3040, + "weighted_flow_matching_loss": 0.015593399200588465, + "weighted_rkd_loss": 0.03768204972147941 + }, + { + "epoch": 0.8874018038987489, + "flow_matching_loss": 0.014762616995722056, + "grad_norm": 0.2894708216190338, + "learning_rate": 2.9999945298001468e-05, + "loss": 0.0485, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013531046360731125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006928990222513675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03399108275771141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9936603724956512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3050, + "weighted_flow_matching_loss": 0.014762616995722056, + "weighted_rkd_loss": 0.03377558588981629 + }, + { + "epoch": 0.8903113180098924, + "flow_matching_loss": 0.015059400349855423, + "grad_norm": 0.2905955910682678, + "learning_rate": 2.9999920692376165e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013922777306288481, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006939216842874884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03478477094322443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9936187267303467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3060, + "weighted_flow_matching_loss": 0.015059400349855423, + "weighted_rkd_loss": 0.03456282280385494 + }, + { + "epoch": 0.8932208321210358, + "flow_matching_loss": 0.015474431216716766, + "grad_norm": 0.32862773537635803, + "learning_rate": 2.9999891530171002e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013761948328465223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006912786979228258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0344366829842329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993576979637146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3070, + "weighted_flow_matching_loss": 0.015474431216716766, + "weighted_rkd_loss": 0.03421550281345844 + }, + { + "epoch": 0.8961303462321792, + "flow_matching_loss": 0.01585593419149518, + "grad_norm": 0.36130982637405396, + "learning_rate": 2.9999857811394845e-05, + "loss": 0.0507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013841297756880522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007430551340803504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03511314745992422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9935350835323333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3080, + "weighted_flow_matching_loss": 0.01585593419149518, + "weighted_rkd_loss": 0.03488612100481987 + }, + { + "epoch": 0.8990398603433226, + "flow_matching_loss": 0.015462717972695827, + "grad_norm": 0.3521775007247925, + "learning_rate": 2.9999819536057935e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015283480565994979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007768219010904432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038335179910063744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9934930562973022, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3090, + "weighted_flow_matching_loss": 0.015462717972695827, + "weighted_rkd_loss": 0.03808571752160787 + }, + { + "epoch": 0.9019493744544661, + "flow_matching_loss": 0.014201299380511045, + "grad_norm": 0.3022342622280121, + "learning_rate": 2.9999776704171894e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013270333502441644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007055801944807172, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03359646871685982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9934508860111236, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3100, + "weighted_flow_matching_loss": 0.014201299380511045, + "weighted_rkd_loss": 0.03337644971907139 + }, + { + "epoch": 0.9048588885656096, + "flow_matching_loss": 0.015557982679456472, + "grad_norm": 0.3328924775123596, + "learning_rate": 2.999972931574973e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01431443216279149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007548606162890792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03617746997624636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9934085786342621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3110, + "weighted_flow_matching_loss": 0.015557982679456472, + "weighted_rkd_loss": 0.03593901917338371 + }, + { + "epoch": 0.907768402676753, + "flow_matching_loss": 0.015011092461645604, + "grad_norm": 0.34180182218551636, + "learning_rate": 2.9999677370805858e-05, + "loss": 0.0507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01423235647380352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0074518958106637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03591660857200622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993366140127182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3120, + "weighted_flow_matching_loss": 0.015011092461645604, + "weighted_rkd_loss": 0.03567835185676813 + }, + { + "epoch": 0.9106779167878964, + "flow_matching_loss": 0.014878144022077322, + "grad_norm": 0.27301543951034546, + "learning_rate": 2.9999620869356034e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013062659557908774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006546902284026146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03267222177237272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993323564529419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3130, + "weighted_flow_matching_loss": 0.014878144022077322, + "weighted_rkd_loss": 0.03245408311486244 + }, + { + "epoch": 0.9135874308990398, + "flow_matching_loss": 0.016469755489379168, + "grad_norm": 0.3242233693599701, + "learning_rate": 2.9999559811417436e-05, + "loss": 0.0494, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013298099860548973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006562678841874003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033158878423273565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9932808578014374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3140, + "weighted_flow_matching_loss": 0.016469755489379168, + "weighted_rkd_loss": 0.03293605800718069 + }, + { + "epoch": 0.9164969450101833, + "flow_matching_loss": 0.015050750784575939, + "grad_norm": 0.301935076713562, + "learning_rate": 2.999949419700861e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012949543725699186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006538991909474134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0324380787089467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9932380318641663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3150, + "weighted_flow_matching_loss": 0.015050750784575939, + "weighted_rkd_loss": 0.03221873063594103 + }, + { + "epoch": 0.9194064591213268, + "flow_matching_loss": 0.01569270873442292, + "grad_norm": 0.25489968061447144, + "learning_rate": 2.9999424026149484e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01378588480874896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006427053641527891, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03399882353842258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9931950390338897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3160, + "weighted_flow_matching_loss": 0.01569270873442292, + "weighted_rkd_loss": 0.03376747332513332 + }, + { + "epoch": 0.9223159732324702, + "flow_matching_loss": 0.01733054742217064, + "grad_norm": 0.27658072113990784, + "learning_rate": 2.9999349298861376e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014324584137648344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006970630632713437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03561979867517948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9931519269943238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3170, + "weighted_flow_matching_loss": 0.01733054742217064, + "weighted_rkd_loss": 0.0353758666664362 + }, + { + "epoch": 0.9252254873436137, + "flow_matching_loss": 0.015303074941039085, + "grad_norm": 0.2758451998233795, + "learning_rate": 2.9999270015166983e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013230278063565493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0067199356853961945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033180491626262666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9931086778640748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3180, + "weighted_flow_matching_loss": 0.015303074941039085, + "weighted_rkd_loss": 0.03295184094458818 + }, + { + "epoch": 0.928135001454757, + "flow_matching_loss": 0.016254966706037523, + "grad_norm": 0.29991811513900757, + "learning_rate": 2.999918617509039e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013848759327083826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006629893137142062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03432741183787584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9930652976036072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3190, + "weighted_flow_matching_loss": 0.016254966706037523, + "weighted_rkd_loss": 0.03408938068896532 + }, + { + "epoch": 0.9310445155659005, + "flow_matching_loss": 0.015291947964578867, + "grad_norm": 0.3422044813632965, + "learning_rate": 2.9999097778657074e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013891274482011795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006968850549310446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034751400165259835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9930217802524567, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3200, + "weighted_flow_matching_loss": 0.015291947964578867, + "weighted_rkd_loss": 0.034508905559778216 + }, + { + "epoch": 0.9339540296770439, + "flow_matching_loss": 0.01588517930358648, + "grad_norm": 0.3209543824195862, + "learning_rate": 2.9999004825893878e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012617228180170059, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006180873280391097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03141532950103283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9929781138896943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3210, + "weighted_flow_matching_loss": 0.01588517930358648, + "weighted_rkd_loss": 0.031194749660789965 + }, + { + "epoch": 0.9368635437881874, + "flow_matching_loss": 0.017672102246433497, + "grad_norm": 0.30598872900009155, + "learning_rate": 2.9998907316829043e-05, + "loss": 0.0507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013366899359971285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006548310443758964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033282109163701536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9929343283176422, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3220, + "weighted_flow_matching_loss": 0.017672102246433497, + "weighted_rkd_loss": 0.03304697833955288 + }, + { + "epoch": 0.9397730578993309, + "flow_matching_loss": 0.015626129508018494, + "grad_norm": 0.328184574842453, + "learning_rate": 2.999880525149219e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012833956256508827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006170598976314068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03183851148933172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9928904056549073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3230, + "weighted_flow_matching_loss": 0.015626129508018494, + "weighted_rkd_loss": 0.0316121393814683 + }, + { + "epoch": 0.9426825720104742, + "flow_matching_loss": 0.015052932593971492, + "grad_norm": 0.2641550898551941, + "learning_rate": 2.999869862991432e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013242319785058498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006482576904818416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032967216335237026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9928463578224183, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3240, + "weighted_flow_matching_loss": 0.015052932593971492, + "weighted_rkd_loss": 0.03273137360811233 + }, + { + "epoch": 0.9455920861216177, + "flow_matching_loss": 0.01642437167465687, + "grad_norm": 0.31893980503082275, + "learning_rate": 2.9998587452127822e-05, + "loss": 0.0479, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012722962815314532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006218162644654512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031664088182151315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9928021490573883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3250, + "weighted_flow_matching_loss": 0.01642437167465687, + "weighted_rkd_loss": 0.031436176225543025 + }, + { + "epoch": 0.9485016002327611, + "flow_matching_loss": 0.014904389064759016, + "grad_norm": 0.3238641917705536, + "learning_rate": 2.999847171816647e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012834237795323134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006501973886042833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03217044994235039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9927578210830689, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3260, + "weighted_flow_matching_loss": 0.014904389064759016, + "weighted_rkd_loss": 0.03193747047334909 + }, + { + "epoch": 0.9514111143439046, + "flow_matching_loss": 0.016816417500376702, + "grad_norm": 0.292336106300354, + "learning_rate": 2.999835142806543e-05, + "loss": 0.0494, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012943418137729168, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006889156950637698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03277599327266216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.992713350057602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3270, + "weighted_flow_matching_loss": 0.016816417500376702, + "weighted_rkd_loss": 0.032537194341421126 + }, + { + "epoch": 0.9543206284550481, + "flow_matching_loss": 0.017864131741225718, + "grad_norm": 0.33475181460380554, + "learning_rate": 2.9998226581861227e-05, + "loss": 0.0525, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013776203989982605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00730495979078114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03485736772418022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9926687598228454, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3280, + "weighted_flow_matching_loss": 0.017864131741225718, + "weighted_rkd_loss": 0.034601837024092676 + }, + { + "epoch": 0.9572301425661914, + "flow_matching_loss": 0.01823303159326315, + "grad_norm": 0.36418256163597107, + "learning_rate": 2.9998097179591793e-05, + "loss": 0.0537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01431244220584631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007059816503897309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0356847008690238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9926240146160126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3290, + "weighted_flow_matching_loss": 0.01823303159326315, + "weighted_rkd_loss": 0.035421447083354 + }, + { + "epoch": 0.9601396566773349, + "flow_matching_loss": 0.01633493723347783, + "grad_norm": 0.28353768587112427, + "learning_rate": 2.9997963221296443e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013775703217834234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006543573271483183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03409497942775488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9925791501998902, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3300, + "weighted_flow_matching_loss": 0.01633493723347783, + "weighted_rkd_loss": 0.03384199980646372 + }, + { + "epoch": 0.9630491707884783, + "flow_matching_loss": 0.014964774437248706, + "grad_norm": 0.3804660737514496, + "learning_rate": 2.9997824707015858e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014677056297659873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007182023860514164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03653613664209843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9925341427326202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3310, + "weighted_flow_matching_loss": 0.014964774437248706, + "weighted_rkd_loss": 0.036263360269367696 + }, + { + "epoch": 0.9659586848996218, + "flow_matching_loss": 0.01514307949692011, + "grad_norm": 0.31462016701698303, + "learning_rate": 2.9997681636792126e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012772108893841505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006196879036724568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03174109663814306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9924890041351319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3320, + "weighted_flow_matching_loss": 0.01514307949692011, + "weighted_rkd_loss": 0.03150268103927374 + }, + { + "epoch": 0.9688681990107652, + "flow_matching_loss": 0.017497471626847984, + "grad_norm": 0.328044593334198, + "learning_rate": 2.9997534010668707e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013769804686307906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0068277019076049324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034367311745882034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9924437284469605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3330, + "weighted_flow_matching_loss": 0.017497471626847984, + "weighted_rkd_loss": 0.03410760667175054 + }, + { + "epoch": 0.9717777131219086, + "flow_matching_loss": 0.01675566369667649, + "grad_norm": 0.3200617730617523, + "learning_rate": 2.9997381828690435e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013499019760638475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006818383419886232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03381642270833254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9923983097076416, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3340, + "weighted_flow_matching_loss": 0.01675566369667649, + "weighted_rkd_loss": 0.03355937115848064 + }, + { + "epoch": 0.9746872272330521, + "flow_matching_loss": 0.017077705916017295, + "grad_norm": 0.25318464636802673, + "learning_rate": 2.9997225090903552e-05, + "loss": 0.049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012738603726029396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006722087087109685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03219929486513138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9923527717590332, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3350, + "weighted_flow_matching_loss": 0.017077705916017295, + "weighted_rkd_loss": 0.03195306174457073 + }, + { + "epoch": 0.9775967413441955, + "flow_matching_loss": 0.017478240374475717, + "grad_norm": 0.27484187483787537, + "learning_rate": 2.9997063797355664e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013164785038679839, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006770618818700313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033100188709795475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9923070907592774, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3360, + "weighted_flow_matching_loss": 0.017478240374475717, + "weighted_rkd_loss": 0.03284555319696665 + }, + { + "epoch": 0.980506255455339, + "flow_matching_loss": 0.015467933844774961, + "grad_norm": 0.3461960554122925, + "learning_rate": 2.9996897948095768e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013476421963423491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006641691830009222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03359453584998846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9922612607479095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3370, + "weighted_flow_matching_loss": 0.015467933844774961, + "weighted_rkd_loss": 0.03333453014492989 + }, + { + "epoch": 0.9834157695664824, + "flow_matching_loss": 0.016628720983862876, + "grad_norm": 0.3040219247341156, + "learning_rate": 2.9996727543174244e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013252530619502068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006672873394563794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0331779345870018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9922153174877166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3380, + "weighted_flow_matching_loss": 0.016628720983862876, + "weighted_rkd_loss": 0.03291963841766119 + }, + { + "epoch": 0.9863252836776258, + "flow_matching_loss": 0.015185752511024475, + "grad_norm": 0.3099946081638336, + "learning_rate": 2.9996552582642857e-05, + "loss": 0.0499, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014076033141463995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006844558380544185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03499662522226572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9921692371368408, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3390, + "weighted_flow_matching_loss": 0.015185752511024475, + "weighted_rkd_loss": 0.03472260758280754 + }, + { + "epoch": 0.9892347977887693, + "flow_matching_loss": 0.01593189127743244, + "grad_norm": 0.2994530498981476, + "learning_rate": 2.9996373066554763e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01263839825987816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006110511114820838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031387307494878766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9921230137348175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3400, + "weighted_flow_matching_loss": 0.01593189127743244, + "weighted_rkd_loss": 0.031140111573040485 + }, + { + "epoch": 0.9921443118999127, + "flow_matching_loss": 0.016197618283331394, + "grad_norm": 0.3752201199531555, + "learning_rate": 2.999618899496448e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013025300670415163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00658504138700664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032635643519461154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9920766651630402, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3410, + "weighted_flow_matching_loss": 0.016197618283331394, + "weighted_rkd_loss": 0.03237703144550323 + }, + { + "epoch": 0.9950538260110562, + "flow_matching_loss": 0.017729467991739512, + "grad_norm": 0.28726065158843994, + "learning_rate": 2.9996000367927936e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013350828923285008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006711013242602348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03341267090290785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9920301735401154, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3420, + "weighted_flow_matching_loss": 0.017729467991739512, + "weighted_rkd_loss": 0.03314637746661901 + }, + { + "epoch": 0.9979633401221996, + "flow_matching_loss": 0.01600938579067588, + "grad_norm": 0.3349139392375946, + "learning_rate": 2.9995807185502425e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014150815457105637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0073745451401919125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03567617628723383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9919835448265075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3430, + "weighted_flow_matching_loss": 0.01600938579067588, + "weighted_rkd_loss": 0.03539015408605337 + }, + { + "epoch": 1.000872854233343, + "flow_matching_loss": 0.015453858580440284, + "grad_norm": 0.3230513334274292, + "learning_rate": 2.9995609447746636e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013709688279777765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0065496339928358795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033969010785222055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9919367849826812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3440, + "weighted_flow_matching_loss": 0.015453858580440284, + "weighted_rkd_loss": 0.033695089630782604 + }, + { + "epoch": 1.0037823683444864, + "flow_matching_loss": 0.016649214550852776, + "grad_norm": 0.30332618951797485, + "learning_rate": 2.999540715472063e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012971850018948317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006479532597586513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032423232682049274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.991889888048172, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3450, + "weighted_flow_matching_loss": 0.016649214550852776, + "weighted_rkd_loss": 0.03216025903820992 + }, + { + "epoch": 1.00669188245563, + "flow_matching_loss": 0.015603832155466079, + "grad_norm": 0.31449416279792786, + "learning_rate": 2.999520030648586e-05, + "loss": 0.0479, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013007541373372077, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006527466094121337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032542549073696136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9918428361415863, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3460, + "weighted_flow_matching_loss": 0.015603832155466079, + "weighted_rkd_loss": 0.03227711021900177 + }, + { + "epoch": 1.0096013965667734, + "flow_matching_loss": 0.01540463725104928, + "grad_norm": 0.3576584458351135, + "learning_rate": 2.9994988903105163e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013941035233438015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070622112601995465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03494428228586912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9917956948280334, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3470, + "weighted_flow_matching_loss": 0.01540463725104928, + "weighted_rkd_loss": 0.03465753737837076 + }, + { + "epoch": 1.0125109106779169, + "flow_matching_loss": 0.014891549944877625, + "grad_norm": 0.2982793152332306, + "learning_rate": 2.999477294464276e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01306690312922001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006304234731942415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03243804145604372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9917483866214752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3480, + "weighted_flow_matching_loss": 0.014891549944877625, + "weighted_rkd_loss": 0.03217037320137024 + }, + { + "epoch": 1.0154204247890601, + "flow_matching_loss": 0.014983420353382826, + "grad_norm": 0.25626054406166077, + "learning_rate": 2.9994552431164247e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012353840749710798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00592660354450345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030634284764528275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9917009532451629, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3490, + "weighted_flow_matching_loss": 0.014983420353382826, + "weighted_rkd_loss": 0.03038006220012903 + }, + { + "epoch": 1.0183299389002036, + "flow_matching_loss": 0.015589979942888021, + "grad_norm": 0.25957342982292175, + "learning_rate": 2.9994327362736617e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013116538152098655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006278548901900649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03251162562519312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9916533827781677, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3500, + "weighted_flow_matching_loss": 0.015589979942888021, + "weighted_rkd_loss": 0.03224028181284666 + }, + { + "epoch": 1.021239453011347, + "flow_matching_loss": 0.015476343780755996, + "grad_norm": 0.4470846951007843, + "learning_rate": 2.9994097739428233e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01389129888266325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006716629257425666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03449922688305378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9916056811809539, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3510, + "weighted_flow_matching_loss": 0.015476343780755996, + "weighted_rkd_loss": 0.03420960288494825 + }, + { + "epoch": 1.0241489671224906, + "flow_matching_loss": 0.015537706296890974, + "grad_norm": 0.3164004683494568, + "learning_rate": 2.9993863561308856e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012898360006511211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006372371688485146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0321690920740366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9915578424930572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3520, + "weighted_flow_matching_loss": 0.015537706296890974, + "weighted_rkd_loss": 0.03189749326556921 + }, + { + "epoch": 1.027058481233634, + "flow_matching_loss": 0.016863851994276046, + "grad_norm": 0.3188573122024536, + "learning_rate": 2.9993624828449616e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012838217057287693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00600641006603837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031682844087481496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9915098667144775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3530, + "weighted_flow_matching_loss": 0.016863851994276046, + "weighted_rkd_loss": 0.03141384776681662 + }, + { + "epoch": 1.0299679953447773, + "flow_matching_loss": 0.014310327637940646, + "grad_norm": 0.2846975326538086, + "learning_rate": 2.9993381540923037e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013417029101401568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006459559127688408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03329361788928509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9914617657661438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3540, + "weighted_flow_matching_loss": 0.014310327637940646, + "weighted_rkd_loss": 0.033009367249906066 + }, + { + "epoch": 1.0328775094559208, + "flow_matching_loss": 0.016332763712853193, + "grad_norm": 0.38914310932159424, + "learning_rate": 2.9993133698803025e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012463630735874176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006047731731086969, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030974992923438548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9914135038852692, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3550, + "weighted_flow_matching_loss": 0.016332763712853193, + "weighted_rkd_loss": 0.030708985216915608 + }, + { + "epoch": 1.0357870235670643, + "flow_matching_loss": 0.016394430305808783, + "grad_norm": 0.3016633093357086, + "learning_rate": 2.9992881302164862e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013696244545280933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006722383666783571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03411487322300673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9913651466369628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3560, + "weighted_flow_matching_loss": 0.016394430305808783, + "weighted_rkd_loss": 0.0338203052058816 + }, + { + "epoch": 1.0386965376782078, + "flow_matching_loss": 0.01518883416429162, + "grad_norm": 0.3391616642475128, + "learning_rate": 2.9992624351085226e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013158178050071, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006128526432439685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03244488313794136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9913166344165802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3570, + "weighted_flow_matching_loss": 0.01518883416429162, + "weighted_rkd_loss": 0.0321631507948041 + }, + { + "epoch": 1.0416060517893513, + "flow_matching_loss": 0.01570876147598028, + "grad_norm": 0.3371601998806, + "learning_rate": 2.9992362845642167e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012008488364517688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005656319158151746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029673296213150024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9912679851055145, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3580, + "weighted_flow_matching_loss": 0.01570876147598028, + "weighted_rkd_loss": 0.029414176754653455 + }, + { + "epoch": 1.0445155659004945, + "flow_matching_loss": 0.016197027638554573, + "grad_norm": 0.26958349347114563, + "learning_rate": 2.9992096785915132e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012588575109839439, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006168701220303774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03134585116058588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9912192106246949, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3590, + "weighted_flow_matching_loss": 0.016197027638554573, + "weighted_rkd_loss": 0.031070634350180627 + }, + { + "epoch": 1.047425080011638, + "flow_matching_loss": 0.015513250138610601, + "grad_norm": 0.3321782946586609, + "learning_rate": 2.999182617198493e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012647702638059855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062749015633016825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03157030642032623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9911702871322632, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3600, + "weighted_flow_matching_loss": 0.015513250138610601, + "weighted_rkd_loss": 0.03129154983907938 + }, + { + "epoch": 1.0503345941227815, + "flow_matching_loss": 0.016439586784690617, + "grad_norm": 0.27063673734664917, + "learning_rate": 2.999155100393378e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013171904254704714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006433915812522173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03277772422879934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9911212384700775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3610, + "weighted_flow_matching_loss": 0.016439586784690617, + "weighted_rkd_loss": 0.03248667120933533 + }, + { + "epoch": 1.053244108233925, + "flow_matching_loss": 0.015418269392102956, + "grad_norm": 0.32533231377601624, + "learning_rate": 2.9991271281845256e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011663525272160769, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005801833979785442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029128884710371495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9910720527172089, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3620, + "weighted_flow_matching_loss": 0.015418269392102956, + "weighted_rkd_loss": 0.028868814557790758 + }, + { + "epoch": 1.0561536223450685, + "flow_matching_loss": 0.017005786765366793, + "grad_norm": 0.2748161852359772, + "learning_rate": 2.9990987005804347e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012252340093255043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005987075343728066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030491754971444608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9910227239131928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3630, + "weighted_flow_matching_loss": 0.017005786765366793, + "weighted_rkd_loss": 0.030218029022216798 + }, + { + "epoch": 1.0590631364562118, + "flow_matching_loss": 0.015339843276888133, + "grad_norm": 0.3721497356891632, + "learning_rate": 2.999069817589739e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012751359306275845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006018516141921282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031521234475076196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9909732639789581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3640, + "weighted_flow_matching_loss": 0.015339843276888133, + "weighted_rkd_loss": 0.031236670352518557 + }, + { + "epoch": 1.0619726505673552, + "flow_matching_loss": 0.018697452265769243, + "grad_norm": 0.29762002825737, + "learning_rate": 2.9990404792212142e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01282310001552105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006236485065892338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031882685236632825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9909236788749695, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3650, + "weighted_flow_matching_loss": 0.018697452265769243, + "weighted_rkd_loss": 0.03159331809729338 + }, + { + "epoch": 1.0648821646784987, + "flow_matching_loss": 0.016391906049102546, + "grad_norm": 0.3289112448692322, + "learning_rate": 2.9990106854837715e-05, + "loss": 0.049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013232946675270797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006465160055086017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03293105307966471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9908739566802979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3660, + "weighted_flow_matching_loss": 0.016391906049102546, + "weighted_rkd_loss": 0.03263054732233286 + }, + { + "epoch": 1.0677916787896422, + "flow_matching_loss": 0.015229456499218941, + "grad_norm": 0.45488956570625305, + "learning_rate": 2.9989804363864615e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013132138177752495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006498193461447954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032762469910085204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9908241033554077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3670, + "weighted_flow_matching_loss": 0.015229456499218941, + "weighted_rkd_loss": 0.03246180433779955 + }, + { + "epoch": 1.0707011929007855, + "flow_matching_loss": 0.016764361318200825, + "grad_norm": 0.3497217893600464, + "learning_rate": 2.998949731938473e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013070935383439064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006565719889476895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03270759098231792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9907740950584412, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3680, + "weighted_flow_matching_loss": 0.016764361318200825, + "weighted_rkd_loss": 0.032405823655426505 + }, + { + "epoch": 1.073610707011929, + "flow_matching_loss": 0.015689233783632517, + "grad_norm": 0.3379392921924591, + "learning_rate": 2.9989185721491338e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01276415828615427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006574302818626166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03210261967033148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.990723979473114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3690, + "weighted_flow_matching_loss": 0.015689233783632517, + "weighted_rkd_loss": 0.03180484902113676 + }, + { + "epoch": 1.0765202211230724, + "flow_matching_loss": 0.01705035110935569, + "grad_norm": 0.2968686521053314, + "learning_rate": 2.998886957027909e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012142827454954385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006207308731973171, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030492963455617427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.990673714876175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3700, + "weighted_flow_matching_loss": 0.01705035110935569, + "weighted_rkd_loss": 0.030208564735949038 + }, + { + "epoch": 1.079429735234216, + "flow_matching_loss": 0.013846580032259225, + "grad_norm": 0.2982233762741089, + "learning_rate": 2.998854886584402e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011782051250338554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005673827836290002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029237930290400983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9906233131885529, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3710, + "weighted_flow_matching_loss": 0.013846580032259225, + "weighted_rkd_loss": 0.028963791951537133 + }, + { + "epoch": 1.0823392493453594, + "flow_matching_loss": 0.01528568435460329, + "grad_norm": 0.30043095350265503, + "learning_rate": 2.9988223608283557e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013119084667414427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006489028688520193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0327271981164813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9905727803707123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3720, + "weighted_flow_matching_loss": 0.01528568435460329, + "weighted_rkd_loss": 0.03241867907345295 + }, + { + "epoch": 1.0852487634565027, + "flow_matching_loss": 0.015579220093786716, + "grad_norm": 0.3321259021759033, + "learning_rate": 2.9987893797696502e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013076508417725563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006367254443466664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032520271465182306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9905221104621887, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3730, + "weighted_flow_matching_loss": 0.015579220093786716, + "weighted_rkd_loss": 0.03221207670867443 + }, + { + "epoch": 1.0881582775676462, + "flow_matching_loss": 0.014773452002555133, + "grad_norm": 0.2846035361289978, + "learning_rate": 2.9987559434183046e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01224450059235096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006124319881200791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03061332106590271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9904713213443757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3740, + "weighted_flow_matching_loss": 0.014773452002555133, + "weighted_rkd_loss": 0.03032158799469471 + }, + { + "epoch": 1.0910677916787896, + "flow_matching_loss": 0.01749791381880641, + "grad_norm": 0.2585802376270294, + "learning_rate": 2.998722051784476e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012352348398417234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006182640697807074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03088733684271574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9904203891754151, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3750, + "weighted_flow_matching_loss": 0.01749791381880641, + "weighted_rkd_loss": 0.03059146720916033 + }, + { + "epoch": 1.0939773057899331, + "flow_matching_loss": 0.0155531900934875, + "grad_norm": 0.28608429431915283, + "learning_rate": 2.9986877048784594e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0129368063993752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006872921995818615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032746535167098044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.990369313955307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3760, + "weighted_flow_matching_loss": 0.0155531900934875, + "weighted_rkd_loss": 0.03243116866797209 + }, + { + "epoch": 1.0968868199010766, + "flow_matching_loss": 0.015530622471123934, + "grad_norm": 0.26085376739501953, + "learning_rate": 2.9986529027106885e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012472053430974484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005840214202180505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030784321017563343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9903181195259094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3770, + "weighted_flow_matching_loss": 0.015530622471123934, + "weighted_rkd_loss": 0.03048628754913807 + }, + { + "epoch": 1.0997963340122199, + "flow_matching_loss": 0.01609187861904502, + "grad_norm": 0.25584256649017334, + "learning_rate": 2.998617645291735e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012159755267202855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005858974950388074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030178485624492168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9902667760848999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3780, + "weighted_flow_matching_loss": 0.01609187861904502, + "weighted_rkd_loss": 0.029884745739400385 + }, + { + "epoch": 1.1027058481233634, + "flow_matching_loss": 0.016272437945008277, + "grad_norm": 0.2580440044403076, + "learning_rate": 2.9985819326323097e-05, + "loss": 0.0465, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01242948854342103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005634658457711339, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030493636056780814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9902153015136719, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3790, + "weighted_flow_matching_loss": 0.016272437945008277, + "weighted_rkd_loss": 0.030195297673344612 + }, + { + "epoch": 1.1056153622345068, + "flow_matching_loss": 0.017178004421293736, + "grad_norm": 0.27097851037979126, + "learning_rate": 2.9985457647432613e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012672234885394573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005908349296078086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031252818554639815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9901636958122253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3800, + "weighted_flow_matching_loss": 0.017178004421293736, + "weighted_rkd_loss": 0.030945413187146185 + }, + { + "epoch": 1.1085248763456503, + "flow_matching_loss": 0.014515696000307799, + "grad_norm": 0.24243274331092834, + "learning_rate": 2.998509141635576e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01262426320463419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006233498360961676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0314820246770978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9901119589805603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3810, + "weighted_flow_matching_loss": 0.014515696000307799, + "weighted_rkd_loss": 0.031170726381242277 + }, + { + "epoch": 1.1114343904567938, + "flow_matching_loss": 0.016527270339429377, + "grad_norm": 0.2725295424461365, + "learning_rate": 2.99847206332038e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012515289522707463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006165928905829787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03119650799781084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9900600790977478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3820, + "weighted_flow_matching_loss": 0.016527270339429377, + "weighted_rkd_loss": 0.0308864314109087 + }, + { + "epoch": 1.114343904567937, + "flow_matching_loss": 0.01549617424607277, + "grad_norm": 0.31777870655059814, + "learning_rate": 2.9984345298089356e-05, + "loss": 0.0464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012523784581571818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006187325296923518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031234894692897797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9900080680847168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3830, + "weighted_flow_matching_loss": 0.01549617424607277, + "weighted_rkd_loss": 0.030922814272344112 + }, + { + "epoch": 1.1172534186790806, + "flow_matching_loss": 0.015519543271511792, + "grad_norm": 0.2923346161842346, + "learning_rate": 2.998396541112645e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011801216471940278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00561711578629911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02921954859048128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9899559199810029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3840, + "weighted_flow_matching_loss": 0.015519543271511792, + "weighted_rkd_loss": 0.028926081582903863 + }, + { + "epoch": 1.120162932790224, + "flow_matching_loss": 0.017273545358330013, + "grad_norm": 0.273937463760376, + "learning_rate": 2.998358097243048e-05, + "loss": 0.0464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011898272670805454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005595404049381613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029391949623823167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9899036467075348, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3850, + "weighted_flow_matching_loss": 0.017273545358330013, + "weighted_rkd_loss": 0.029095190204679967 + }, + { + "epoch": 1.1230724469013675, + "flow_matching_loss": 0.015316135063767434, + "grad_norm": 0.34680166840553284, + "learning_rate": 2.998319198211823e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012783889658749104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006438274448737502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03200605381280184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9898512363433838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3860, + "weighted_flow_matching_loss": 0.015316135063767434, + "weighted_rkd_loss": 0.03168121837079525 + }, + { + "epoch": 1.1259819610125108, + "flow_matching_loss": 0.01512721749022603, + "grad_norm": 0.2648478150367737, + "learning_rate": 2.998279844030786e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012070967350155115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005909764440730214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030051699094474315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9897986829280854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3870, + "weighted_flow_matching_loss": 0.01512721749022603, + "weighted_rkd_loss": 0.029745131358504297 + }, + { + "epoch": 1.1288914751236543, + "flow_matching_loss": 0.015614660456776619, + "grad_norm": 0.2586710453033447, + "learning_rate": 2.9982400347118926e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01262418096885085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00639050486497581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031638866662979125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9897460043430328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3880, + "weighted_flow_matching_loss": 0.015614660456776619, + "weighted_rkd_loss": 0.03131444901227951 + }, + { + "epoch": 1.1318009892347978, + "flow_matching_loss": 0.015809947811067104, + "grad_norm": 0.3250906467437744, + "learning_rate": 2.9981997702672353e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012781326938420535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061707722023129465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03173342570662498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9896931827068329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3890, + "weighted_flow_matching_loss": 0.015809947811067104, + "weighted_rkd_loss": 0.031406381726264955 + }, + { + "epoch": 1.1347105033459413, + "flow_matching_loss": 0.015429712273180485, + "grad_norm": 0.25817984342575073, + "learning_rate": 2.9981590507090454e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011710861138999462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005817229673266411, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029238951951265336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9896402418613434, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3900, + "weighted_flow_matching_loss": 0.015429712273180485, + "weighted_rkd_loss": 0.02893606759607792 + }, + { + "epoch": 1.1376200174570847, + "flow_matching_loss": 0.017466566897928713, + "grad_norm": 0.2551655173301697, + "learning_rate": 2.9981178760496927e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01208475772291422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005646894499659538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02981640975922346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9895871460437775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3910, + "weighted_flow_matching_loss": 0.017466566897928713, + "weighted_rkd_loss": 0.029505963996052743 + }, + { + "epoch": 1.140529531568228, + "flow_matching_loss": 0.016687026992440224, + "grad_norm": 0.3189692497253418, + "learning_rate": 2.9980762463016848e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011168255656957626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005886372830718756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02822288405150175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.989533931016922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3920, + "weighted_flow_matching_loss": 0.016687026992440224, + "weighted_rkd_loss": 0.027927501127123833 + }, + { + "epoch": 1.1434390456793715, + "flow_matching_loss": 0.015280418749898671, + "grad_norm": 0.29144904017448425, + "learning_rate": 2.9980341614776673e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012738159950822591, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006545344460755586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032021664083004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9894805788993836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3930, + "weighted_flow_matching_loss": 0.015280418749898671, + "weighted_rkd_loss": 0.031684857234358786 + }, + { + "epoch": 1.146348559790515, + "flow_matching_loss": 0.016688871569931506, + "grad_norm": 0.3343261480331421, + "learning_rate": 2.9979916215904247e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012462745793163777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006061210064217448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03098670169711113, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9894270896911621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3940, + "weighted_flow_matching_loss": 0.016688871569931506, + "weighted_rkd_loss": 0.030659079924225806 + }, + { + "epoch": 1.1492580739016585, + "flow_matching_loss": 0.017801901884377004, + "grad_norm": 0.23812705278396606, + "learning_rate": 2.99794862665288e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012156054377555847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005880860518664122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030192968994379045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9893734753131866, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3950, + "weighted_flow_matching_loss": 0.017801901884377004, + "weighted_rkd_loss": 0.029872100055217742 + }, + { + "epoch": 1.152167588012802, + "flow_matching_loss": 0.016334368754178284, + "grad_norm": 0.2923268973827362, + "learning_rate": 2.9979051766780938e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01155077638104558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005855512898415327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028957065567374228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9893197119235992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3960, + "weighted_flow_matching_loss": 0.016334368754178284, + "weighted_rkd_loss": 0.02864778507500887 + }, + { + "epoch": 1.1550771021239452, + "flow_matching_loss": 0.014421662129461766, + "grad_norm": 0.29588693380355835, + "learning_rate": 2.9978612716792647e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01105430070310831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005324276583269239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02743287794291973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9892658293247223, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3970, + "weighted_flow_matching_loss": 0.014421662129461766, + "weighted_rkd_loss": 0.02713836878538132 + }, + { + "epoch": 1.1579866162350887, + "flow_matching_loss": 0.01568668643012643, + "grad_norm": 0.32115283608436584, + "learning_rate": 2.99781691166973e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012350194249302148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006136539904400706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030836928449571133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9892118036746979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3980, + "weighted_flow_matching_loss": 0.01568668643012643, + "weighted_rkd_loss": 0.030504229292273522 + }, + { + "epoch": 1.1608961303462322, + "flow_matching_loss": 0.015853261388838292, + "grad_norm": 0.32044318318367004, + "learning_rate": 2.9977720966629646e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012819794565439224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005977633502334356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03161722309887409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.989157646894455, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3990, + "weighted_flow_matching_loss": 0.015853261388838292, + "weighted_rkd_loss": 0.03127440810203552 + }, + { + "epoch": 1.1638056444573757, + "flow_matching_loss": 0.016058335732668638, + "grad_norm": 0.28831058740615845, + "learning_rate": 2.997726826672583e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012541523482650519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006088323937729001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03117137048393488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.989103353023529, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4000, + "weighted_flow_matching_loss": 0.016058335732668638, + "weighted_rkd_loss": 0.030831710435450077 + }, + { + "epoch": 1.1667151585685192, + "flow_matching_loss": 0.017083024978637694, + "grad_norm": 0.3354160189628601, + "learning_rate": 2.9976811017123368e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012885598931461573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005933664366602897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03170486204326153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9890489220619202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4010, + "weighted_flow_matching_loss": 0.017083024978637694, + "weighted_rkd_loss": 0.03135764915496111 + }, + { + "epoch": 1.1696246726796624, + "flow_matching_loss": 0.015708048082888126, + "grad_norm": 0.2773033082485199, + "learning_rate": 2.9976349217961158e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011570034641772508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005833460204303265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028973529674112798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9889943599700928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4020, + "weighted_flow_matching_loss": 0.015708048082888126, + "weighted_rkd_loss": 0.028654685243964195 + }, + { + "epoch": 1.172534186790806, + "flow_matching_loss": 0.01631864979863167, + "grad_norm": 0.325560986995697, + "learning_rate": 2.9975882869379482e-05, + "loss": 0.0455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011821576952934265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005834061978384852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029477215930819513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9889396727085114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4030, + "weighted_flow_matching_loss": 0.01631864979863167, + "weighted_rkd_loss": 0.02915120255202055 + }, + { + "epoch": 1.1754437009019494, + "flow_matching_loss": 0.015244421549141407, + "grad_norm": 0.2960995137691498, + "learning_rate": 2.997541197152001e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011686866264790296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005704355845227838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029078088141977786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9888848423957824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4040, + "weighted_flow_matching_loss": 0.015244421549141407, + "weighted_rkd_loss": 0.028754884377121925 + }, + { + "epoch": 1.1783532150130929, + "flow_matching_loss": 0.01604560688138008, + "grad_norm": 0.357870876789093, + "learning_rate": 2.9974936524525774e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013037064857780934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006263620872050524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032337750867009164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9888298809528351, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4050, + "weighted_flow_matching_loss": 0.01604560688138008, + "weighted_rkd_loss": 0.031976531445980075 + }, + { + "epoch": 1.1812627291242364, + "flow_matching_loss": 0.017365876119583845, + "grad_norm": 0.28251007199287415, + "learning_rate": 2.9974456528541222e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012403829582035541, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006145427701994777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030953087285161017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9887747704982758, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4060, + "weighted_flow_matching_loss": 0.017365876119583845, + "weighted_rkd_loss": 0.0306056410074234 + }, + { + "epoch": 1.1841722432353796, + "flow_matching_loss": 0.015554477181285619, + "grad_norm": 0.28031206130981445, + "learning_rate": 2.9973971983712147e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012586492672562599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061803657561540605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031353351473808286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9887195527553558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4070, + "weighted_flow_matching_loss": 0.015554477181285619, + "weighted_rkd_loss": 0.030999713391065598 + }, + { + "epoch": 1.187081757346523, + "flow_matching_loss": 0.013379156310111284, + "grad_norm": 0.2771613895893097, + "learning_rate": 2.9973482890185753e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011802960373461246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00545972641557455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029065647535026072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9886641800403595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4080, + "weighted_flow_matching_loss": 0.013379156310111284, + "weighted_rkd_loss": 0.028736188448965548 + }, + { + "epoch": 1.1899912714576666, + "flow_matching_loss": 0.01650505941361189, + "grad_norm": 0.31591323018074036, + "learning_rate": 2.9972989248110606e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011955253593623639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005882153427228332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029792660474777223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9886086881160736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4090, + "weighted_flow_matching_loss": 0.01650505941361189, + "weighted_rkd_loss": 0.02945328950881958 + }, + { + "epoch": 1.19290078556881, + "flow_matching_loss": 0.014569080527871847, + "grad_norm": 0.27015021443367004, + "learning_rate": 2.9972491057636665e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011916804779320955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005989770358428359, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02982338033616543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9885530591011047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4100, + "weighted_flow_matching_loss": 0.014569080527871847, + "weighted_rkd_loss": 0.02948201633989811 + }, + { + "epoch": 1.1958102996799536, + "flow_matching_loss": 0.01677712593227625, + "grad_norm": 0.3331674337387085, + "learning_rate": 2.9971988318915268e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012223812378942966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006067536771297455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030515161342918873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9884972870349884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4110, + "weighted_flow_matching_loss": 0.01677712593227625, + "weighted_rkd_loss": 0.0301642095670104 + }, + { + "epoch": 1.1987198137910968, + "flow_matching_loss": 0.015401473268866539, + "grad_norm": 0.32896333932876587, + "learning_rate": 2.997148103209913e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011633305344730616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057807477656751875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02904735840857029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.988441389799118, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4120, + "weighted_flow_matching_loss": 0.015401473268866539, + "weighted_rkd_loss": 0.028711626864969732 + }, + { + "epoch": 1.2016293279022403, + "flow_matching_loss": 0.016325051710009576, + "grad_norm": 0.27578750252723694, + "learning_rate": 2.9970969197342356e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012139895465224981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006030951160937548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03031074199825525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9883853554725647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4130, + "weighted_flow_matching_loss": 0.016325051710009576, + "weighted_rkd_loss": 0.02995868492871523 + }, + { + "epoch": 1.2045388420133838, + "flow_matching_loss": 0.01504892585799098, + "grad_norm": 0.38844773173332214, + "learning_rate": 2.9970452814800422e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011989373341202736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006025130162015557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030003876611590385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9883291900157929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4140, + "weighted_flow_matching_loss": 0.01504892585799098, + "weighted_rkd_loss": 0.02965365517884493 + }, + { + "epoch": 1.2074483561245273, + "flow_matching_loss": 0.015126300230622291, + "grad_norm": 0.24974791705608368, + "learning_rate": 2.99699318846302e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011765951570123435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005708995694294572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029240898415446283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.988272887468338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4150, + "weighted_flow_matching_loss": 0.015126300230622291, + "weighted_rkd_loss": 0.02889801226556301 + }, + { + "epoch": 1.2103578702356708, + "flow_matching_loss": 0.015436310693621636, + "grad_norm": 0.3510286211967468, + "learning_rate": 2.9969406406989927e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012060373183339834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005802864208817482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02992361057549715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9882164537906647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4160, + "weighted_flow_matching_loss": 0.015436310693621636, + "weighted_rkd_loss": 0.029571009054780006 + }, + { + "epoch": 1.213267384346814, + "flow_matching_loss": 0.01560963150113821, + "grad_norm": 0.28830164670944214, + "learning_rate": 2.996887638203923e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012276879325509072, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005815457459539175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030369216576218605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9881598651409149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4170, + "weighted_flow_matching_loss": 0.01560963150113821, + "weighted_rkd_loss": 0.030009655095636845 + }, + { + "epoch": 1.2161768984579575, + "flow_matching_loss": 0.015430605970323086, + "grad_norm": 0.257860392332077, + "learning_rate": 2.9968341809939116e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012083979416638612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005599512159824372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029767471179366112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9881031811237335, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4180, + "weighted_flow_matching_loss": 0.015430605970323086, + "weighted_rkd_loss": 0.029413321055471897 + }, + { + "epoch": 1.219086412569101, + "flow_matching_loss": 0.01755090355873108, + "grad_norm": 0.2947503626346588, + "learning_rate": 2.9967802690851985e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011762381624430418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006023313477635383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029548076912760735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9880463421344757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4190, + "weighted_flow_matching_loss": 0.01755090355873108, + "weighted_rkd_loss": 0.029194844886660576 + }, + { + "epoch": 1.2219959266802445, + "flow_matching_loss": 0.015418660547584295, + "grad_norm": 0.35973918437957764, + "learning_rate": 2.9967259024941595e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011990603059530258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005837728129699826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029818933829665183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9879893720149994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4200, + "weighted_flow_matching_loss": 0.015418660547584295, + "weighted_rkd_loss": 0.02946077398955822 + }, + { + "epoch": 1.2249054407913877, + "flow_matching_loss": 0.01623535994440317, + "grad_norm": 0.28992557525634766, + "learning_rate": 2.9966710812373097e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012080925889313221, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005971877789124847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030133729986846446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9879322588443756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4210, + "weighted_flow_matching_loss": 0.01623535994440317, + "weighted_rkd_loss": 0.02977011650800705 + }, + { + "epoch": 1.2278149549025312, + "flow_matching_loss": 0.017178228218108417, + "grad_norm": 0.2752803564071655, + "learning_rate": 2.996615805331303e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011014973092824221, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005389459058642387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027419405244290828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9878750205039978, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4220, + "weighted_flow_matching_loss": 0.017178228218108417, + "weighted_rkd_loss": 0.027086925506591798 + }, + { + "epoch": 1.2307244690136747, + "flow_matching_loss": 0.015215567965060472, + "grad_norm": 0.25838789343833923, + "learning_rate": 2.99656007479293e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011257686093449593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057209473568946125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028236319310963154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9878176510334015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4230, + "weighted_flow_matching_loss": 0.015215567965060472, + "weighted_rkd_loss": 0.027892350032925606 + }, + { + "epoch": 1.2336339831248182, + "flow_matching_loss": 0.01570466151461005, + "grad_norm": 0.30548954010009766, + "learning_rate": 2.9965038896391213e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011148328706622124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005280932039022445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027577589452266692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9877601444721222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4240, + "weighted_flow_matching_loss": 0.01570466151461005, + "weighted_rkd_loss": 0.027240022644400598 + }, + { + "epoch": 1.2365434972359617, + "flow_matching_loss": 0.016791129671037196, + "grad_norm": 0.2599770426750183, + "learning_rate": 2.9964472498869434e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011263151746243238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00544657614082098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027972879633307458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9877025008201599, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4250, + "weighted_flow_matching_loss": 0.016791129671037196, + "weighted_rkd_loss": 0.027628862112760545 + }, + { + "epoch": 1.239453011347105, + "flow_matching_loss": 0.01531132571399212, + "grad_norm": 0.3259899318218231, + "learning_rate": 2.996390155553603e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011977021861821414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006098329043015838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030052373372018337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9876447379589081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4260, + "weighted_flow_matching_loss": 0.01531132571399212, + "weighted_rkd_loss": 0.029681045189499856 + }, + { + "epoch": 1.2423625254582484, + "flow_matching_loss": 0.015623761713504792, + "grad_norm": 0.2723449170589447, + "learning_rate": 2.9963326066564426e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010436586849391461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050939205568283795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025967094488441943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9875868260860443, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4270, + "weighted_flow_matching_loss": 0.015623761713504792, + "weighted_rkd_loss": 0.025644771195948123 + }, + { + "epoch": 1.245272039569392, + "flow_matching_loss": 0.01565769677981734, + "grad_norm": 0.32124751806259155, + "learning_rate": 2.996274603212945e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011597502510994673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005511705810204148, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028706710785627365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.987528783082962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4280, + "weighted_flow_matching_loss": 0.01565769677981734, + "weighted_rkd_loss": 0.0283486757427454 + }, + { + "epoch": 1.2481815536805354, + "flow_matching_loss": 0.01715508783236146, + "grad_norm": 0.2602415680885315, + "learning_rate": 2.9962161452407296e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011647560261189937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057805829215794805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029075703397393227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9874706089496612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4290, + "weighted_flow_matching_loss": 0.01715508783236146, + "weighted_rkd_loss": 0.028711409494280814 + }, + { + "epoch": 1.2510910677916787, + "flow_matching_loss": 0.015951192285865546, + "grad_norm": 0.27380672097206116, + "learning_rate": 2.996157232757555e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011300671380013227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005372051638551057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02797339465469122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.987412303686142, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4300, + "weighted_flow_matching_loss": 0.015951192285865546, + "weighted_rkd_loss": 0.027621274627745152 + }, + { + "epoch": 1.2540005819028222, + "flow_matching_loss": 0.015190057177096605, + "grad_norm": 0.25020161271095276, + "learning_rate": 2.996097865781316e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011943488661199809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005624269694089889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029511247389018537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9873538553714752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4310, + "weighted_flow_matching_loss": 0.015190057177096605, + "weighted_rkd_loss": 0.029138055630028248 + }, + { + "epoch": 1.2569100960139656, + "flow_matching_loss": 0.01450390089303255, + "grad_norm": 0.2660652697086334, + "learning_rate": 2.996038044330048e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011793280020356178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005599042121320963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029185602068901063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9872952818870544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4320, + "weighted_flow_matching_loss": 0.01450390089303255, + "weighted_rkd_loss": 0.028814787045121193 + }, + { + "epoch": 1.2598196101251091, + "flow_matching_loss": 0.01572409775108099, + "grad_norm": 0.26360514760017395, + "learning_rate": 2.9959777684219233e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010309815872460604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004899899568408727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025519531220197678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9872365713119506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4330, + "weighted_flow_matching_loss": 0.01572409775108099, + "weighted_rkd_loss": 0.02519379686564207 + }, + { + "epoch": 1.2627291242362526, + "flow_matching_loss": 0.01513874065130949, + "grad_norm": 0.3088046908378601, + "learning_rate": 2.9959170380752508e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011209197714924813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005259190266951918, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02767758574336767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9871777355670929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4340, + "weighted_flow_matching_loss": 0.01513874065130949, + "weighted_rkd_loss": 0.027322699874639512 + }, + { + "epoch": 1.2656386383473959, + "flow_matching_loss": 0.016426373180001973, + "grad_norm": 0.3044181764125824, + "learning_rate": 2.99585585330848e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011502168141305447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005772610241547227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02877694647759199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9871187508106232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4350, + "weighted_flow_matching_loss": 0.016426373180001973, + "weighted_rkd_loss": 0.028406261838972567 + }, + { + "epoch": 1.2685481524585394, + "flow_matching_loss": 0.016845317371189595, + "grad_norm": 0.2606402039527893, + "learning_rate": 2.9957942141401967e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01109478883445263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005532497959211469, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027722075767815112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9870596408843995, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4360, + "weighted_flow_matching_loss": 0.016845317371189595, + "weighted_rkd_loss": 0.027363353595137595 + }, + { + "epoch": 1.2714576665696828, + "flow_matching_loss": 0.01415220545604825, + "grad_norm": 0.33161601424217224, + "learning_rate": 2.995732120589125e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011276703700423241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005466339876875281, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028019747324287893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9870004057884216, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4370, + "weighted_flow_matching_loss": 0.01415220545604825, + "weighted_rkd_loss": 0.027655455097556115 + }, + { + "epoch": 1.2743671806808263, + "flow_matching_loss": 0.015483355056494475, + "grad_norm": 0.3019312620162964, + "learning_rate": 2.9956695726741276e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011770815961062909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0055736774578690525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029115309193730354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9869410216808319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4380, + "weighted_flow_matching_loss": 0.015483355056494475, + "weighted_rkd_loss": 0.028735104016959667 + }, + { + "epoch": 1.2772766947919698, + "flow_matching_loss": 0.016859203577041626, + "grad_norm": 0.2794947028160095, + "learning_rate": 2.995606570414205e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011201841291040182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005291707953438163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027695390582084655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9868814945220947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4390, + "weighted_flow_matching_loss": 0.016859203577041626, + "weighted_rkd_loss": 0.02733205780386925 + }, + { + "epoch": 1.280186208903113, + "flow_matching_loss": 0.016895103082060815, + "grad_norm": 0.2757188677787781, + "learning_rate": 2.9955431138284955e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011243888083845377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005472135171294213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02795991152524948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9868218660354614, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4400, + "weighted_flow_matching_loss": 0.016895103082060815, + "weighted_rkd_loss": 0.027591456845402717 + }, + { + "epoch": 1.2830957230142566, + "flow_matching_loss": 0.01633802568539977, + "grad_norm": 0.27499160170555115, + "learning_rate": 2.9954792029362754e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011431543482467532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005743327317759395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028606414049863815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9867620885372161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4410, + "weighted_flow_matching_loss": 0.01633802568539977, + "weighted_rkd_loss": 0.028227742575109006 + }, + { + "epoch": 1.2860052371254, + "flow_matching_loss": 0.016216258518397808, + "grad_norm": 0.283979594707489, + "learning_rate": 2.9954148377569598e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010460616275668144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005130497878417373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026051730662584306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9867021799087524, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4420, + "weighted_flow_matching_loss": 0.016216258518397808, + "weighted_rkd_loss": 0.025705312378704548 + }, + { + "epoch": 1.2889147512365435, + "flow_matching_loss": 0.017318779323250055, + "grad_norm": 0.2575053572654724, + "learning_rate": 2.9953500183101002e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011119201965630054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005442480277270079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02768088411539793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9866421341896057, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4430, + "weighted_flow_matching_loss": 0.017318779323250055, + "weighted_rkd_loss": 0.027311134338378906 + }, + { + "epoch": 1.291824265347687, + "flow_matching_loss": 0.01613399339839816, + "grad_norm": 0.28743207454681396, + "learning_rate": 2.9952847446153877e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011326804291456938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005212845280766487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027866453863680363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.986581951379776, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4440, + "weighted_flow_matching_loss": 0.01613399339839816, + "weighted_rkd_loss": 0.0274925472214818 + }, + { + "epoch": 1.2947337794588303, + "flow_matching_loss": 0.018347480986267328, + "grad_norm": 0.28332194685935974, + "learning_rate": 2.9952190166926498e-05, + "loss": 0.0453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01116673331707716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005018309224396944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027351775765419008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9865216493606568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4450, + "weighted_flow_matching_loss": 0.018347480986267328, + "weighted_rkd_loss": 0.026983148790895937 + }, + { + "epoch": 1.2976432935699738, + "flow_matching_loss": 0.016275725793093443, + "grad_norm": 0.233274906873703, + "learning_rate": 2.9951528345618546e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01119851702824235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005411933967843651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027808968164026737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9864611983299255, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4460, + "weighted_flow_matching_loss": 0.016275725793093443, + "weighted_rkd_loss": 0.027432483062148093 + }, + { + "epoch": 1.3005528076811173, + "flow_matching_loss": 0.01699780151247978, + "grad_norm": 0.26756614446640015, + "learning_rate": 2.995086198243105e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011261493060737848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005635647382587194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028158633783459664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9864006221294404, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4470, + "weighted_flow_matching_loss": 0.01699780151247978, + "weighted_rkd_loss": 0.027775727212429047 + }, + { + "epoch": 1.3034623217922607, + "flow_matching_loss": 0.01824235003441572, + "grad_norm": 0.25973019003868103, + "learning_rate": 2.9950191077566434e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011403738241642713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00567076881416142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028478245623409747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9863399267196655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4480, + "weighted_flow_matching_loss": 0.01824235003441572, + "weighted_rkd_loss": 0.028089249134063722 + }, + { + "epoch": 1.3063718359034042, + "flow_matching_loss": 0.014792430680245161, + "grad_norm": 0.25923076272010803, + "learning_rate": 2.9949515631228515e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010298336949199438, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004905517562292516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025502191483974458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9862790703773499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4490, + "weighted_flow_matching_loss": 0.014792430680245161, + "weighted_rkd_loss": 0.025152284651994705 + }, + { + "epoch": 1.3092813500145475, + "flow_matching_loss": 0.015441843681037426, + "grad_norm": 0.26937440037727356, + "learning_rate": 2.994883564362246e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011424543708562851, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005616527376696467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02846561446785927, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9862180888652802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4500, + "weighted_flow_matching_loss": 0.015441843681037426, + "weighted_rkd_loss": 0.028073333203792572 + }, + { + "epoch": 1.312190864125691, + "flow_matching_loss": 0.015964278392493723, + "grad_norm": 0.28625234961509705, + "learning_rate": 2.994815111495484e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010650001186877488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00497000296600163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026270005293190478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9861569762229919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4510, + "weighted_flow_matching_loss": 0.015964278392493723, + "weighted_rkd_loss": 0.02590632289648056 + }, + { + "epoch": 1.3151003782368345, + "flow_matching_loss": 0.014619791880249977, + "grad_norm": 0.3148649036884308, + "learning_rate": 2.9947462045433594e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012140804529190063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005807665549218655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030089274607598783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9860957384109497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4520, + "weighted_flow_matching_loss": 0.014619791880249977, + "weighted_rkd_loss": 0.029670915752649307 + }, + { + "epoch": 1.318009892347978, + "flow_matching_loss": 0.016705347411334514, + "grad_norm": 0.2664354741573334, + "learning_rate": 2.9946768435268045e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01158783994615078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005692280968651176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028867961280047892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.98603435754776, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4530, + "weighted_flow_matching_loss": 0.016705347411334514, + "weighted_rkd_loss": 0.02846479993313551 + }, + { + "epoch": 1.3209194064591214, + "flow_matching_loss": 0.014711481984704733, + "grad_norm": 0.284700483083725, + "learning_rate": 2.9946070284668884e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011335507966578007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005203226348385215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027874242141842843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9859728455543518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4540, + "weighted_flow_matching_loss": 0.014711481984704733, + "weighted_rkd_loss": 0.02748325001448393 + }, + { + "epoch": 1.3238289205702647, + "flow_matching_loss": 0.015370065812021494, + "grad_norm": 0.27125754952430725, + "learning_rate": 2.994536759384821e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011392591148614883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005423990311101079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028209172748029233, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9859112083911896, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4550, + "weighted_flow_matching_loss": 0.015370065812021494, + "weighted_rkd_loss": 0.027811748906970024 + }, + { + "epoch": 1.3267384346814082, + "flow_matching_loss": 0.01616780124604702, + "grad_norm": 0.27009862661361694, + "learning_rate": 2.994466036301946e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01118941167369485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005351496301591397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02773031983524561, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9858494281768799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4560, + "weighted_flow_matching_loss": 0.01616780124604702, + "weighted_rkd_loss": 0.027337906695902348 + }, + { + "epoch": 1.3296479487925517, + "flow_matching_loss": 0.016839444916695357, + "grad_norm": 0.3170153498649597, + "learning_rate": 2.9943948592397494e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010712444595992566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005405138991773128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026830027997493743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9857875108718872, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4570, + "weighted_flow_matching_loss": 0.016839444916695357, + "weighted_rkd_loss": 0.026448684744536877 + }, + { + "epoch": 1.3325574629036951, + "flow_matching_loss": 0.016673833038657902, + "grad_norm": 0.3116300404071808, + "learning_rate": 2.994323228219851e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01150734182447195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00538751243147999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028402195870876314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9857254743576049, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4580, + "weighted_flow_matching_loss": 0.016673833038657902, + "weighted_rkd_loss": 0.027996756508946417 + }, + { + "epoch": 1.3354669770148386, + "flow_matching_loss": 0.015540473256260156, + "grad_norm": 0.30885574221611023, + "learning_rate": 2.9942511432640113e-05, + "loss": 0.0424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010902607254683972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005429906165227294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02723512053489685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9856632888317108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4590, + "weighted_flow_matching_loss": 0.015540473256260156, + "weighted_rkd_loss": 0.026844626292586327 + }, + { + "epoch": 1.338376491125982, + "flow_matching_loss": 0.017066449392586946, + "grad_norm": 0.2346208095550537, + "learning_rate": 2.9941786043941278e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01081798765808344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005201239604502916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026837214827537537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9856009840965271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4600, + "weighted_flow_matching_loss": 0.017066449392586946, + "weighted_rkd_loss": 0.026450826786458492 + }, + { + "epoch": 1.3412860052371254, + "flow_matching_loss": 0.01671601040288806, + "grad_norm": 0.27311304211616516, + "learning_rate": 2.9941056116322357e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010635056160390377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005552603863179684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02682271581143141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9855385422706604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4610, + "weighted_flow_matching_loss": 0.01671601040288806, + "weighted_rkd_loss": 0.02643483765423298 + }, + { + "epoch": 1.3441955193482689, + "flow_matching_loss": 0.016524933092296124, + "grad_norm": 0.29457172751426697, + "learning_rate": 2.994032165000508e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011260752519592642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005435364693403244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027956869453191757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9854759693145752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4620, + "weighted_flow_matching_loss": 0.016524933092296124, + "weighted_rkd_loss": 0.02755083404481411 + }, + { + "epoch": 1.3471050334594123, + "flow_matching_loss": 0.014580247085541486, + "grad_norm": 0.2863948941230774, + "learning_rate": 2.9939582645212566e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010947143565863371, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005329720047302544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02722400724887848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9854132533073425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4630, + "weighted_flow_matching_loss": 0.014580247085541486, + "weighted_rkd_loss": 0.02682689353823662 + }, + { + "epoch": 1.3500145475705558, + "flow_matching_loss": 0.018473496474325658, + "grad_norm": 0.23746736347675323, + "learning_rate": 2.9938839102169303e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010936431773006915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004933289578184485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026806153170764447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9853504121303558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4640, + "weighted_flow_matching_loss": 0.018473496474325658, + "weighted_rkd_loss": 0.0264134518802166 + }, + { + "epoch": 1.352924061681699, + "flow_matching_loss": 0.015128001477569342, + "grad_norm": 0.2749095559120178, + "learning_rate": 2.993809102110116e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010715643782168627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00507135116495192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02650263886898756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9852874398231506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4650, + "weighted_flow_matching_loss": 0.015128001477569342, + "weighted_rkd_loss": 0.026112750731408597 + }, + { + "epoch": 1.3558335757928426, + "flow_matching_loss": 0.014979046490043401, + "grad_norm": 0.30144065618515015, + "learning_rate": 2.9937338402235373e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011360440030694008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005494261765852571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028215141966938972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9852243065834045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4660, + "weighted_flow_matching_loss": 0.014979046490043401, + "weighted_rkd_loss": 0.027798249386250973 + }, + { + "epoch": 1.358743089903986, + "flow_matching_loss": 0.016647800151258708, + "grad_norm": 0.2722247540950775, + "learning_rate": 2.993658124580058e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011401648633182049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005513199511915446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0283164968714118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9851610898971558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4670, + "weighted_flow_matching_loss": 0.016647800151258708, + "weighted_rkd_loss": 0.02789630778133869 + }, + { + "epoch": 1.3616526040151296, + "flow_matching_loss": 0.015222217608243226, + "grad_norm": 0.29295840859413147, + "learning_rate": 2.9935819552026785e-05, + "loss": 0.0455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012487652525305749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005746214278042316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030721519514918326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9850977063179016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4680, + "weighted_flow_matching_loss": 0.015222217608243226, + "weighted_rkd_loss": 0.03026368487626314 + }, + { + "epoch": 1.364562118126273, + "flow_matching_loss": 0.015265952330082655, + "grad_norm": 0.3696195185184479, + "learning_rate": 2.9935053321145368e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011680952087044717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005918906256556511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029280810430645943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9850341975688934, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4690, + "weighted_flow_matching_loss": 0.015265952330082655, + "weighted_rkd_loss": 0.02884253431111574 + }, + { + "epoch": 1.3674716322374163, + "flow_matching_loss": 0.017328743264079093, + "grad_norm": 0.2787955403327942, + "learning_rate": 2.9934282553389088e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011754173971712589, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057662364561110735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02927458453923464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9849705696105957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4700, + "weighted_flow_matching_loss": 0.017328743264079093, + "weighted_rkd_loss": 0.028834605403244496 + }, + { + "epoch": 1.3703811463485598, + "flow_matching_loss": 0.014761111326515674, + "grad_norm": 0.2304742932319641, + "learning_rate": 2.9933507248992084e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010649273544549942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005174128874205053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026472675800323486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.984906792640686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4710, + "weighted_flow_matching_loss": 0.014761111326515674, + "weighted_rkd_loss": 0.026073134690523147 + }, + { + "epoch": 1.3732906604597033, + "flow_matching_loss": 0.013840542919933797, + "grad_norm": 0.3928302824497223, + "learning_rate": 2.9932727408189876e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011580667458474636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005572572303935886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028733907081186772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9848428785800933, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4720, + "weighted_flow_matching_loss": 0.013840542919933797, + "weighted_rkd_loss": 0.02829840164631605 + }, + { + "epoch": 1.3762001745708465, + "flow_matching_loss": 0.014519343804568052, + "grad_norm": 0.28296440839767456, + "learning_rate": 2.9931943031219356e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01060384102165699, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004870258900336921, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026077941246330737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9847788393497467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4730, + "weighted_flow_matching_loss": 0.014519343804568052, + "weighted_rkd_loss": 0.025680994614958763 + }, + { + "epoch": 1.3791096886819902, + "flow_matching_loss": 0.015917461831122637, + "grad_norm": 0.26377609372138977, + "learning_rate": 2.9931154118318803e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011339655332267284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005476656323298812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028155967220664026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9847146689891815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4740, + "weighted_flow_matching_loss": 0.015917461831122637, + "weighted_rkd_loss": 0.02772558182477951 + }, + { + "epoch": 1.3820192027931335, + "flow_matching_loss": 0.015052027348428965, + "grad_norm": 0.23230843245983124, + "learning_rate": 2.9930360669727858e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010426886193454266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004887704364955425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025741476751863956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9846503734588623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4750, + "weighted_flow_matching_loss": 0.015052027348428965, + "weighted_rkd_loss": 0.02534637302160263 + }, + { + "epoch": 1.384928716904277, + "flow_matching_loss": 0.01660174913704395, + "grad_norm": 0.38880935311317444, + "learning_rate": 2.992956268568755e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011382163036614656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00534019460901618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028104520961642267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9845859229564666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4760, + "weighted_flow_matching_loss": 0.01660174913704395, + "weighted_rkd_loss": 0.02767126578837633 + }, + { + "epoch": 1.3878382310154205, + "flow_matching_loss": 0.015879571717232466, + "grad_norm": 0.36786705255508423, + "learning_rate": 2.99287601664403e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01184660792350769, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005766458017751575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0294596740975976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9845213532447815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4770, + "weighted_flow_matching_loss": 0.015879571717232466, + "weighted_rkd_loss": 0.02900365721434355 + }, + { + "epoch": 1.3907477451265637, + "flow_matching_loss": 0.0163159754127264, + "grad_norm": 0.28286486864089966, + "learning_rate": 2.992795311222988e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010539864283055067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048142530024051665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0258939815685153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9844566524028778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4780, + "weighted_flow_matching_loss": 0.0163159754127264, + "weighted_rkd_loss": 0.025491480343043806 + }, + { + "epoch": 1.3936572592377072, + "flow_matching_loss": 0.015647327806800605, + "grad_norm": 0.2492673248052597, + "learning_rate": 2.9927141523301453e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010057377628982067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050810981076210735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025195853412151338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9843918144702911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4790, + "weighted_flow_matching_loss": 0.015647327806800605, + "weighted_rkd_loss": 0.02480257544666529 + }, + { + "epoch": 1.3965667733488507, + "flow_matching_loss": 0.014703121222555637, + "grad_norm": 0.2825538218021393, + "learning_rate": 2.992632539990156e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010611401777714491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049919809214770796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026214784383773802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9843268454074859, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4800, + "weighted_flow_matching_loss": 0.014703121222555637, + "weighted_rkd_loss": 0.025803944654762746 + }, + { + "epoch": 1.3994762874599942, + "flow_matching_loss": 0.015092565212398767, + "grad_norm": 0.36747369170188904, + "learning_rate": 2.9925504742278117e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011307812016457319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005729617550969124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028345241397619247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9842617452144623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4810, + "weighted_flow_matching_loss": 0.015092565212398767, + "weighted_rkd_loss": 0.02789910975843668 + }, + { + "epoch": 1.4023858015711377, + "flow_matching_loss": 0.015646854881197214, + "grad_norm": 0.3262200355529785, + "learning_rate": 2.9924679550680416e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011627135612070561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005472672265022993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028726943396031858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9841965079307556, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4820, + "weighted_flow_matching_loss": 0.015646854881197214, + "weighted_rkd_loss": 0.02827298678457737 + }, + { + "epoch": 1.405295315682281, + "flow_matching_loss": 0.014500455744564534, + "grad_norm": 0.2543656826019287, + "learning_rate": 2.9923849825359133e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011110718036070466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005181124247610569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027402560226619244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9841311395168304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4830, + "weighted_flow_matching_loss": 0.014500455744564534, + "weighted_rkd_loss": 0.02696773912757635 + }, + { + "epoch": 1.4082048297934244, + "flow_matching_loss": 0.01711323643103242, + "grad_norm": 0.3127693831920624, + "learning_rate": 2.9923015566566317e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010656991600990295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004916771268472076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02623075433075428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9840656518936157, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4840, + "weighted_flow_matching_loss": 0.01711323643103242, + "weighted_rkd_loss": 0.025812776200473307 + }, + { + "epoch": 1.411114343904568, + "flow_matching_loss": 0.015118444431573152, + "grad_norm": 0.30079084634780884, + "learning_rate": 2.992217677455539e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011650298349559307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005383284413255751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028683881275355816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9840000033378601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4850, + "weighted_flow_matching_loss": 0.015118444431573152, + "weighted_rkd_loss": 0.028224910609424114 + }, + { + "epoch": 1.4140238580157114, + "flow_matching_loss": 0.01714483918622136, + "grad_norm": 0.3106515109539032, + "learning_rate": 2.9921333449581153e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01075666407123208, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005317620793357491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026830948889255524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9839342415332795, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4860, + "weighted_flow_matching_loss": 0.01714483918622136, + "weighted_rkd_loss": 0.02639987114816904 + }, + { + "epoch": 1.4169333721268549, + "flow_matching_loss": 0.013981352653354406, + "grad_norm": 0.24586281180381775, + "learning_rate": 2.9920485591899795e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010665399860590697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005089581524953246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026420381478965284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9838683426380157, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4870, + "weighted_flow_matching_loss": 0.013981352653354406, + "weighted_rkd_loss": 0.025994164124131202 + }, + { + "epoch": 1.4198428862379981, + "flow_matching_loss": 0.01545362239703536, + "grad_norm": 0.29441332817077637, + "learning_rate": 2.991963320176886e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011996132507920265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005969913210719824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029962178133428097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9838023126125336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4880, + "weighted_flow_matching_loss": 0.01545362239703536, + "weighted_rkd_loss": 0.02947693448513746 + }, + { + "epoch": 1.4227524003491416, + "flow_matching_loss": 0.01603837814182043, + "grad_norm": 0.2946488857269287, + "learning_rate": 2.991877627944729e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011272125039249658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005754859931766987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028299110382795332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9837361514568329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4890, + "weighted_flow_matching_loss": 0.01603837814182043, + "weighted_rkd_loss": 0.02783887889236212 + }, + { + "epoch": 1.4256619144602851, + "flow_matching_loss": 0.016880910936743022, + "grad_norm": 0.3193369507789612, + "learning_rate": 2.9917914825195393e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011528273206204176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005623571667820215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028680118173360823, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9836698472499847, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4900, + "weighted_flow_matching_loss": 0.016880910936743022, + "weighted_rkd_loss": 0.02821170650422573 + }, + { + "epoch": 1.4285714285714286, + "flow_matching_loss": 0.016661809477955102, + "grad_norm": 0.2914353013038635, + "learning_rate": 2.991704883927486e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01112956665456295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005022003478370607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027281136810779573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9836034178733826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4910, + "weighted_flow_matching_loss": 0.016661809477955102, + "weighted_rkd_loss": 0.02683382835239172 + }, + { + "epoch": 1.431480942682572, + "flow_matching_loss": 0.01642534779384732, + "grad_norm": 0.3282681405544281, + "learning_rate": 2.991617832194875e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011307290010154247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056558924727141855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028270472213625907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9835368514060974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4920, + "weighted_flow_matching_loss": 0.01642534779384732, + "weighted_rkd_loss": 0.027805030159652234 + }, + { + "epoch": 1.4343904567937154, + "flow_matching_loss": 0.018094405345618726, + "grad_norm": 0.2550683319568634, + "learning_rate": 2.9915303273481504e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010589493531733751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004977558553218841, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02615654580295086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9834701418876648, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4930, + "weighted_flow_matching_loss": 0.018094405345618726, + "weighted_rkd_loss": 0.0257241690531373 + }, + { + "epoch": 1.4372999709048588, + "flow_matching_loss": 0.015306400414556266, + "grad_norm": 0.303922563791275, + "learning_rate": 2.9914423694138942e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011005380190908909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005237299483269453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027248059771955012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9834033370018005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4940, + "weighted_flow_matching_loss": 0.015306400414556266, + "weighted_rkd_loss": 0.026795860193669797 + }, + { + "epoch": 1.4402094850160023, + "flow_matching_loss": 0.014181260857731104, + "grad_norm": 0.2988678216934204, + "learning_rate": 2.9913539584188253e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010579347051680088, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005158142652362585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02631683647632599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9833363831043244, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4950, + "weighted_flow_matching_loss": 0.014181260857731104, + "weighted_rkd_loss": 0.025878296233713627 + }, + { + "epoch": 1.4431189991271458, + "flow_matching_loss": 0.01516383932903409, + "grad_norm": 0.29053518176078796, + "learning_rate": 2.9912650943898008e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010508168395608664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004916211008094251, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025932547822594642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9832692861557006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4960, + "weighted_flow_matching_loss": 0.01516383932903409, + "weighted_rkd_loss": 0.025498666800558567 + }, + { + "epoch": 1.4460285132382893, + "flow_matching_loss": 0.014713722653687, + "grad_norm": 0.32646873593330383, + "learning_rate": 2.9911757773538148e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01113414838910103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053100124932825565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027578309178352356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9832020580768586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4970, + "weighted_flow_matching_loss": 0.014713722653687, + "weighted_rkd_loss": 0.027114991284906863 + }, + { + "epoch": 1.4489380273494326, + "flow_matching_loss": 0.015332086849957705, + "grad_norm": 0.2492109090089798, + "learning_rate": 2.9910860073380005e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009911021031439305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004757074406370521, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024579116329550742, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9831346988677978, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4980, + "weighted_flow_matching_loss": 0.015332086849957705, + "weighted_rkd_loss": 0.02416458111256361 + }, + { + "epoch": 1.451847541460576, + "flow_matching_loss": 0.014684156980365515, + "grad_norm": 0.2936452329158783, + "learning_rate": 2.9909957843696267e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011699477769434452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005278093600645661, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028677049465477465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9830672144889832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4990, + "weighted_flow_matching_loss": 0.014684156980365515, + "weighted_rkd_loss": 0.028191477246582507 + }, + { + "epoch": 1.4547570555717195, + "flow_matching_loss": 0.014576353318989276, + "grad_norm": 0.3324815630912781, + "learning_rate": 2.990905108476101e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01078219348564744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052208029199391605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026785189844667913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.982999587059021, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5000, + "weighted_flow_matching_loss": 0.014576353318989276, + "weighted_rkd_loss": 0.02632980179041624 + }, + { + "epoch": 1.457666569682863, + "flow_matching_loss": 0.017003392707556485, + "grad_norm": 0.27027297019958496, + "learning_rate": 2.9908139796849683e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010459831263870002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005177131667733192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02609679438173771, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9829318344593048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5010, + "weighted_flow_matching_loss": 0.017003392707556485, + "weighted_rkd_loss": 0.025651362352073194 + }, + { + "epoch": 1.4605760837940065, + "flow_matching_loss": 0.016139674186706542, + "grad_norm": 0.25955846905708313, + "learning_rate": 2.9907223980239114e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011423656344413757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005092139262706042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027939452044665813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9828639626502991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5020, + "weighted_flow_matching_loss": 0.016139674186706542, + "weighted_rkd_loss": 0.027460655383765697 + }, + { + "epoch": 1.4634855979051498, + "flow_matching_loss": 0.014888795651495456, + "grad_norm": 0.26991429924964905, + "learning_rate": 2.99063036352075e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01070809792727232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00489394050091505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026310136541724205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9827959299087524, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5030, + "weighted_flow_matching_loss": 0.014888795651495456, + "weighted_rkd_loss": 0.025857510231435298 + }, + { + "epoch": 1.4663951120162932, + "flow_matching_loss": 0.014979096408933402, + "grad_norm": 0.25639450550079346, + "learning_rate": 2.9905378762034424e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011872419621795416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005510864360257983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029255703464150427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9827277839183808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5040, + "weighted_flow_matching_loss": 0.014979096408933402, + "weighted_rkd_loss": 0.028750420734286308 + }, + { + "epoch": 1.4693046261274367, + "flow_matching_loss": 0.016773895546793937, + "grad_norm": 0.28577759861946106, + "learning_rate": 2.9904449361000833e-05, + "loss": 0.0424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010472690500319005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005121188517659902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026066569425165654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9826594948768616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5050, + "weighted_flow_matching_loss": 0.016773895546793937, + "weighted_rkd_loss": 0.02561454251408577 + }, + { + "epoch": 1.4722141402385802, + "flow_matching_loss": 0.014004754927009344, + "grad_norm": 0.29985547065734863, + "learning_rate": 2.9903515432389056e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010705871693789958, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053973684553056955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026809111796319486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9825910866260529, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5060, + "weighted_flow_matching_loss": 0.014004754927009344, + "weighted_rkd_loss": 0.026342372223734856 + }, + { + "epoch": 1.4751236543497237, + "flow_matching_loss": 0.014831526856869458, + "grad_norm": 0.31296297907829285, + "learning_rate": 2.990257697648279e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010771110840141773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004945311858318746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02648753356188536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9825225353240967, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5070, + "weighted_flow_matching_loss": 0.014831526856869458, + "weighted_rkd_loss": 0.02602456323802471 + }, + { + "epoch": 1.478033168460867, + "flow_matching_loss": 0.01616538278758526, + "grad_norm": 0.3256014287471771, + "learning_rate": 2.9901633993567125e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010423570312559605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004781817854382098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025628958642482758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.982453852891922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5080, + "weighted_flow_matching_loss": 0.01616538278758526, + "weighted_rkd_loss": 0.025179245695471763 + }, + { + "epoch": 1.4809426825720104, + "flow_matching_loss": 0.014070382993668317, + "grad_norm": 0.2749958634376526, + "learning_rate": 2.99006864839285e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011574190389364959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005495480191893875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028643861040472986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9823850333690644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5090, + "weighted_flow_matching_loss": 0.014070382993668317, + "weighted_rkd_loss": 0.0281393576413393 + }, + { + "epoch": 1.483852196683154, + "flow_matching_loss": 0.016462529823184014, + "grad_norm": 0.2968834340572357, + "learning_rate": 2.989973444785476e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010389482509344817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005185231799259782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0259641969576478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9823160886764526, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5100, + "weighted_flow_matching_loss": 0.016462529823184014, + "weighted_rkd_loss": 0.025505034253001213 + }, + { + "epoch": 1.4867617107942974, + "flow_matching_loss": 0.015392791386693716, + "grad_norm": 0.3220158517360687, + "learning_rate": 2.989877788563509e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010178117826581002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047651027562096715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025121338479220868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.982246994972229, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5110, + "weighted_flow_matching_loss": 0.015392791386693716, + "weighted_rkd_loss": 0.024675352312624455 + }, + { + "epoch": 1.489671224905441, + "flow_matching_loss": 0.01659672912210226, + "grad_norm": 0.2600126266479492, + "learning_rate": 2.9897816797560085e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010997162014245988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005121431592851877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02711575571447611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9821778059005737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5120, + "weighted_flow_matching_loss": 0.01659672912210226, + "weighted_rkd_loss": 0.02663248311728239 + }, + { + "epoch": 1.4925807390165842, + "flow_matching_loss": 0.014992323704063892, + "grad_norm": 0.272063672542572, + "learning_rate": 2.989685118392169e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011093649640679359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005467187240719795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027654486522078515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9821084678173065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5130, + "weighted_flow_matching_loss": 0.014992323704063892, + "weighted_rkd_loss": 0.027159717120230197 + }, + { + "epoch": 1.4954902531277277, + "flow_matching_loss": 0.014943788573145867, + "grad_norm": 0.2700706422328949, + "learning_rate": 2.9895881045013232e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010263311723247171, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004846698138862848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025373321399092673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9820389926433564, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5140, + "weighted_flow_matching_loss": 0.014943788573145867, + "weighted_rkd_loss": 0.02491756118834019 + }, + { + "epoch": 1.4983997672388711, + "flow_matching_loss": 0.014737242367118597, + "grad_norm": 0.2913779318332672, + "learning_rate": 2.9894906381129414e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010451946454122663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004925159295089543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02582905236631632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9819693803787232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5150, + "weighted_flow_matching_loss": 0.014737242367118597, + "weighted_rkd_loss": 0.025363336503505706 + }, + { + "epoch": 1.5013092813500144, + "flow_matching_loss": 0.01457502618432045, + "grad_norm": 0.3284592628479004, + "learning_rate": 2.9893927192566316e-05, + "loss": 0.0424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01136061903089285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005619464418850839, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02834070287644863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9818996369838715, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5160, + "weighted_flow_matching_loss": 0.01457502618432045, + "weighted_rkd_loss": 0.02782769612967968 + }, + { + "epoch": 1.504218795461158, + "flow_matching_loss": 0.015263149607926607, + "grad_norm": 0.27267682552337646, + "learning_rate": 2.989294347962139e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010164338629692793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005062860110774637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02539153750985861, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9818297743797302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5170, + "weighted_flow_matching_loss": 0.015263149607926607, + "weighted_rkd_loss": 0.02493020351976156 + }, + { + "epoch": 1.5071283095723014, + "flow_matching_loss": 0.014419104345142842, + "grad_norm": 0.3126676380634308, + "learning_rate": 2.989195524259346e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011440042220056058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00535621247254312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028236296959221363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.981759774684906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5180, + "weighted_flow_matching_loss": 0.014419104345142842, + "weighted_rkd_loss": 0.02772122770547867 + }, + { + "epoch": 1.5100378236834449, + "flow_matching_loss": 0.01427716612815857, + "grad_norm": 0.2927142381668091, + "learning_rate": 2.9890962481782723e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010489669255912304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00498783583752811, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025967174768447877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9816896378993988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5190, + "weighted_flow_matching_loss": 0.01427716612815857, + "weighted_rkd_loss": 0.025491696409881116 + }, + { + "epoch": 1.5129473377945883, + "flow_matching_loss": 0.015086509194225073, + "grad_norm": 0.35532987117767334, + "learning_rate": 2.9889965197490757e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011126839369535447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005364829371683299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027618508040904998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9816193580627441, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5200, + "weighted_flow_matching_loss": 0.015086509194225073, + "weighted_rkd_loss": 0.02711088601499796 + }, + { + "epoch": 1.5158568519057316, + "flow_matching_loss": 0.016268151719123124, + "grad_norm": 0.27802032232284546, + "learning_rate": 2.988896339002052e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011295988969504834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005678465776145458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02827044352889061, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9815489768981933, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5210, + "weighted_flow_matching_loss": 0.016268151719123124, + "weighted_rkd_loss": 0.027748848125338555 + }, + { + "epoch": 1.5187663660168753, + "flow_matching_loss": 0.01595028266310692, + "grad_norm": 0.2134842723608017, + "learning_rate": 2.9887957059676315e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010570163279771805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004875858966261149, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026016185618937017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9814784467220307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5220, + "weighted_flow_matching_loss": 0.01595028266310692, + "weighted_rkd_loss": 0.02553436178714037 + }, + { + "epoch": 1.5216758801280186, + "flow_matching_loss": 0.015106334816664457, + "grad_norm": 0.3115711510181427, + "learning_rate": 2.9886946206763855e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0098333521746099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046629113610833885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024329615756869316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.981407779455185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5230, + "weighted_flow_matching_loss": 0.015106334816664457, + "weighted_rkd_loss": 0.023877236619591714 + }, + { + "epoch": 1.524585394239162, + "flow_matching_loss": 0.01595349656417966, + "grad_norm": 0.2575652301311493, + "learning_rate": 2.9885930831590202e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011291226651519538, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005515784956514835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028098238073289393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9813369750976563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5240, + "weighted_flow_matching_loss": 0.01595349656417966, + "weighted_rkd_loss": 0.027573831751942633 + }, + { + "epoch": 1.5274949083503055, + "flow_matching_loss": 0.015430074092000723, + "grad_norm": 0.25069308280944824, + "learning_rate": 2.98849109344638e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010705020558089019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005185246677137911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026595287956297398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9812660455703736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5250, + "weighted_flow_matching_loss": 0.015430074092000723, + "weighted_rkd_loss": 0.026097049564123155 + }, + { + "epoch": 1.5304044224614488, + "flow_matching_loss": 0.014825723692774772, + "grad_norm": 0.3033854365348816, + "learning_rate": 2.9883886515694474e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011284223478287458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005300272884778678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027868719957768916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9811949908733368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5260, + "weighted_flow_matching_loss": 0.014825723692774772, + "weighted_rkd_loss": 0.027344650030136107 + }, + { + "epoch": 1.5333139365725925, + "flow_matching_loss": 0.01511730458587408, + "grad_norm": 0.30628713965415955, + "learning_rate": 2.988285757559341e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010381461773067712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005011481000110507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025774404779076575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9811238050460815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5270, + "weighted_flow_matching_loss": 0.01511730458587408, + "weighted_rkd_loss": 0.025287831388413905 + }, + { + "epoch": 1.5362234506837358, + "flow_matching_loss": 0.016044924221932887, + "grad_norm": 0.292979896068573, + "learning_rate": 2.988182411447317e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01031127581372857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005016243876889348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02563879545778036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9810524821281433, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5280, + "weighted_flow_matching_loss": 0.016044924221932887, + "weighted_rkd_loss": 0.025153004750609397 + }, + { + "epoch": 1.5391329647948793, + "flow_matching_loss": 0.016785888466984034, + "grad_norm": 0.25503915548324585, + "learning_rate": 2.98807861326477e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010495661199092865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00506539260968566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026056714914739133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.98098104596138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5290, + "weighted_flow_matching_loss": 0.016785888466984034, + "weighted_rkd_loss": 0.025561130419373513 + }, + { + "epoch": 1.5420424789060228, + "flow_matching_loss": 0.014294416457414628, + "grad_norm": 0.2609098553657532, + "learning_rate": 2.9879743630432307e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01026564920321107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00490747825242579, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025438776612281798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9809094429016113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5300, + "weighted_flow_matching_loss": 0.014294416457414628, + "weighted_rkd_loss": 0.024953184835612773 + }, + { + "epoch": 1.544951993017166, + "flow_matching_loss": 0.015655660908669232, + "grad_norm": 0.28236886858940125, + "learning_rate": 2.9878696608143677e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010608669836074114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005458025657571852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0266753651201725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9808377265930176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5310, + "weighted_flow_matching_loss": 0.015655660908669232, + "weighted_rkd_loss": 0.026164171285927296 + }, + { + "epoch": 1.5478615071283097, + "flow_matching_loss": 0.015863559767603876, + "grad_norm": 0.2858717143535614, + "learning_rate": 2.9877645066099862e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010272334469482303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005369101697579026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025913770496845245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9807658791542053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5320, + "weighted_flow_matching_loss": 0.015863559767603876, + "weighted_rkd_loss": 0.0254153311252594 + }, + { + "epoch": 1.550771021239453, + "flow_matching_loss": 0.016437431797385216, + "grad_norm": 0.2990921437740326, + "learning_rate": 2.9876589004620298e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010832840576767921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004972539073787629, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026638220250606536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9806938946247101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5330, + "weighted_flow_matching_loss": 0.016437431797385216, + "weighted_rkd_loss": 0.026123949140310288 + }, + { + "epoch": 1.5536805353505965, + "flow_matching_loss": 0.01647294843569398, + "grad_norm": 0.2833864688873291, + "learning_rate": 2.987552842402579e-05, + "loss": 0.0411, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010290936846286059, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004575086641125381, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025156960263848303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9806217789649964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5340, + "weighted_flow_matching_loss": 0.01647294843569398, + "weighted_rkd_loss": 0.024669476971030235 + }, + { + "epoch": 1.55659004946174, + "flow_matching_loss": 0.0151178739964962, + "grad_norm": 0.2703924775123596, + "learning_rate": 2.9874463324638515e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010114171449095011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004652228020131588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024880571104586125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9805495321750641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5350, + "weighted_flow_matching_loss": 0.0151178739964962, + "weighted_rkd_loss": 0.024396654777228834 + }, + { + "epoch": 1.5594995635728832, + "flow_matching_loss": 0.014320675190538168, + "grad_norm": 0.2505466341972351, + "learning_rate": 2.9873393706782016e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010727415047585964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005162394652143121, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026617224514484405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9804771542549133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5360, + "weighted_flow_matching_loss": 0.014320675190538168, + "weighted_rkd_loss": 0.026097549498081206 + }, + { + "epoch": 1.562409077684027, + "flow_matching_loss": 0.016691526770591734, + "grad_norm": 0.25889885425567627, + "learning_rate": 2.987231957078122e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010472187586128712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00492242204491049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02586679719388485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9804046511650085, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5370, + "weighted_flow_matching_loss": 0.016691526770591734, + "weighted_rkd_loss": 0.02535996288061142 + }, + { + "epoch": 1.5653185917951702, + "flow_matching_loss": 0.014269305765628815, + "grad_norm": 0.24030418694019318, + "learning_rate": 2.9871240916962414e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009975122241303324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004902978800237179, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024853223375976084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9803320169448853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5380, + "weighted_flow_matching_loss": 0.014269305765628815, + "weighted_rkd_loss": 0.02436444479972124 + }, + { + "epoch": 1.5682281059063137, + "flow_matching_loss": 0.015166699141263961, + "grad_norm": 0.29857370257377625, + "learning_rate": 2.9870157745653274e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010318018915131687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005075115151703358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02571115307509899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.98025923371315, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5390, + "weighted_flow_matching_loss": 0.015166699141263961, + "weighted_rkd_loss": 0.02520357947796583 + }, + { + "epoch": 1.5711376200174572, + "flow_matching_loss": 0.015290232561528682, + "grad_norm": 0.23591434955596924, + "learning_rate": 2.9869070057182838e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010098335286602377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046863863477483395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024883056804537774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9801863372325897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5400, + "weighted_flow_matching_loss": 0.015290232561528682, + "weighted_rkd_loss": 0.024389995262026787 + }, + { + "epoch": 1.5740471341286004, + "flow_matching_loss": 0.014702382031828165, + "grad_norm": 0.23105870187282562, + "learning_rate": 2.986797785188151e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009571698866784573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004352514445781708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02349591236561537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.980113297700882, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5410, + "weighted_flow_matching_loss": 0.014702382031828165, + "weighted_rkd_loss": 0.02302867416292429 + }, + { + "epoch": 1.576956648239744, + "flow_matching_loss": 0.01471664709970355, + "grad_norm": 0.2580011785030365, + "learning_rate": 2.9866881130081074e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009863391751423478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004483613534830511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024210396967828274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9800401329994202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5420, + "weighted_flow_matching_loss": 0.01471664709970355, + "weighted_rkd_loss": 0.023727196641266345 + }, + { + "epoch": 1.5798661623508874, + "flow_matching_loss": 0.0157160303555429, + "grad_norm": 0.28588810563087463, + "learning_rate": 2.986577989211469e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009629614558070898, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046039534034207465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02386318240314722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9799668371677399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5430, + "weighted_flow_matching_loss": 0.0157160303555429, + "weighted_rkd_loss": 0.02338511496782303 + }, + { + "epoch": 1.5827756764620309, + "flow_matching_loss": 0.014558604080229997, + "grad_norm": 0.23975956439971924, + "learning_rate": 2.9864674138316887e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00973142171278596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004431354883126915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023894198052585126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9798934042453766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5440, + "weighted_flow_matching_loss": 0.014558604080229997, + "weighted_rkd_loss": 0.023413790576159953 + }, + { + "epoch": 1.5856851905731744, + "flow_matching_loss": 0.01617082264274359, + "grad_norm": 0.2899096608161926, + "learning_rate": 2.9863563869023554e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010760476440191269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005097774392925203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026618727296590806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9798198342323303, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5450, + "weighted_flow_matching_loss": 0.01617082264274359, + "weighted_rkd_loss": 0.026081542298197747 + }, + { + "epoch": 1.5885947046843176, + "flow_matching_loss": 0.01588248210027814, + "grad_norm": 0.31475400924682617, + "learning_rate": 2.9862449084571972e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01095920018851757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005308617651462555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027227018028497696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9797461450099945, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5460, + "weighted_flow_matching_loss": 0.01588248210027814, + "weighted_rkd_loss": 0.026675580814480782 + }, + { + "epoch": 1.5915042187954611, + "flow_matching_loss": 0.015882621798664332, + "grad_norm": 0.3024378716945648, + "learning_rate": 2.9861329785300774e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010889915190637112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005149328499101102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026929158717393875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9796723127365112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5470, + "weighted_flow_matching_loss": 0.015882621798664332, + "weighted_rkd_loss": 0.026381755992770196 + }, + { + "epoch": 1.5944137329066046, + "flow_matching_loss": 0.014474795758724212, + "grad_norm": 0.3754127621650696, + "learning_rate": 2.9860205971549978e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01014743596315384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004832484386861324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025127356313169, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9795983672142029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5480, + "weighted_flow_matching_loss": 0.014474795758724212, + "weighted_rkd_loss": 0.024614678323268892 + }, + { + "epoch": 1.597323247017748, + "flow_matching_loss": 0.015335299726575612, + "grad_norm": 0.2897777855396271, + "learning_rate": 2.9859077643660974e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01075655473396182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005144184874370694, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026657294295728207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9795242726802826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5490, + "weighted_flow_matching_loss": 0.015335299726575612, + "weighted_rkd_loss": 0.026111454516649247 + }, + { + "epoch": 1.6002327611288916, + "flow_matching_loss": 0.014649754855781794, + "grad_norm": 0.22665800154209137, + "learning_rate": 2.985794480197651e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01013543177396059, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004743651021271944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025014514848589898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9794500589370727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5500, + "weighted_flow_matching_loss": 0.014649754855781794, + "weighted_rkd_loss": 0.024500468745827675 + }, + { + "epoch": 1.6031422752400348, + "flow_matching_loss": 0.015950732119381427, + "grad_norm": 0.27847445011138916, + "learning_rate": 2.9856807446840716e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010563763510435819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004849057109095156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025976583920419218, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.97937570810318, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5510, + "weighted_flow_matching_loss": 0.015950732119381427, + "weighted_rkd_loss": 0.02544087152928114 + }, + { + "epoch": 1.6060517893511783, + "flow_matching_loss": 0.01589626641944051, + "grad_norm": 0.2590188980102539, + "learning_rate": 2.9855665578599093e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009955349378287792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00492962053976953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024840319342911244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9793012320995331, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5520, + "weighted_flow_matching_loss": 0.01589626641944051, + "weighted_rkd_loss": 0.024326134845614432 + }, + { + "epoch": 1.6089613034623218, + "flow_matching_loss": 0.017155612539499997, + "grad_norm": 0.30467087030410767, + "learning_rate": 2.9854519197598504e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01037179259583354, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005112443980760872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02585602942854166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9792266070842743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5530, + "weighted_flow_matching_loss": 0.017155612539499997, + "weighted_rkd_loss": 0.02531896997243166 + }, + { + "epoch": 1.6118708175734653, + "flow_matching_loss": 0.014358546026051045, + "grad_norm": 0.27660536766052246, + "learning_rate": 2.9853368304187197e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010777992568910122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004915503971278668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02647148920223117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9791518688201905, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5540, + "weighted_flow_matching_loss": 0.014358546026051045, + "weighted_rkd_loss": 0.02591962171718478 + }, + { + "epoch": 1.6147803316846088, + "flow_matching_loss": 0.015376094356179237, + "grad_norm": 0.2962818145751953, + "learning_rate": 2.9852212898714783e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009850421641021966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004750081966631114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024450925551354884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9790769875049591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5550, + "weighted_flow_matching_loss": 0.015376094356179237, + "weighted_rkd_loss": 0.02393929157406092 + }, + { + "epoch": 1.617689845795752, + "flow_matching_loss": 0.016560736671090127, + "grad_norm": 0.2859225869178772, + "learning_rate": 2.9851052981532233e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010202006343752146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004554877220652997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024958889558911325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9790019869804383, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5560, + "weighted_flow_matching_loss": 0.016560736671090127, + "weighted_rkd_loss": 0.02443480081856251 + }, + { + "epoch": 1.6205993599068955, + "flow_matching_loss": 0.015774458087980746, + "grad_norm": 0.26766642928123474, + "learning_rate": 2.9849888552991908e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009899156028404832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00503799261059612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024836304783821105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9789268434047699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5570, + "weighted_flow_matching_loss": 0.015774458087980746, + "weighted_rkd_loss": 0.024312941171228886 + }, + { + "epoch": 1.623508874018039, + "flow_matching_loss": 0.01601511174812913, + "grad_norm": 0.25901854038238525, + "learning_rate": 2.9848719613447535e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009874464850872755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004699723771773278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024448653683066368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9788515746593476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5580, + "weighted_flow_matching_loss": 0.01601511174812913, + "weighted_rkd_loss": 0.023931550979614257 + }, + { + "epoch": 1.6264183881291823, + "flow_matching_loss": 0.015059714019298554, + "grad_norm": 0.3025739789009094, + "learning_rate": 2.9847546163254194e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010170658212155103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046188171254470944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024960133619606494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9787761688232421, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5590, + "weighted_flow_matching_loss": 0.015059714019298554, + "weighted_rkd_loss": 0.024430355802178384 + }, + { + "epoch": 1.629327902240326, + "flow_matching_loss": 0.014040351100265979, + "grad_norm": 0.254093199968338, + "learning_rate": 2.984636820276836e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009674861282110214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004629438696429134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023979161120951175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9787006378173828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5600, + "weighted_flow_matching_loss": 0.014040351100265979, + "weighted_rkd_loss": 0.02346840538084507 + }, + { + "epoch": 1.6322374163514692, + "flow_matching_loss": 0.014378631580621004, + "grad_norm": 0.3095475435256958, + "learning_rate": 2.984518573234786e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010418618191033601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005152963986620307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02599020041525364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9786249697208405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5610, + "weighted_flow_matching_loss": 0.014378631580621004, + "weighted_rkd_loss": 0.02543463874608278 + }, + { + "epoch": 1.6351469304626127, + "flow_matching_loss": 0.015362306404858828, + "grad_norm": 0.29661890864372253, + "learning_rate": 2.98439987523519e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010887590982019902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005078270542435348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026853452436625957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9785491704940796, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5620, + "weighted_flow_matching_loss": 0.015362306404858828, + "weighted_rkd_loss": 0.02627745233476162 + }, + { + "epoch": 1.6380564445737562, + "flow_matching_loss": 0.01423096638172865, + "grad_norm": 0.23538701236248016, + "learning_rate": 2.9842807263141052e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0107145338319242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005175903136841953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02660497073084116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9784732460975647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5630, + "weighted_flow_matching_loss": 0.01423096638172865, + "weighted_rkd_loss": 0.026032267324626445 + }, + { + "epoch": 1.6409659586848995, + "flow_matching_loss": 0.01724084410816431, + "grad_norm": 0.32410478591918945, + "learning_rate": 2.9841611265077256e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011098826490342616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050474964315071705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02724514976143837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9783971965312958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5640, + "weighted_flow_matching_loss": 0.01724084410816431, + "weighted_rkd_loss": 0.02665654718875885 + }, + { + "epoch": 1.6438754727960432, + "flow_matching_loss": 0.01634203614667058, + "grad_norm": 0.24156451225280762, + "learning_rate": 2.9840410758523832e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01023393338546157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004597794264554977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025065660662949085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9783210098743439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5650, + "weighted_flow_matching_loss": 0.01634203614667058, + "weighted_rkd_loss": 0.024522225745022296 + }, + { + "epoch": 1.6467849869071864, + "flow_matching_loss": 0.016258678771555422, + "grad_norm": 0.23640209436416626, + "learning_rate": 2.9839205743845453e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010415041632950307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004905977682210505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02573606073856354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9782446920871735, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5660, + "weighted_flow_matching_loss": 0.016258678771555422, + "weighted_rkd_loss": 0.02517620474100113 + }, + { + "epoch": 1.64969450101833, + "flow_matching_loss": 0.015836291667073964, + "grad_norm": 0.23963557183742523, + "learning_rate": 2.983799622140818e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009143257373943925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004348263936117291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022634778171777725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9781682312488555, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5670, + "weighted_flow_matching_loss": 0.015836291667073964, + "weighted_rkd_loss": 0.022140614688396454 + }, + { + "epoch": 1.6526040151294734, + "flow_matching_loss": 0.01558458162471652, + "grad_norm": 0.3015161454677582, + "learning_rate": 2.9836782191579425e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01139493053779006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057506530545651914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02854051422327757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9780916452407837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5680, + "weighted_flow_matching_loss": 0.01558458162471652, + "weighted_rkd_loss": 0.027915243431925772 + }, + { + "epoch": 1.6555135292406167, + "flow_matching_loss": 0.014739287551492453, + "grad_norm": 0.2978746294975281, + "learning_rate": 2.983556365472799e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010719756595790386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005112967500463128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026552480645477772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9780149400234223, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5690, + "weighted_flow_matching_loss": 0.014739287551492453, + "weighted_rkd_loss": 0.025968693755567075 + }, + { + "epoch": 1.6584230433517604, + "flow_matching_loss": 0.015259434841573239, + "grad_norm": 0.3024307191371918, + "learning_rate": 2.9834340611224022e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010051833745092154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005009913956746459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02511358130723238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9779380857944489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5700, + "weighted_flow_matching_loss": 0.015259434841573239, + "weighted_rkd_loss": 0.024559522047638893 + }, + { + "epoch": 1.6613325574629036, + "flow_matching_loss": 0.014548598509281873, + "grad_norm": 0.2647421956062317, + "learning_rate": 2.9833113061439057e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010660436423495411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050947961397469045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026415669173002244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9778611123561859, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5710, + "weighted_flow_matching_loss": 0.014548598509281873, + "weighted_rkd_loss": 0.025830836035311222 + }, + { + "epoch": 1.6642420715740471, + "flow_matching_loss": 0.015684116818010806, + "grad_norm": 0.29639631509780884, + "learning_rate": 2.9831881005745986e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009591517876833678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044017932144924995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02358482889831066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9777840077877045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5720, + "weighted_flow_matching_loss": 0.015684116818010806, + "weighted_rkd_loss": 0.02306085843592882 + }, + { + "epoch": 1.6671515856851906, + "flow_matching_loss": 0.018237083684653044, + "grad_norm": 0.26607903838157654, + "learning_rate": 2.983064444451908e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009871066454797984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004677088838070631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02441922165453434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9777067720890045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5730, + "weighted_flow_matching_loss": 0.018237083684653044, + "weighted_rkd_loss": 0.023874838277697562 + }, + { + "epoch": 1.6700610997963339, + "flow_matching_loss": 0.017362096719443797, + "grad_norm": 0.30207374691963196, + "learning_rate": 2.9829403378133974e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010576813109219074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005074102710932493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026227728836238383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.977629405260086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5740, + "weighted_flow_matching_loss": 0.017362096719443797, + "weighted_rkd_loss": 0.02564096264541149 + }, + { + "epoch": 1.6729706139074776, + "flow_matching_loss": 0.015380278788506984, + "grad_norm": 0.25149136781692505, + "learning_rate": 2.9828157806967668e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00987101043574512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048157258192077276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024557746574282648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9775519073009491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5750, + "weighted_flow_matching_loss": 0.015380278788506984, + "weighted_rkd_loss": 0.024006476625800133 + }, + { + "epoch": 1.6758801280186209, + "flow_matching_loss": 0.017006926611065863, + "grad_norm": 0.2612192630767822, + "learning_rate": 2.9826907731398533e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010505951102823019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005292333848774433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02630423605442047, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9774742722511292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5760, + "weighted_flow_matching_loss": 0.017006926611065863, + "weighted_rkd_loss": 0.025711746513843538 + }, + { + "epoch": 1.6787896421297643, + "flow_matching_loss": 0.016084355302155018, + "grad_norm": 0.24589794874191284, + "learning_rate": 2.9825653151806315e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00955626885406673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044681050581857566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023580642975866796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9773965120315552, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5770, + "weighted_flow_matching_loss": 0.016084355302155018, + "weighted_rkd_loss": 0.023047642596066 + }, + { + "epoch": 1.6816991562409078, + "flow_matching_loss": 0.01807961044833064, + "grad_norm": 0.2564457058906555, + "learning_rate": 2.9824394068572114e-05, + "loss": 0.0411, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009529603272676468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004479330219328404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023538536950945856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9773186206817627, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5780, + "weighted_flow_matching_loss": 0.01807961044833064, + "weighted_rkd_loss": 0.023004632629454135 + }, + { + "epoch": 1.684608670352051, + "flow_matching_loss": 0.015963610634207724, + "grad_norm": 0.28044435381889343, + "learning_rate": 2.9823130482078415e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00978642595000565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004518468398600817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024091320298612117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9772405922412872, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5790, + "weighted_flow_matching_loss": 0.015963610634207724, + "weighted_rkd_loss": 0.023543019965291023 + }, + { + "epoch": 1.6875181844631948, + "flow_matching_loss": 0.014743242971599102, + "grad_norm": 0.2580192983150482, + "learning_rate": 2.9821862392709054e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009617903782054783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004334699548780918, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02357050683349371, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9771624445915222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5800, + "weighted_flow_matching_loss": 0.014743242971599102, + "weighted_rkd_loss": 0.023032196797430515 + }, + { + "epoch": 1.690427698574338, + "flow_matching_loss": 0.016129953879863025, + "grad_norm": 0.32394006848335266, + "learning_rate": 2.982058980084925e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010737307090312242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005077702272683382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026552316546440125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9770841538906098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5810, + "weighted_flow_matching_loss": 0.016129953879863025, + "weighted_rkd_loss": 0.02594387363642454 + }, + { + "epoch": 1.6933372126854815, + "flow_matching_loss": 0.01600977508351207, + "grad_norm": 0.35138556361198425, + "learning_rate": 2.9819312706885577e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01048330394551158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004972370737232268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025938978791236876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9770057439804077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5820, + "weighted_flow_matching_loss": 0.01600977508351207, + "weighted_rkd_loss": 0.025342515669763088 + }, + { + "epoch": 1.696246726796625, + "flow_matching_loss": 0.014572138991206884, + "grad_norm": 0.24727988243103027, + "learning_rate": 2.9818031111205986e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010096015082672239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004636789415962994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024828819558024406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9769271969795227, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5830, + "weighted_flow_matching_loss": 0.014572138991206884, + "weighted_rkd_loss": 0.024255982600152493 + }, + { + "epoch": 1.6991562409077683, + "flow_matching_loss": 0.016754224617034196, + "grad_norm": 0.25630906224250793, + "learning_rate": 2.981674501419979e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010022000037133693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00461772158741951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024661721661686897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9768485248088836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5840, + "weighted_flow_matching_loss": 0.016754224617034196, + "weighted_rkd_loss": 0.024090772680938243 + }, + { + "epoch": 1.702065755018912, + "flow_matching_loss": 0.01593845384195447, + "grad_norm": 0.27846968173980713, + "learning_rate": 2.981545441625767e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009935825504362584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004712823661975563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024584474600851537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9767697155475616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5850, + "weighted_flow_matching_loss": 0.01593845384195447, + "weighted_rkd_loss": 0.024013359658420087 + }, + { + "epoch": 1.7049752691300553, + "flow_matching_loss": 0.014285833109170198, + "grad_norm": 0.2525630593299866, + "learning_rate": 2.9814159317771682e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01097814692184329, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004899016115814447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0268553102388978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9766907632350922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5860, + "weighted_flow_matching_loss": 0.014285833109170198, + "weighted_rkd_loss": 0.026229295879602432 + }, + { + "epoch": 1.7078847832411987, + "flow_matching_loss": 0.017010523844510315, + "grad_norm": 0.23893937468528748, + "learning_rate": 2.9812859719135236e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010214293655008078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004656570870429278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025085158180445433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9766117036342621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5870, + "weighted_flow_matching_loss": 0.017010523844510315, + "weighted_rkd_loss": 0.024498459696769715 + }, + { + "epoch": 1.7107942973523422, + "flow_matching_loss": 0.01703493148088455, + "grad_norm": 0.27012065052986145, + "learning_rate": 2.981155562074312e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010301128588616848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005091688386164606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02569394577294588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9765324890613556, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5880, + "weighted_flow_matching_loss": 0.01703493148088455, + "weighted_rkd_loss": 0.025090941041707993 + }, + { + "epoch": 1.7137038114634855, + "flow_matching_loss": 0.016573453694581984, + "grad_norm": 0.3032994270324707, + "learning_rate": 2.9810247022991483e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008998549822717906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041947057005018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02219180539250374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.976453173160553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5890, + "weighted_flow_matching_loss": 0.016573453694581984, + "weighted_rkd_loss": 0.021669242531061172 + }, + { + "epoch": 1.7166133255746292, + "flow_matching_loss": 0.016368495672941206, + "grad_norm": 0.2577274441719055, + "learning_rate": 2.9808933926277842e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009385054837912321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004396921768784523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023167031444609167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9763737142086029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5900, + "weighted_flow_matching_loss": 0.016368495672941206, + "weighted_rkd_loss": 0.02261976059526205 + }, + { + "epoch": 1.7195228396857725, + "flow_matching_loss": 0.015892619267106056, + "grad_norm": 0.23049040138721466, + "learning_rate": 2.9807616331001078e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009876437857747078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004744696221314371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024497572146356106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9762941241264343, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5910, + "weighted_flow_matching_loss": 0.015892619267106056, + "weighted_rkd_loss": 0.02391684800386429 + }, + { + "epoch": 1.722432353796916, + "flow_matching_loss": 0.016219139005988835, + "grad_norm": 0.25467929244041443, + "learning_rate": 2.9806294237561452e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010504578985273839, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00506090484559536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026070063561201097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9762143909931182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5920, + "weighted_flow_matching_loss": 0.016219139005988835, + "weighted_rkd_loss": 0.025449962168931962 + }, + { + "epoch": 1.7253418679080594, + "flow_matching_loss": 0.017282426357269287, + "grad_norm": 0.32481545209884644, + "learning_rate": 2.980496764636057e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009865832235664129, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004533000336959958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024264664761722086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9761345386505127, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5930, + "weighted_flow_matching_loss": 0.017282426357269287, + "weighted_rkd_loss": 0.023685581423342227 + }, + { + "epoch": 1.7282513820192027, + "flow_matching_loss": 0.01708027198910713, + "grad_norm": 0.28403645753860474, + "learning_rate": 2.9803636557801417e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009860913921147584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004533199570141732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024255027435719966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9760545551776886, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5940, + "weighted_flow_matching_loss": 0.01708027198910713, + "weighted_rkd_loss": 0.023674241080880166 + }, + { + "epoch": 1.7311608961303462, + "flow_matching_loss": 0.016236407961696387, + "grad_norm": 0.27801719307899475, + "learning_rate": 2.980230097228835e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009460871480405331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004739907430484891, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02366165053099394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.975974440574646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5950, + "weighted_flow_matching_loss": 0.016236407961696387, + "weighted_rkd_loss": 0.02309312503784895 + }, + { + "epoch": 1.7340704102414897, + "flow_matching_loss": 0.01604061173275113, + "grad_norm": 0.2623654901981354, + "learning_rate": 2.9800960890227076e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01017582924105227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00490072222892195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025252380594611167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9758942008018494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5960, + "weighted_flow_matching_loss": 0.01604061173275113, + "weighted_rkd_loss": 0.024643680825829507 + }, + { + "epoch": 1.7369799243526332, + "flow_matching_loss": 0.017401599884033205, + "grad_norm": 0.2539287805557251, + "learning_rate": 2.979961631202468e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009867012081667782, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00464655919931829, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02438058312982321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9758138358592987, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5970, + "weighted_flow_matching_loss": 0.017401599884033205, + "weighted_rkd_loss": 0.02379093263298273 + }, + { + "epoch": 1.7398894384637766, + "flow_matching_loss": 0.01550693791359663, + "grad_norm": 0.2960910201072693, + "learning_rate": 2.9798267238089613e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010229590255767107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004768451629206538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02522763218730688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9757333278656006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5980, + "weighted_flow_matching_loss": 0.01550693791359663, + "weighted_rkd_loss": 0.024615422263741492 + }, + { + "epoch": 1.74279895257492, + "flow_matching_loss": 0.01607047375291586, + "grad_norm": 0.23986704647541046, + "learning_rate": 2.9796913668831684e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010291899647563696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004568980657495558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025152779929339886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9756526827812195, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5990, + "weighted_flow_matching_loss": 0.01607047375291586, + "weighted_rkd_loss": 0.024540402740240098 + }, + { + "epoch": 1.7457084666860634, + "flow_matching_loss": 0.015077634993940593, + "grad_norm": 0.248790442943573, + "learning_rate": 2.979555560466207e-05, + "loss": 0.0396, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010168180940672756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00483530645724386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02517166845500469, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9755719125270843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6000, + "weighted_flow_matching_loss": 0.015077634993940593, + "weighted_rkd_loss": 0.024556737393140793 + }, + { + "epoch": 1.7486179807972069, + "flow_matching_loss": 0.014785032533109189, + "grad_norm": 0.23123109340667725, + "learning_rate": 2.9794193045993317e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009829327231273055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046413364354521034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024299991130828858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9754910230636596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6010, + "weighted_flow_matching_loss": 0.014785032533109189, + "weighted_rkd_loss": 0.023704473860561848 + }, + { + "epoch": 1.7515274949083504, + "flow_matching_loss": 0.015966060385107995, + "grad_norm": 0.26873645186424255, + "learning_rate": 2.9792825993239337e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009891259483993053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004799494426697493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024582013487815857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.975409984588623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6020, + "weighted_flow_matching_loss": 0.015966060385107995, + "weighted_rkd_loss": 0.02397751435637474 + }, + { + "epoch": 1.7544370090194938, + "flow_matching_loss": 0.014716342557221651, + "grad_norm": 0.22892434895038605, + "learning_rate": 2.9791454446815404e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009664166439324617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00465497188270092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023983304761350153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9753288447856903, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6030, + "weighted_flow_matching_loss": 0.014716342557221651, + "weighted_rkd_loss": 0.02339162528514862 + }, + { + "epoch": 1.757346523130637, + "flow_matching_loss": 0.015601846668869258, + "grad_norm": 0.2910483181476593, + "learning_rate": 2.9790078407138154e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011245938017964363, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005373169877566397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02786504626274109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9752475559711457, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6040, + "weighted_flow_matching_loss": 0.015601846668869258, + "weighted_rkd_loss": 0.02717527821660042 + }, + { + "epoch": 1.7602560372417806, + "flow_matching_loss": 0.016416683420538904, + "grad_norm": 0.286709189414978, + "learning_rate": 2.9788697874625588e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009450165089219809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004561441531404853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023461771570146082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9751661360263825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6050, + "weighted_flow_matching_loss": 0.016416683420538904, + "weighted_rkd_loss": 0.02287914287298918 + }, + { + "epoch": 1.763165551352924, + "flow_matching_loss": 0.014821319561451673, + "grad_norm": 0.22783686220645905, + "learning_rate": 2.9787312849697088e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01016847686842084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004572576633654535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02490953039377928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9750845730304718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6060, + "weighted_flow_matching_loss": 0.014821319561451673, + "weighted_rkd_loss": 0.02428888939321041 + }, + { + "epoch": 1.7660750654640673, + "flow_matching_loss": 0.017947612702846526, + "grad_norm": 0.3050621449947357, + "learning_rate": 2.978592333277338e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011314783804118633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005555104813538492, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028184672072529792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9750029027462006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6070, + "weighted_flow_matching_loss": 0.017947612702846526, + "weighted_rkd_loss": 0.027480106987059116 + }, + { + "epoch": 1.768984579575211, + "flow_matching_loss": 0.015716791711747648, + "grad_norm": 0.24856941401958466, + "learning_rate": 2.9784529324276557e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009352440712973475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00439438852481544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023099270090460776, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9749211013317108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6080, + "weighted_flow_matching_loss": 0.015716791711747648, + "weighted_rkd_loss": 0.02251999620348215 + }, + { + "epoch": 1.7718940936863543, + "flow_matching_loss": 0.015540011133998633, + "grad_norm": 0.2746303081512451, + "learning_rate": 2.9783130824630093e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00990780177526176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004702808079309761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024518411420285702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.974839162826538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6090, + "weighted_flow_matching_loss": 0.015540011133998633, + "weighted_rkd_loss": 0.02390152346342802 + }, + { + "epoch": 1.7748036077974978, + "flow_matching_loss": 0.0162017360329628, + "grad_norm": 0.246710866689682, + "learning_rate": 2.9781727834258806e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010314496699720621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004779854114167393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025408847257494926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9747570931911469, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6100, + "weighted_flow_matching_loss": 0.0162017360329628, + "weighted_rkd_loss": 0.024767486192286015 + }, + { + "epoch": 1.7777131219086413, + "flow_matching_loss": 0.013416532427072525, + "grad_norm": 0.27560776472091675, + "learning_rate": 2.97803203535889e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009920273767784238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004728748230263591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024569295533001422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9746748805046082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6110, + "weighted_flow_matching_loss": 0.013416532427072525, + "weighted_rkd_loss": 0.023947051353752614 + }, + { + "epoch": 1.7806226360197845, + "flow_matching_loss": 0.016651680693030357, + "grad_norm": 0.2757640480995178, + "learning_rate": 2.977890838304792e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009757869224995375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046862689545378085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024202007427811624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9745925664901733, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6120, + "weighted_flow_matching_loss": 0.016651680693030357, + "weighted_rkd_loss": 0.02358703836798668 + }, + { + "epoch": 1.7835321501309283, + "flow_matching_loss": 0.015304743405431509, + "grad_norm": 0.296671986579895, + "learning_rate": 2.9777491923064782e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010194363119080663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004913543025031686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02530226930975914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9745100975036621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6130, + "weighted_flow_matching_loss": 0.015304743405431509, + "weighted_rkd_loss": 0.02465730793774128 + }, + { + "epoch": 1.7864416642420715, + "flow_matching_loss": 0.01683132452890277, + "grad_norm": 0.28243109583854675, + "learning_rate": 2.977607097406978e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01024941699579358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004809132288210094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025307966582477092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9744275152683258, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6140, + "weighted_flow_matching_loss": 0.01683132452890277, + "weighted_rkd_loss": 0.02466079443693161 + }, + { + "epoch": 1.789351178353215, + "flow_matching_loss": 0.014257131610065699, + "grad_norm": 0.28242290019989014, + "learning_rate": 2.9774645536494552e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009524838346987963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004737461800687015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023787138424813746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9743447959423065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6150, + "weighted_flow_matching_loss": 0.014257131610065699, + "weighted_rkd_loss": 0.023176855966448785 + }, + { + "epoch": 1.7922606924643585, + "flow_matching_loss": 0.014538587909191847, + "grad_norm": 0.21161818504333496, + "learning_rate": 2.9773215610772117e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009991664486005903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004757033358328044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024740362353622914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9742619454860687, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6160, + "weighted_flow_matching_loss": 0.014538587909191847, + "weighted_rkd_loss": 0.02410355918109417 + }, + { + "epoch": 1.7951702065755017, + "flow_matching_loss": 0.014967251848429442, + "grad_norm": 0.24007782340049744, + "learning_rate": 2.9771781197336837e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009568548994138837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004504567477852106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023641665279865266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9741789698600769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6170, + "weighted_flow_matching_loss": 0.014967251848429442, + "weighted_rkd_loss": 0.02303117122501135 + }, + { + "epoch": 1.7980797206866455, + "flow_matching_loss": 0.015823858696967362, + "grad_norm": 0.2550787627696991, + "learning_rate": 2.977034229662446e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009553584037348628, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004453298728913068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02356046698987484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9740958452224732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6180, + "weighted_flow_matching_loss": 0.015823858696967362, + "weighted_rkd_loss": 0.022950175404548644 + }, + { + "epoch": 1.8009892347977887, + "flow_matching_loss": 0.014575982745736838, + "grad_norm": 0.21718959510326385, + "learning_rate": 2.976889890907208e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009614462777972222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004490429954603315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023719355277717114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9740126132965088, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6190, + "weighted_flow_matching_loss": 0.014575982745736838, + "weighted_rkd_loss": 0.023102983087301254 + }, + { + "epoch": 1.8038987489089322, + "flow_matching_loss": 0.015587755758315325, + "grad_norm": 0.25262922048568726, + "learning_rate": 2.9767451035118164e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009551443764939905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044112668838351965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02351415455341339, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973929226398468, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6200, + "weighted_flow_matching_loss": 0.015587755758315325, + "weighted_rkd_loss": 0.022901079431176186 + }, + { + "epoch": 1.8068082630200757, + "flow_matching_loss": 0.015173747204244137, + "grad_norm": 0.28351494669914246, + "learning_rate": 2.9765998675202536e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009584170673042536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004392954451031983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023561296053230762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9738457441329956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6210, + "weighted_flow_matching_loss": 0.015173747204244137, + "weighted_rkd_loss": 0.0229450486600399 + }, + { + "epoch": 1.809717777131219, + "flow_matching_loss": 0.014916244428604841, + "grad_norm": 0.2593556344509125, + "learning_rate": 2.9764541829766382e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008678568713366985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042015281040221454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021558665670454502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9737621128559113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6220, + "weighted_flow_matching_loss": 0.014916244428604841, + "weighted_rkd_loss": 0.020992994867265226 + }, + { + "epoch": 1.8126272912423627, + "flow_matching_loss": 0.015082623530179261, + "grad_norm": 0.21824122965335846, + "learning_rate": 2.9763080499252253e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009618757292628288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004581079981289804, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023818594962358476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9736783623695373, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6230, + "weighted_flow_matching_loss": 0.015082623530179261, + "weighted_rkd_loss": 0.02319166176021099 + }, + { + "epoch": 1.815536805353506, + "flow_matching_loss": 0.014880022685974836, + "grad_norm": 0.23212012648582458, + "learning_rate": 2.9761614684104064e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009233899926766754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004218098684214055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022685898654162883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973594468832016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6240, + "weighted_flow_matching_loss": 0.014880022685974836, + "weighted_rkd_loss": 0.022086855210363864 + }, + { + "epoch": 1.8184463194646494, + "flow_matching_loss": 0.015080843307077884, + "grad_norm": 0.2179505079984665, + "learning_rate": 2.9760144384767095e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009195305546745658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004414147860370576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022804758884012698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973510456085205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6250, + "weighted_flow_matching_loss": 0.015080843307077884, + "weighted_rkd_loss": 0.022200648486614228 + }, + { + "epoch": 1.821355833575793, + "flow_matching_loss": 0.015534594282507896, + "grad_norm": 0.31799694895744324, + "learning_rate": 2.9758669601687986e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010440502315759659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004840079718269408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025721084140241147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9734263002872467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6260, + "weighted_flow_matching_loss": 0.015534594282507896, + "weighted_rkd_loss": 0.025037580542266367 + }, + { + "epoch": 1.8242653476869362, + "flow_matching_loss": 0.013545552641153336, + "grad_norm": 0.25426143407821655, + "learning_rate": 2.9757190335314722e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00998372696340084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004857541620731353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02482499536126852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9733420252799988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6270, + "weighted_flow_matching_loss": 0.013545552641153336, + "weighted_rkd_loss": 0.024163195863366127 + }, + { + "epoch": 1.8271748617980799, + "flow_matching_loss": 0.015215404238551855, + "grad_norm": 0.260172575712204, + "learning_rate": 2.975570658609668e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00993276946246624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004524811683222651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02439035065472126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9732576131820678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6280, + "weighted_flow_matching_loss": 0.015215404238551855, + "weighted_rkd_loss": 0.023738092742860317 + }, + { + "epoch": 1.8300843759092231, + "flow_matching_loss": 0.01477318610996008, + "grad_norm": 0.22411537170410156, + "learning_rate": 2.9754218354484582e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008437740337103605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003996246424503624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020871727168560027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973173075914383, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6290, + "weighted_flow_matching_loss": 0.01477318610996008, + "weighted_rkd_loss": 0.020311796851456166 + }, + { + "epoch": 1.8329938900203666, + "flow_matching_loss": 0.01557460641488433, + "grad_norm": 0.2338072806596756, + "learning_rate": 2.9752725640930504e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008590241940692068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003996433853171766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021176917664706706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973088413476944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6300, + "weighted_flow_matching_loss": 0.01557460641488433, + "weighted_rkd_loss": 0.020607028901576997 + }, + { + "epoch": 1.83590340413151, + "flow_matching_loss": 0.016000485606491567, + "grad_norm": 0.26019126176834106, + "learning_rate": 2.9751228445887905e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009442441072314978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004674506885930896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02355938907712698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9730036079883575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6310, + "weighted_flow_matching_loss": 0.016000485606491567, + "weighted_rkd_loss": 0.02292337492108345 + }, + { + "epoch": 1.8388129182426534, + "flow_matching_loss": 0.014819631911814212, + "grad_norm": 0.2497170865535736, + "learning_rate": 2.9749726769811588e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009457009518519044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004240534431301057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02315455339848995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9729186773300171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6320, + "weighted_flow_matching_loss": 0.014819631911814212, + "weighted_rkd_loss": 0.022527495212852956 + }, + { + "epoch": 1.841722432353797, + "flow_matching_loss": 0.017328306566923857, + "grad_norm": 0.27962055802345276, + "learning_rate": 2.9748220613157717e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009112284146249294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004244274739176035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022468843124806882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9728336274623871, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6330, + "weighted_flow_matching_loss": 0.017328306566923857, + "weighted_rkd_loss": 0.021858416870236395 + }, + { + "epoch": 1.8446319464649403, + "flow_matching_loss": 0.01441214932128787, + "grad_norm": 0.25686267018318176, + "learning_rate": 2.9746709976383832e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008908756216987968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004400512436404824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02221802473068237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9727484285831451, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6340, + "weighted_flow_matching_loss": 0.01441214932128787, + "weighted_rkd_loss": 0.02161253457888961 + }, + { + "epoch": 1.8475414605760838, + "flow_matching_loss": 0.01693856744095683, + "grad_norm": 0.2589597702026367, + "learning_rate": 2.974519485994882e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009578121546655893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004594463109970093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02375070620328188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9726631224155426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6350, + "weighted_flow_matching_loss": 0.01693856744095683, + "weighted_rkd_loss": 0.023101423308253287 + }, + { + "epoch": 1.8504509746872273, + "flow_matching_loss": 0.016537949349731206, + "grad_norm": 0.26077598333358765, + "learning_rate": 2.9743675264312934e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01052542170509696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004689330374822021, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025740174017846583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9725776791572571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6360, + "weighted_flow_matching_loss": 0.016537949349731206, + "weighted_rkd_loss": 0.02503432650119066 + }, + { + "epoch": 1.8533604887983706, + "flow_matching_loss": 0.015613204706460238, + "grad_norm": 0.30482274293899536, + "learning_rate": 2.9742151189937784e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009506537206470966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00429082999471575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02330390438437462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9724920988082886, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6370, + "weighted_flow_matching_loss": 0.015613204706460238, + "weighted_rkd_loss": 0.022662882506847382 + }, + { + "epoch": 1.8562700029095143, + "flow_matching_loss": 0.01561249690130353, + "grad_norm": 0.24684220552444458, + "learning_rate": 2.974062263728635e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008555841306224466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004010263923555612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021121946536004542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9724063873291016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6380, + "weighted_flow_matching_loss": 0.01561249690130353, + "weighted_rkd_loss": 0.020539114251732826 + }, + { + "epoch": 1.8591795170206575, + "flow_matching_loss": 0.015711492486298086, + "grad_norm": 0.31714221835136414, + "learning_rate": 2.973908960682296e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009575933683663607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004582341434434057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023734208755195142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9723205506801605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6390, + "weighted_flow_matching_loss": 0.015711492486298086, + "weighted_rkd_loss": 0.023077249899506568 + }, + { + "epoch": 1.862089031131801, + "flow_matching_loss": 0.01701487898826599, + "grad_norm": 0.2902095913887024, + "learning_rate": 2.9737552099013313e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009017643006518482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004250901751220226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022286187671124936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.97223459482193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6400, + "weighted_flow_matching_loss": 0.01701487898826599, + "weighted_rkd_loss": 0.021667441911995412 + }, + { + "epoch": 1.8649985452429445, + "flow_matching_loss": 0.014295717515051366, + "grad_norm": 0.2659444808959961, + "learning_rate": 2.9736010114324462e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00935589475557208, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004362650774419307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023074440099298953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9721485018730164, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6410, + "weighted_flow_matching_loss": 0.014295717515051366, + "weighted_rkd_loss": 0.02243180200457573 + }, + { + "epoch": 1.8679080593540878, + "flow_matching_loss": 0.015283175464719533, + "grad_norm": 0.2744116187095642, + "learning_rate": 2.9734463653224813e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01018256451934576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004741389257833362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025106518156826498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9720622777938843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6420, + "weighted_flow_matching_loss": 0.015283175464719533, + "weighted_rkd_loss": 0.024405128322541714 + }, + { + "epoch": 1.8708175734652313, + "flow_matching_loss": 0.017166535556316375, + "grad_norm": 0.26245588064193726, + "learning_rate": 2.9732912716184155e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009556342381983995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004313371982425452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0234260568395257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9719759404659272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6430, + "weighted_flow_matching_loss": 0.017166535556316375, + "weighted_rkd_loss": 0.02276954483240843 + }, + { + "epoch": 1.8737270875763747, + "flow_matching_loss": 0.015501542296260596, + "grad_norm": 0.27370861172676086, + "learning_rate": 2.9731357303673607e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009945720620453357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047753539867699145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024666795134544374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.971889442205429, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6440, + "weighted_flow_matching_loss": 0.015501542296260596, + "weighted_rkd_loss": 0.023973411321640013 + }, + { + "epoch": 1.8766366016875182, + "flow_matching_loss": 0.014895827136933803, + "grad_norm": 0.24585901200771332, + "learning_rate": 2.9729797416165677e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00836603636853397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003903323062695563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020635395869612695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9718028366565704, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6450, + "weighted_flow_matching_loss": 0.014895827136933803, + "weighted_rkd_loss": 0.02005354966968298 + }, + { + "epoch": 1.8795461157986617, + "flow_matching_loss": 0.0147015867754817, + "grad_norm": 0.24782361090183258, + "learning_rate": 2.9728233054134202e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00934792528860271, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004176377062685787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02287222780287266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9717160880565643, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6460, + "weighted_flow_matching_loss": 0.0147015867754817, + "weighted_rkd_loss": 0.022225321270525455 + }, + { + "epoch": 1.882455629909805, + "flow_matching_loss": 0.016201211884617806, + "grad_norm": 0.24344182014465332, + "learning_rate": 2.9726664218054405e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00936891851015389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004339480586349964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02307731769979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9716292321681976, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6470, + "weighted_flow_matching_loss": 0.016201211884617806, + "weighted_rkd_loss": 0.022422605007886887 + }, + { + "epoch": 1.8853651440209485, + "flow_matching_loss": 0.015385365206748247, + "grad_norm": 0.2870752215385437, + "learning_rate": 2.9725090908402856e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009937100671231746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004987783683463931, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024861984699964524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.971542227268219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6480, + "weighted_flow_matching_loss": 0.015385365206748247, + "weighted_rkd_loss": 0.024154504388570787 + }, + { + "epoch": 1.888274658132092, + "flow_matching_loss": 0.015713028889149428, + "grad_norm": 0.2813125550746918, + "learning_rate": 2.972351312565748e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009722680691629649, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004612308135256172, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024057669565081596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9714551031589508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6490, + "weighted_flow_matching_loss": 0.015713028889149428, + "weighted_rkd_loss": 0.023370900191366673 + }, + { + "epoch": 1.8911841722432354, + "flow_matching_loss": 0.01548419166356325, + "grad_norm": 0.28429508209228516, + "learning_rate": 2.9721930870297567e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010715754656121134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005073684453964234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026505194045603274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9713678359985352, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6500, + "weighted_flow_matching_loss": 0.01548419166356325, + "weighted_rkd_loss": 0.02574629709124565 + }, + { + "epoch": 1.894093686354379, + "flow_matching_loss": 0.016495776642113925, + "grad_norm": 0.24182897806167603, + "learning_rate": 2.9720344142803766e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010406632721424103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004967381106689572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025780646689236165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.97128044962883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6510, + "weighted_flow_matching_loss": 0.016495776642113925, + "weighted_rkd_loss": 0.02504024710506201 + }, + { + "epoch": 1.8970032004655222, + "flow_matching_loss": 0.01595498016104102, + "grad_norm": 0.29002365469932556, + "learning_rate": 2.9718752943658085e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009414530918002128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004440881218761206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023269942961633205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9711929261684418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6520, + "weighted_flow_matching_loss": 0.01595498016104102, + "weighted_rkd_loss": 0.022599578835070133 + }, + { + "epoch": 1.8999127145766657, + "flow_matching_loss": 0.015185806434601545, + "grad_norm": 0.2653600573539734, + "learning_rate": 2.9717157273343885e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008954695519059896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004271752596832812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022181143797934056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9711052894592285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6530, + "weighted_flow_matching_loss": 0.015185806434601545, + "weighted_rkd_loss": 0.021540215983986855 + }, + { + "epoch": 1.9028222286878091, + "flow_matching_loss": 0.014621087722480297, + "grad_norm": 0.2870291769504547, + "learning_rate": 2.971555713234589e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010572974290698766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005139599926769734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026285548508167268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9710175156593323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6540, + "weighted_flow_matching_loss": 0.014621087722480297, + "weighted_rkd_loss": 0.025523700937628745 + }, + { + "epoch": 1.9057317427989524, + "flow_matching_loss": 0.01546249222010374, + "grad_norm": 0.25499048829078674, + "learning_rate": 2.9713952521150176e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010524574853479862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005027992045506835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026077141612768175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.970929616689682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6550, + "weighted_flow_matching_loss": 0.01546249222010374, + "weighted_rkd_loss": 0.025319065153598785 + }, + { + "epoch": 1.9086412569100961, + "flow_matching_loss": 0.016827996261417864, + "grad_norm": 0.19865551590919495, + "learning_rate": 2.971234344024419e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008213498443365098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038364206906408072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02026341762393713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9708415806293488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6560, + "weighted_flow_matching_loss": 0.016827996261417864, + "weighted_rkd_loss": 0.01967259291559458 + }, + { + "epoch": 1.9115507710212394, + "flow_matching_loss": 0.01494403900578618, + "grad_norm": 0.24818409979343414, + "learning_rate": 2.971072989011672e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009628817904740572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004440847760997712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023698483407497407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9707534313201904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6570, + "weighted_flow_matching_loss": 0.01494403900578618, + "weighted_rkd_loss": 0.02300535887479782 + }, + { + "epoch": 1.9144602851323829, + "flow_matching_loss": 0.015884912759065627, + "grad_norm": 0.260219544172287, + "learning_rate": 2.9709111871257927e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010425819177180528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004769248515367508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025620886869728566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9706651270389557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6580, + "weighted_flow_matching_loss": 0.015884912759065627, + "weighted_rkd_loss": 0.024869347363710402 + }, + { + "epoch": 1.9173697992435264, + "flow_matching_loss": 0.015044184681028128, + "grad_norm": 0.30024710297584534, + "learning_rate": 2.9707489384159318e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009998711198568344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048771381843835115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0248745609074831, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9705767035484314, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6590, + "weighted_flow_matching_loss": 0.015044184681028128, + "weighted_rkd_loss": 0.024142602272331715 + }, + { + "epoch": 1.9202793133546696, + "flow_matching_loss": 0.01588432341814041, + "grad_norm": 0.24514208734035492, + "learning_rate": 2.9705862429313763e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009581303037703038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004415233759209514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023577839881181718, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9704881608486176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6600, + "weighted_flow_matching_loss": 0.01588432341814041, + "weighted_rkd_loss": 0.022882049903273582 + }, + { + "epoch": 1.9231888274658133, + "flow_matching_loss": 0.015809511300176382, + "grad_norm": 0.2683047652244568, + "learning_rate": 2.9704231007215488e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008831934770569205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004165814747102559, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02182968445122242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9703994870185852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6610, + "weighted_flow_matching_loss": 0.015809511300176382, + "weighted_rkd_loss": 0.021183496993035077 + }, + { + "epoch": 1.9260983415769566, + "flow_matching_loss": 0.014903751295059919, + "grad_norm": 0.3179374635219574, + "learning_rate": 2.9702595118360077e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009011352434754372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004284603102132678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02230730764567852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9703106820583344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6620, + "weighted_flow_matching_loss": 0.014903751295059919, + "weighted_rkd_loss": 0.021644986048340798 + }, + { + "epoch": 1.9290078556881, + "flow_matching_loss": 0.017050268687307836, + "grad_norm": 0.28156113624572754, + "learning_rate": 2.9700954763244465e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008725520689040422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004199460637755692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021650501899421214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9702217400074005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6630, + "weighted_flow_matching_loss": 0.017050268687307836, + "weighted_rkd_loss": 0.02100577186793089 + }, + { + "epoch": 1.9319173697992436, + "flow_matching_loss": 0.013745280634611844, + "grad_norm": 0.29055625200271606, + "learning_rate": 2.9699309942366953e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009301518695428968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004210395691916347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022813433408737184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9701326727867127, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6640, + "weighted_flow_matching_loss": 0.013745280634611844, + "weighted_rkd_loss": 0.02213207557797432 + }, + { + "epoch": 1.9348268839103868, + "flow_matching_loss": 0.01658735442906618, + "grad_norm": 0.22677116096019745, + "learning_rate": 2.969766065622719e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00879402463324368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004214522358961403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0218025716021657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9700434803962708, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6650, + "weighted_flow_matching_loss": 0.01658735442906618, + "weighted_rkd_loss": 0.02114943154156208 + }, + { + "epoch": 1.9377363980215305, + "flow_matching_loss": 0.015225237980484963, + "grad_norm": 0.24736584722995758, + "learning_rate": 2.96960069053262e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009336192719638347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00439123110845685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02306361645460129, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9699541628360748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6660, + "weighted_flow_matching_loss": 0.015225237980484963, + "weighted_rkd_loss": 0.022370684891939163 + }, + { + "epoch": 1.9406459121326738, + "flow_matching_loss": 0.015811032522469758, + "grad_norm": 0.24543538689613342, + "learning_rate": 2.9694348690166327e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00872745392844081, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004062827513553202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021517734974622726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9698647022247314, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6670, + "weighted_flow_matching_loss": 0.015811032522469758, + "weighted_rkd_loss": 0.020869305357336998 + }, + { + "epoch": 1.9435554262438173, + "flow_matching_loss": 0.017815256770700217, + "grad_norm": 0.28058329224586487, + "learning_rate": 2.969268601125131e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009252038318663836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00418920440133661, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022693281061947346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9697751224040985, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6680, + "weighted_flow_matching_loss": 0.017815256770700217, + "weighted_rkd_loss": 0.022007356956601144 + }, + { + "epoch": 1.9464649403549608, + "flow_matching_loss": 0.015417904406785966, + "grad_norm": 0.24071775376796722, + "learning_rate": 2.969101886908622e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008123451750725508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039002982433885337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02014720169827342, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9696854233741761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6690, + "weighted_flow_matching_loss": 0.015417904406785966, + "weighted_rkd_loss": 0.019536447897553443 + }, + { + "epoch": 1.949374454466104, + "flow_matching_loss": 0.015961439441889526, + "grad_norm": 0.27665194869041443, + "learning_rate": 2.968934726417749e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010349937621504069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004753590119071305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025453465431928633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9695955812931061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6700, + "weighted_flow_matching_loss": 0.015961439441889526, + "weighted_rkd_loss": 0.024679509177803993 + }, + { + "epoch": 1.9522839685772477, + "flow_matching_loss": 0.015425755828619003, + "grad_norm": 0.31035393476486206, + "learning_rate": 2.968767119703291e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009833405259996653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004421438067220151, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02408824861049652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9695056259632111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6710, + "weighted_flow_matching_loss": 0.015425755828619003, + "weighted_rkd_loss": 0.023353716358542442 + }, + { + "epoch": 1.955193482688391, + "flow_matching_loss": 0.01670852638781071, + "grad_norm": 0.36933615803718567, + "learning_rate": 2.9685990668161628e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009744721464812756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004601724026724696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02409116718918085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9694155216217041, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6720, + "weighted_flow_matching_loss": 0.01670852638781071, + "weighted_rkd_loss": 0.023354312218725682 + }, + { + "epoch": 1.9581029967995345, + "flow_matching_loss": 0.01426249099895358, + "grad_norm": 0.2776140868663788, + "learning_rate": 2.9684305678074137e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009003191022202373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004305842355825007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022312224097549915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9693252980709076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6730, + "weighted_flow_matching_loss": 0.01426249099895358, + "weighted_rkd_loss": 0.021627813577651978 + }, + { + "epoch": 1.961012510910678, + "flow_matching_loss": 0.01603299705311656, + "grad_norm": 0.25859329104423523, + "learning_rate": 2.9682616227282304e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008899574168026447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045874809147790074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02238662913441658, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.969234949350357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6740, + "weighted_flow_matching_loss": 0.01603299705311656, + "weighted_rkd_loss": 0.02169789783656597 + }, + { + "epoch": 1.9639220250218212, + "flow_matching_loss": 0.016134611144661904, + "grad_norm": 0.25196540355682373, + "learning_rate": 2.968092231629933e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00869304183870554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004220852558501065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021606936119496822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9691444575786591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6750, + "weighted_flow_matching_loss": 0.016134611144661904, + "weighted_rkd_loss": 0.020940236374735834 + }, + { + "epoch": 1.966831539132965, + "flow_matching_loss": 0.01568955313414335, + "grad_norm": 0.287132203578949, + "learning_rate": 2.967922394563978e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009570077247917651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044150118250399824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023555166274309158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.969053852558136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6760, + "weighted_flow_matching_loss": 0.01568955313414335, + "weighted_rkd_loss": 0.02282620957121253 + }, + { + "epoch": 1.9697410532441082, + "flow_matching_loss": 0.01529375957325101, + "grad_norm": 0.26037517189979553, + "learning_rate": 2.9677521115819585e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009966526553034783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044878642074763775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0244209174066782, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9689631164073944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6770, + "weighted_flow_matching_loss": 0.01529375957325101, + "weighted_rkd_loss": 0.023662971891462804 + }, + { + "epoch": 1.9726505673552517, + "flow_matching_loss": 0.015240975562483073, + "grad_norm": 0.21747931838035583, + "learning_rate": 2.967581382735601e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009411730710417033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004220910114236176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023044371604919435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9688722431659699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6780, + "weighted_flow_matching_loss": 0.015240975562483073, + "weighted_rkd_loss": 0.022327067516744136 + }, + { + "epoch": 1.9755600814663952, + "flow_matching_loss": 0.015880063734948636, + "grad_norm": 0.29544103145599365, + "learning_rate": 2.9674102080767685e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009546031104400754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004436412453651428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02352847419679165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9687812566757202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6790, + "weighted_flow_matching_loss": 0.015880063734948636, + "weighted_rkd_loss": 0.022793937474489212 + }, + { + "epoch": 1.9784695955775384, + "flow_matching_loss": 0.015506590157747269, + "grad_norm": 0.2612156271934509, + "learning_rate": 2.9672385876574597e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009529721830040217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004508869862183929, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02356831356883049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9686901271343231, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6800, + "weighted_flow_matching_loss": 0.015506590157747269, + "weighted_rkd_loss": 0.022830374538898468 + }, + { + "epoch": 1.9813791096886821, + "flow_matching_loss": 0.016948509402573107, + "grad_norm": 0.21899975836277008, + "learning_rate": 2.967066521529808e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010588317038491368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005169051908887922, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026345686055719852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9685988783836365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6810, + "weighted_flow_matching_loss": 0.016948509402573107, + "weighted_rkd_loss": 0.025518437661230565 + }, + { + "epoch": 1.9842886237998254, + "flow_matching_loss": 0.0157421350479126, + "grad_norm": 0.25707104802131653, + "learning_rate": 2.966894009746083e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009892819821834565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047030591173097495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024488698691129684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9685074925422669, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6820, + "weighted_flow_matching_loss": 0.0157421350479126, + "weighted_rkd_loss": 0.02371752355247736 + }, + { + "epoch": 1.9871981379109689, + "flow_matching_loss": 0.01521225729957223, + "grad_norm": 0.27428045868873596, + "learning_rate": 2.9667210523586888e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00898303436115384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004356921347789466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022322990000247955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9684159815311432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6830, + "weighted_flow_matching_loss": 0.01521225729957223, + "weighted_rkd_loss": 0.02161792851984501 + }, + { + "epoch": 1.9901076520221124, + "flow_matching_loss": 0.016126382723450662, + "grad_norm": 0.26686686277389526, + "learning_rate": 2.966547649420165e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00938558615744114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004465580778196454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023236753046512605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9683243453502655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6840, + "weighted_flow_matching_loss": 0.016126382723450662, + "weighted_rkd_loss": 0.022500685788691042 + }, + { + "epoch": 1.9930171661332556, + "flow_matching_loss": 0.015873071737587452, + "grad_norm": 0.2698989510536194, + "learning_rate": 2.9663738009831877e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009114963607862592, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004250459372997284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02248038649559021, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9682325720787048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6850, + "weighted_flow_matching_loss": 0.015873071737587452, + "weighted_rkd_loss": 0.021766234561800955 + }, + { + "epoch": 1.9959266802443993, + "flow_matching_loss": 0.01543284598737955, + "grad_norm": 0.2317487746477127, + "learning_rate": 2.966199507100567e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009140518307685853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004418553435243666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02269959021359682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9681406855583191, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6860, + "weighted_flow_matching_loss": 0.01543284598737955, + "weighted_rkd_loss": 0.02197640500962734 + }, + { + "epoch": 1.9988361943555426, + "flow_matching_loss": 0.014575435593724251, + "grad_norm": 0.33275195956230164, + "learning_rate": 2.9660247678252484e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010422384086996316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004913370148278773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02575813848525286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9680486619472504, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6870, + "weighted_flow_matching_loss": 0.014575435593724251, + "weighted_rkd_loss": 0.024935150146484376 + }, + { + "epoch": 2.001745708466686, + "flow_matching_loss": 0.017009069491177798, + "grad_norm": 0.2666349411010742, + "learning_rate": 2.965849583210314e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009228843497112393, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043805979657918215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022838285006582738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9679565131664276, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6880, + "weighted_flow_matching_loss": 0.017009069491177798, + "weighted_rkd_loss": 0.022106486558914184 + }, + { + "epoch": 2.0046552225778296, + "flow_matching_loss": 0.0159749211743474, + "grad_norm": 0.25090348720550537, + "learning_rate": 2.9656739533089794e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009237935626879334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004334510955959559, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022810382209718226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9678642332553864, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6890, + "weighted_flow_matching_loss": 0.0159749211743474, + "weighted_rkd_loss": 0.022077380679547787 + }, + { + "epoch": 2.007564736688973, + "flow_matching_loss": 0.014771359506994486, + "grad_norm": 0.2669006884098053, + "learning_rate": 2.9654978781745968e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00868907943367958, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004031210578978062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02140936944633722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9677718281745911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6900, + "weighted_flow_matching_loss": 0.014771359506994486, + "weighted_rkd_loss": 0.020719356276094912 + }, + { + "epoch": 2.0104742508001165, + "flow_matching_loss": 0.016152662876993417, + "grad_norm": 0.2632644474506378, + "learning_rate": 2.9653213578606534e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008597496012225748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003948796587064862, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021143788285553456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9676792860031128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6910, + "weighted_flow_matching_loss": 0.016152662876993417, + "weighted_rkd_loss": 0.02046039067208767 + }, + { + "epoch": 2.01338376491126, + "flow_matching_loss": 0.015960295870900153, + "grad_norm": 0.2942262589931488, + "learning_rate": 2.9651443924207704e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009354142751544715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004550224845297634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02325851023197174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9675866305828095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6920, + "weighted_flow_matching_loss": 0.015960295870900153, + "weighted_rkd_loss": 0.022504595387727023 + }, + { + "epoch": 2.016293279022403, + "flow_matching_loss": 0.01467772601172328, + "grad_norm": 0.26057446002960205, + "learning_rate": 2.964966981908706e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009383555315434933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00431686791125685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023083978705108164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9674938321113586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6930, + "weighted_flow_matching_loss": 0.01467772601172328, + "weighted_rkd_loss": 0.022333642840385436 + }, + { + "epoch": 2.019202793133547, + "flow_matching_loss": 0.015590764489024878, + "grad_norm": 0.28841009736061096, + "learning_rate": 2.964789126378353e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008726380905136466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004001471167430282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021454232931137084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9674009084701538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6940, + "weighted_flow_matching_loss": 0.015590764489024878, + "weighted_rkd_loss": 0.020754813961684705 + }, + { + "epoch": 2.02211230724469, + "flow_matching_loss": 0.016402169968932866, + "grad_norm": 0.3067300319671631, + "learning_rate": 2.9646108258837388e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009170599281787872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042913539102301, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022632552683353423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9673078656196594, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6950, + "weighted_flow_matching_loss": 0.016402169968932866, + "weighted_rkd_loss": 0.021892659738659857 + }, + { + "epoch": 2.0250218213558338, + "flow_matching_loss": 0.014994826074689627, + "grad_norm": 0.26797932386398315, + "learning_rate": 2.9644320804790267e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00907712015323341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003999138600192964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022153378650546073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9672146916389466, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6960, + "weighted_flow_matching_loss": 0.014994826074689627, + "weighted_rkd_loss": 0.021427078172564507 + }, + { + "epoch": 2.027931335466977, + "flow_matching_loss": 0.014565586857497693, + "grad_norm": 0.233096182346344, + "learning_rate": 2.9642528902185142e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008573372289538384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038349613780155776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02098170556128025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9671213805675507, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6970, + "weighted_flow_matching_loss": 0.014565586857497693, + "weighted_rkd_loss": 0.02029182929545641 + }, + { + "epoch": 2.0308408495781203, + "flow_matching_loss": 0.016518320143222808, + "grad_norm": 0.2160651981830597, + "learning_rate": 2.9640732551566354e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008990080887451768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004188111145049333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022168273106217384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9670279562473297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6980, + "weighted_flow_matching_loss": 0.016518320143222808, + "weighted_rkd_loss": 0.021437456645071507 + }, + { + "epoch": 2.033750363689264, + "flow_matching_loss": 0.014065660163760185, + "grad_norm": 0.22274291515350342, + "learning_rate": 2.9638931753479578e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008902748534455895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003957556607201695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0217630535364151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9669343948364257, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6990, + "weighted_flow_matching_loss": 0.014065660163760185, + "weighted_rkd_loss": 0.021043462306261064 + }, + { + "epoch": 2.0366598778004072, + "flow_matching_loss": 0.015311667043715716, + "grad_norm": 0.21895506978034973, + "learning_rate": 2.9637126508471858e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009163614502176643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004425217118114233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022752446122467518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9668406963348388, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7000, + "weighted_flow_matching_loss": 0.015311667043715716, + "weighted_rkd_loss": 0.021998011507093906 + }, + { + "epoch": 2.039569391911551, + "flow_matching_loss": 0.013314617238938809, + "grad_norm": 0.2744390666484833, + "learning_rate": 2.9635316817091577e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009165288275107741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004252132284455002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022582709230482577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9667468786239624, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7010, + "weighted_flow_matching_loss": 0.013314617238938809, + "weighted_rkd_loss": 0.021831716038286687 + }, + { + "epoch": 2.042478906022694, + "flow_matching_loss": 0.01503255432471633, + "grad_norm": 0.23413516581058502, + "learning_rate": 2.9633502679888467e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008633951470255852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003925678762607276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021193581726402046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9666529357433319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7020, + "weighted_flow_matching_loss": 0.01503255432471633, + "weighted_rkd_loss": 0.020486854296177626 + }, + { + "epoch": 2.0453884201338375, + "flow_matching_loss": 0.016566891223192215, + "grad_norm": 0.25213634967803955, + "learning_rate": 2.963168409741362e-05, + "loss": 0.0396, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009624146344140171, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004534307308495045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02378259990364313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9665588676929474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7030, + "weighted_flow_matching_loss": 0.016566891223192215, + "weighted_rkd_loss": 0.02298722416162491 + }, + { + "epoch": 2.048297934244981, + "flow_matching_loss": 0.014432305470108986, + "grad_norm": 0.24463823437690735, + "learning_rate": 2.9629861070219467e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009171195700764656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004279614798724652, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022622006386518477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9664646685123444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7040, + "weighted_flow_matching_loss": 0.014432305470108986, + "weighted_rkd_loss": 0.02186336014419794 + }, + { + "epoch": 2.0512074483561245, + "flow_matching_loss": 0.015681408159434796, + "grad_norm": 0.26108062267303467, + "learning_rate": 2.9628033598859807e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00928059802390635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004636541754007339, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023197737894952296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9663703262805938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7050, + "weighted_flow_matching_loss": 0.015681408159434796, + "weighted_rkd_loss": 0.022417615354061126 + }, + { + "epoch": 2.054116962467268, + "flow_matching_loss": 0.01537886904552579, + "grad_norm": 0.24546298384666443, + "learning_rate": 2.9626201683889764e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00898818951100111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004191708099097013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022168087027966975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9662758767604828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7060, + "weighted_flow_matching_loss": 0.01537886904552579, + "weighted_rkd_loss": 0.021420486457645892 + }, + { + "epoch": 2.0570264765784114, + "flow_matching_loss": 0.01540916096419096, + "grad_norm": 0.2686702311038971, + "learning_rate": 2.9624365325865835e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010399426333606243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00473602875135839, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025534881465137004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9661812901496887, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7070, + "weighted_flow_matching_loss": 0.01540916096419096, + "weighted_rkd_loss": 0.024671308137476443 + }, + { + "epoch": 2.0599359906895547, + "flow_matching_loss": 0.015412051882594823, + "grad_norm": 0.2452821582555771, + "learning_rate": 2.962252452534585e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009006883157417179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004469330725260079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022483097203075886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9660865962505341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7080, + "weighted_flow_matching_loss": 0.015412051882594823, + "weighted_rkd_loss": 0.021720581874251365 + }, + { + "epoch": 2.0628455048006984, + "flow_matching_loss": 0.016964547988027334, + "grad_norm": 0.2694118022918701, + "learning_rate": 2.9620679282889006e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009583312645554542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00444131987169385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023607945069670677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.965991735458374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7090, + "weighted_flow_matching_loss": 0.016964547988027334, + "weighted_rkd_loss": 0.022805065102875232 + }, + { + "epoch": 2.0657550189118417, + "flow_matching_loss": 0.015154062863439321, + "grad_norm": 0.2764928936958313, + "learning_rate": 2.9618829599055833e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009370063431560993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004576708609238267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02331683561205864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9658967733383179, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7100, + "weighted_flow_matching_loss": 0.015154062863439321, + "weighted_rkd_loss": 0.022521655820310117 + }, + { + "epoch": 2.0686645330229854, + "flow_matching_loss": 0.014866102393716574, + "grad_norm": 0.23176954686641693, + "learning_rate": 2.961697547440821e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008315931679680944, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041263469262048606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020758210495114327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9658016741275788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7110, + "weighted_flow_matching_loss": 0.014866102393716574, + "weighted_rkd_loss": 0.020048279128968716 + }, + { + "epoch": 2.0715740471341286, + "flow_matching_loss": 0.015892816241830587, + "grad_norm": 0.2912958562374115, + "learning_rate": 2.9615116909509383e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00895428191870451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041157621657475826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02202432621270418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9657064497470855, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7120, + "weighted_flow_matching_loss": 0.015892816241830587, + "weighted_rkd_loss": 0.021269052289426327 + }, + { + "epoch": 2.074483561245272, + "flow_matching_loss": 0.015178527031093837, + "grad_norm": 0.24923548102378845, + "learning_rate": 2.9613253904923924e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008957009855657816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004149204515852034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022063224203884603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9656111001968384, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7130, + "weighted_flow_matching_loss": 0.015178527031093837, + "weighted_rkd_loss": 0.021304501593112944 + }, + { + "epoch": 2.0773930753564156, + "flow_matching_loss": 0.016057768184691668, + "grad_norm": 0.20109790563583374, + "learning_rate": 2.9611386461217772e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008685815101489424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038469032850116493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02121853344142437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9655156195163727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7140, + "weighted_flow_matching_loss": 0.016057768184691668, + "weighted_rkd_loss": 0.02048683874309063 + }, + { + "epoch": 2.080302589467559, + "flow_matching_loss": 0.015466212946921586, + "grad_norm": 0.3214528262615204, + "learning_rate": 2.96095145789582e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008886841963976622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042114440817385915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021985128149390222, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9654200077056885, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7150, + "weighted_flow_matching_loss": 0.015466212946921586, + "weighted_rkd_loss": 0.021224877797067165 + }, + { + "epoch": 2.0832121035787026, + "flow_matching_loss": 0.01581126395612955, + "grad_norm": 0.2932658791542053, + "learning_rate": 2.9607638258713845e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008988360175862908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043476286577060815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022324348986148834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9653242766857147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7160, + "weighted_flow_matching_loss": 0.01581126395612955, + "weighted_rkd_loss": 0.021550229378044607 + }, + { + "epoch": 2.086121617689846, + "flow_matching_loss": 0.01494361162185669, + "grad_norm": 0.23357723653316498, + "learning_rate": 2.960575750105468e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009243793692439795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004473871178925037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022961458750069143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.965228408575058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7170, + "weighted_flow_matching_loss": 0.01494361162185669, + "weighted_rkd_loss": 0.022163047455251217 + }, + { + "epoch": 2.089031131800989, + "flow_matching_loss": 0.014823278971016407, + "grad_norm": 0.25799456238746643, + "learning_rate": 2.9603872306552025e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008578636776655912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038761945441365244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021033468097448348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9651324272155761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7180, + "weighted_flow_matching_loss": 0.014823278971016407, + "weighted_rkd_loss": 0.020300055481493474 + }, + { + "epoch": 2.091940645912133, + "flow_matching_loss": 0.015080484561622143, + "grad_norm": 0.2542925477027893, + "learning_rate": 2.9601982675778562e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008698379434645177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004013732401654124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021410491317510605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9650363087654114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7190, + "weighted_flow_matching_loss": 0.015080484561622143, + "weighted_rkd_loss": 0.020661891996860506 + }, + { + "epoch": 2.094850160023276, + "flow_matching_loss": 0.015200870763510465, + "grad_norm": 0.23328052461147308, + "learning_rate": 2.9600088609308294e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009677064372226596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044891306199133394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023843259178102018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9649400651454926, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7200, + "weighted_flow_matching_loss": 0.015200870763510465, + "weighted_rkd_loss": 0.023007346503436565 + }, + { + "epoch": 2.0977596741344193, + "flow_matching_loss": 0.01350554870441556, + "grad_norm": 0.2044268399477005, + "learning_rate": 2.959819010771661e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009043156635016203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004214205872267484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022300519235432147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9648437023162841, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7210, + "weighted_flow_matching_loss": 0.01350554870441556, + "weighted_rkd_loss": 0.02151650656014681 + }, + { + "epoch": 2.100669188245563, + "flow_matching_loss": 0.016664336994290353, + "grad_norm": 0.2508515417575836, + "learning_rate": 2.959628717158021e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008993104845285416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003966385731473565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02195259537547827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9647471904754639, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7220, + "weighted_flow_matching_loss": 0.016664336994290353, + "weighted_rkd_loss": 0.02117871418595314 + }, + { + "epoch": 2.1035787023567063, + "flow_matching_loss": 0.013925072643905878, + "grad_norm": 0.3047352731227875, + "learning_rate": 2.959437980147716e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00952418246306479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004165743174962699, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0232141082175076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.964650571346283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7230, + "weighted_flow_matching_loss": 0.013925072643905878, + "weighted_rkd_loss": 0.02239349177107215 + }, + { + "epoch": 2.10648821646785, + "flow_matching_loss": 0.01589673487469554, + "grad_norm": 0.23182909190654755, + "learning_rate": 2.959246799798687e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008711288077756762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037480764789506794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021170652844011782, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9645538210868836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7240, + "weighted_flow_matching_loss": 0.01589673487469554, + "weighted_rkd_loss": 0.020420261099934577 + }, + { + "epoch": 2.1093977305789933, + "flow_matching_loss": 0.014325914904475212, + "grad_norm": 0.24344590306282043, + "learning_rate": 2.9590551761690095e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009302241168916226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004399229399859905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023003711737692355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9644569337368012, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7250, + "weighted_flow_matching_loss": 0.014325914904475212, + "weighted_rkd_loss": 0.02218606844544411 + }, + { + "epoch": 2.112307244690137, + "flow_matching_loss": 0.015070849563926458, + "grad_norm": 0.2649766802787781, + "learning_rate": 2.9588631093168936e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008421782962977886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040347080677747725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02087827455252409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9643599331378937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7260, + "weighted_flow_matching_loss": 0.015070849563926458, + "weighted_rkd_loss": 0.02013417612761259 + }, + { + "epoch": 2.1152167588012802, + "flow_matching_loss": 0.01542871855199337, + "grad_norm": 0.24779196083545685, + "learning_rate": 2.9586705993006837e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00819797501899302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037964458111673593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02019239580258727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9642628014087677, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7270, + "weighted_flow_matching_loss": 0.01542871855199337, + "weighted_rkd_loss": 0.019470786862075328 + }, + { + "epoch": 2.1181262729124235, + "flow_matching_loss": 0.016290140617638827, + "grad_norm": 0.28432977199554443, + "learning_rate": 2.95847764617886e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009658062225207686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004142547165974975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02345867156982422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9641655266284943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7280, + "weighted_flow_matching_loss": 0.016290140617638827, + "weighted_rkd_loss": 0.02261801976710558 + }, + { + "epoch": 2.121035787023567, + "flow_matching_loss": 0.015303454361855984, + "grad_norm": 0.3150205910205841, + "learning_rate": 2.9582842500100364e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009032397670671344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004318314255215228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022383109480142594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9640681445598602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7290, + "weighted_flow_matching_loss": 0.015303454361855984, + "weighted_rkd_loss": 0.02157881557941437 + }, + { + "epoch": 2.1239453011347105, + "flow_matching_loss": 0.015381499286741018, + "grad_norm": 0.22934885323047638, + "learning_rate": 2.958090410852961e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009150297567248344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004643703135661781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022944298572838306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9639706254005432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7300, + "weighted_flow_matching_loss": 0.015381499286741018, + "weighted_rkd_loss": 0.022117617167532445 + }, + { + "epoch": 2.1268548152458537, + "flow_matching_loss": 0.014639202505350113, + "grad_norm": 0.23543396592140198, + "learning_rate": 2.9578961287665175e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009278488671407103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00417026556096971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022727242670953274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9638729691505432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7310, + "weighted_flow_matching_loss": 0.014639202505350113, + "weighted_rkd_loss": 0.021906186267733573 + }, + { + "epoch": 2.1297643293569974, + "flow_matching_loss": 0.016264568362385035, + "grad_norm": 0.25497567653656006, + "learning_rate": 2.9577014038097238e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008355912752449512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039163172710686926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02062814272940159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9637751996517181, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7320, + "weighted_flow_matching_loss": 0.016264568362385035, + "weighted_rkd_loss": 0.01988090332597494 + }, + { + "epoch": 2.1326738434681407, + "flow_matching_loss": 0.01566125825047493, + "grad_norm": 0.2546565532684326, + "learning_rate": 2.9575062360417324e-05, + "loss": 0.0396, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009995704237371682, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004837479814887047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024828888289630414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9636773109436035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7330, + "weighted_flow_matching_loss": 0.01566125825047493, + "weighted_rkd_loss": 0.02392710018903017 + }, + { + "epoch": 2.1355833575792844, + "flow_matching_loss": 0.014844376780092716, + "grad_norm": 0.21586282551288605, + "learning_rate": 2.957310625521829e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009605405433103443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004415756650269031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02362656779587269, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.963579285144806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7340, + "weighted_flow_matching_loss": 0.014844376780092716, + "weighted_rkd_loss": 0.022766069136559965 + }, + { + "epoch": 2.1384928716904277, + "flow_matching_loss": 0.016035904455929995, + "grad_norm": 0.279528945684433, + "learning_rate": 2.957114572309436e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007997859409078956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037913267035037277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019787045568227767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9634811222553253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7350, + "weighted_flow_matching_loss": 0.016035904455929995, + "weighted_rkd_loss": 0.019064393918961288 + }, + { + "epoch": 2.141402385801571, + "flow_matching_loss": 0.01422981284558773, + "grad_norm": 0.24870102107524872, + "learning_rate": 2.956918076464109e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008048056205734611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037175124511122704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019813624583184718, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9633828639984131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7360, + "weighted_flow_matching_loss": 0.01422981284558773, + "weighted_rkd_loss": 0.019088076986372472 + }, + { + "epoch": 2.1443118999127146, + "flow_matching_loss": 0.016165448259562253, + "grad_norm": 0.28022822737693787, + "learning_rate": 2.9567211380455376e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009408927569165827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004389541037380695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023207396268844604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9632844388484955, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7370, + "weighted_flow_matching_loss": 0.016165448259562253, + "weighted_rkd_loss": 0.02235530000180006 + }, + { + "epoch": 2.147221414023858, + "flow_matching_loss": 0.014637920539826155, + "grad_norm": 0.2753180265426636, + "learning_rate": 2.9565237571135472e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008682715194299816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037344815442338585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02109991144388914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9631859004497528, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7380, + "weighted_flow_matching_loss": 0.014637920539826155, + "weighted_rkd_loss": 0.020323129184544087 + }, + { + "epoch": 2.1501309281350016, + "flow_matching_loss": 0.014568329229950905, + "grad_norm": 0.28930842876434326, + "learning_rate": 2.9563259337280967e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009350078459829093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004399652522988617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023099809326231478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9630872428417205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7390, + "weighted_flow_matching_loss": 0.014568329229950905, + "weighted_rkd_loss": 0.02224707081913948 + }, + { + "epoch": 2.153040442246145, + "flow_matching_loss": 0.015105050057172775, + "grad_norm": 0.2971581518650055, + "learning_rate": 2.9561276679492794e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008906790846958756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003998808516189456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02181239053606987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9629884541034699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7400, + "weighted_flow_matching_loss": 0.015105050057172775, + "weighted_rkd_loss": 0.021005059778690337 + }, + { + "epoch": 2.155949956357288, + "flow_matching_loss": 0.015153993666172028, + "grad_norm": 0.2524949908256531, + "learning_rate": 2.9559289598373236e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00890714987181127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003961570281535387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021775869838893413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9628895461559296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7410, + "weighted_flow_matching_loss": 0.015153993666172028, + "weighted_rkd_loss": 0.020967711042612792 + }, + { + "epoch": 2.158859470468432, + "flow_matching_loss": 0.018020981922745705, + "grad_norm": 0.26184001564979553, + "learning_rate": 2.9557298094525913e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009443998988717794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004347758530639112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02323575634509325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9627904951572418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7420, + "weighted_flow_matching_loss": 0.018020981922745705, + "weighted_rkd_loss": 0.02237118035554886 + }, + { + "epoch": 2.161768984579575, + "flow_matching_loss": 0.015437027625739574, + "grad_norm": 0.25717779994010925, + "learning_rate": 2.955530216855579e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008614228293299674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003820966393686831, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021049422770738603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.962691330909729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7430, + "weighted_flow_matching_loss": 0.015437027625739574, + "weighted_rkd_loss": 0.020264131389558314 + }, + { + "epoch": 2.164678498690719, + "flow_matching_loss": 0.016624892875552178, + "grad_norm": 0.2824056148529053, + "learning_rate": 2.955330182106918e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009720237832516432, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004499246482737362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023939722403883933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9625920295715332, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7440, + "weighted_flow_matching_loss": 0.016624892875552178, + "weighted_rkd_loss": 0.023044191300868988 + }, + { + "epoch": 2.167588012801862, + "flow_matching_loss": 0.014873823057860136, + "grad_norm": 0.2369089424610138, + "learning_rate": 2.9551297052673734e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009021180402487516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004197696503251791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022240057401359082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9624926090240479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7450, + "weighted_flow_matching_loss": 0.014873823057860136, + "weighted_rkd_loss": 0.021405871957540512 + }, + { + "epoch": 2.1704975269130053, + "flow_matching_loss": 0.01498074121773243, + "grad_norm": 0.22725093364715576, + "learning_rate": 2.954928786397845e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008869923511520028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040426639607176185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021782510727643967, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.962393057346344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7460, + "weighted_flow_matching_loss": 0.01498074121773243, + "weighted_rkd_loss": 0.020963391289114952 + }, + { + "epoch": 2.173407041024149, + "flow_matching_loss": 0.01623020078986883, + "grad_norm": 0.2259262502193451, + "learning_rate": 2.9547274255593665e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009053291147574782, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044793854234740135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022585967741906642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9622933864593506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7470, + "weighted_flow_matching_loss": 0.01623020078986883, + "weighted_rkd_loss": 0.021734347194433214 + }, + { + "epoch": 2.1763165551352923, + "flow_matching_loss": 0.014841917902231216, + "grad_norm": 0.24063001573085785, + "learning_rate": 2.9545256228131058e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009630156168714165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042286323150619864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02348894448950887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9621935725212097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7480, + "weighted_flow_matching_loss": 0.014841917902231216, + "weighted_rkd_loss": 0.02260101493448019 + }, + { + "epoch": 2.179226069246436, + "flow_matching_loss": 0.014774658996611834, + "grad_norm": 0.2318609654903412, + "learning_rate": 2.9543233782203653e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008925336925312877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004202661709859968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022053335234522818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9620936512947083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7490, + "weighted_flow_matching_loss": 0.014774658996611834, + "weighted_rkd_loss": 0.02121741771697998 + }, + { + "epoch": 2.1821355833575793, + "flow_matching_loss": 0.01569085903465748, + "grad_norm": 0.2897837162017822, + "learning_rate": 2.954120691842582e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00907512023113668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004048253316432238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022198493778705596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9619935750961304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7500, + "weighted_flow_matching_loss": 0.01569085903465748, + "weighted_rkd_loss": 0.021354756504297256 + }, + { + "epoch": 2.1850450974687226, + "flow_matching_loss": 0.017297944892197846, + "grad_norm": 0.23373021185398102, + "learning_rate": 2.9539175637413264e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007874056650325655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00352859606500715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019276709109544755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9618934154510498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7510, + "weighted_flow_matching_loss": 0.017297944892197846, + "weighted_rkd_loss": 0.018542111571878194 + }, + { + "epoch": 2.1879546115798663, + "flow_matching_loss": 0.01581384753808379, + "grad_norm": 0.20298795402050018, + "learning_rate": 2.9537139939783033e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008623139606788755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003811828838661313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02105810809880495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9617931008338928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7520, + "weighted_flow_matching_loss": 0.01581384753808379, + "weighted_rkd_loss": 0.02025359347462654 + }, + { + "epoch": 2.1908641256910095, + "flow_matching_loss": 0.014749114867299795, + "grad_norm": 0.21448980271816254, + "learning_rate": 2.9535099826153516e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007911852188408374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036606153938919307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01948432009667158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9616926670074463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7530, + "weighted_flow_matching_loss": 0.014749114867299795, + "weighted_rkd_loss": 0.01873793890699744 + }, + { + "epoch": 2.1937736398021532, + "flow_matching_loss": 0.015256935823708772, + "grad_norm": 0.22816726565361023, + "learning_rate": 2.9533055297144454e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007789082638919353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034938240889459847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019071989320218564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9615920960903168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7540, + "weighted_flow_matching_loss": 0.015256935823708772, + "weighted_rkd_loss": 0.018339498434215784 + }, + { + "epoch": 2.1966831539132965, + "flow_matching_loss": 0.015111944917589427, + "grad_norm": 0.20595292747020721, + "learning_rate": 2.953100635337691e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007792859291657806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035678629064932466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019153581466525793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9614914238452912, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7550, + "weighted_flow_matching_loss": 0.015111944917589427, + "weighted_rkd_loss": 0.018416019156575203 + }, + { + "epoch": 2.1995926680244398, + "flow_matching_loss": 0.014692298974841833, + "grad_norm": 0.2819809317588806, + "learning_rate": 2.952895299547331e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008229779405519366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038523828377947213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020311941765248774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.961390596628189, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7560, + "weighted_flow_matching_loss": 0.014692298974841833, + "weighted_rkd_loss": 0.01952767074108124 + }, + { + "epoch": 2.2025021821355835, + "flow_matching_loss": 0.015049576573073865, + "grad_norm": 0.27828994393348694, + "learning_rate": 2.9526895224057402e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008745177555829287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040947113186120985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021585066989064216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9612896561622619, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7570, + "weighted_flow_matching_loss": 0.015049576573073865, + "weighted_rkd_loss": 0.020749499462544917 + }, + { + "epoch": 2.2054116962467267, + "flow_matching_loss": 0.01620026780292392, + "grad_norm": 0.20922182500362396, + "learning_rate": 2.9524833039754284e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008626367244869471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004129714705049992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02138244900852442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9611885845661163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7580, + "weighted_flow_matching_loss": 0.01620026780292392, + "weighted_rkd_loss": 0.020552631467580795 + }, + { + "epoch": 2.2083212103578704, + "flow_matching_loss": 0.01610681163147092, + "grad_norm": 0.23104682564735413, + "learning_rate": 2.9522766443190398e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00876571349799633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004081659251824021, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021613086201250555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9610873878002166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7590, + "weighted_flow_matching_loss": 0.01610681163147092, + "weighted_rkd_loss": 0.020772080309689044 + }, + { + "epoch": 2.2112307244690137, + "flow_matching_loss": 0.016334795299917458, + "grad_norm": 0.20867086946964264, + "learning_rate": 2.9520695434993516e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00880726077593863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004139565769582987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02175408722832799, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.960986065864563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7600, + "weighted_flow_matching_loss": 0.016334795299917458, + "weighted_rkd_loss": 0.02090544020757079 + }, + { + "epoch": 2.214140238580157, + "flow_matching_loss": 0.016881780046969653, + "grad_norm": 0.26276805996894836, + "learning_rate": 2.9518620015792757e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008220486063510179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003793908073566854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02023488041013479, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9608846187591553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7610, + "weighted_flow_matching_loss": 0.016881780046969653, + "weighted_rkd_loss": 0.019443398527801036 + }, + { + "epoch": 2.2170497526913007, + "flow_matching_loss": 0.015404819510877132, + "grad_norm": 0.2895834743976593, + "learning_rate": 2.951654018621858e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008506193291395903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040412662550807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02105365265160799, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9607830345630646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7620, + "weighted_flow_matching_loss": 0.015404819510877132, + "weighted_rkd_loss": 0.020228003058582543 + }, + { + "epoch": 2.219959266802444, + "flow_matching_loss": 0.014190452173352242, + "grad_norm": 0.30538779497146606, + "learning_rate": 2.951445594690278e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009144906233996154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004129581595771014, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0224193942733109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9606813490390778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7630, + "weighted_flow_matching_loss": 0.014190452173352242, + "weighted_rkd_loss": 0.021537877433001996 + }, + { + "epoch": 2.2228687809135876, + "flow_matching_loss": 0.015092996414750814, + "grad_norm": 0.23068712651729584, + "learning_rate": 2.95123672984785e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008780109276995064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004100469779223204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021660688426345588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9605795204639435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7640, + "weighted_flow_matching_loss": 0.015092996414750814, + "weighted_rkd_loss": 0.020806836243718863 + }, + { + "epoch": 2.225778295024731, + "flow_matching_loss": 0.015771908033639193, + "grad_norm": 0.22806867957115173, + "learning_rate": 2.9510274241580207e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008777597872540354, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038799728034064175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021435168478637932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9604775607585907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7650, + "weighted_flow_matching_loss": 0.015771908033639193, + "weighted_rkd_loss": 0.02058801017701626 + }, + { + "epoch": 2.228687809135874, + "flow_matching_loss": 0.015944305062294006, + "grad_norm": 0.24796351790428162, + "learning_rate": 2.9508176776843726e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009146303636953235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004301793477497995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022594400867819785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9603754878044128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7660, + "weighted_flow_matching_loss": 0.015944305062294006, + "weighted_rkd_loss": 0.021699168346822263 + }, + { + "epoch": 2.231597323247018, + "flow_matching_loss": 0.014560336992144585, + "grad_norm": 0.23137392103672028, + "learning_rate": 2.95060749049062e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008381892694160342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038561819586902858, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020619967579841615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.960273277759552, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7670, + "weighted_flow_matching_loss": 0.014560336992144585, + "weighted_rkd_loss": 0.01980077922344208 + }, + { + "epoch": 2.234506837358161, + "flow_matching_loss": 0.016136565804481508, + "grad_norm": 0.2753266990184784, + "learning_rate": 2.9503968626406133e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00800532870925963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034954698290675878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019506126828491687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9601709485054016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7680, + "weighted_flow_matching_loss": 0.016136565804481508, + "weighted_rkd_loss": 0.018729200772941112 + }, + { + "epoch": 2.237416351469305, + "flow_matching_loss": 0.016132725402712822, + "grad_norm": 0.25776761770248413, + "learning_rate": 2.9501857941983354e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008584325714036823, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004203406628221273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021372058242559434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9600684642791748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7690, + "weighted_flow_matching_loss": 0.016132725402712822, + "weighted_rkd_loss": 0.020518633350729943 + }, + { + "epoch": 2.240325865580448, + "flow_matching_loss": 0.01445256881415844, + "grad_norm": 0.2581033706665039, + "learning_rate": 2.9499742852279025e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009173361025750637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003973843855783343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02232056576758623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9599658966064453, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7700, + "weighted_flow_matching_loss": 0.01445256881415844, + "weighted_rkd_loss": 0.02142697237432003 + }, + { + "epoch": 2.2432353796915914, + "flow_matching_loss": 0.017285640724003314, + "grad_norm": 0.2583281695842743, + "learning_rate": 2.9497623357935666e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008309533027932048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003746149828657508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020365215837955475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9598631918430328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7710, + "weighted_flow_matching_loss": 0.017285640724003314, + "weighted_rkd_loss": 0.019547808356583117 + }, + { + "epoch": 2.246144893802735, + "flow_matching_loss": 0.016055378876626492, + "grad_norm": 0.22970178723335266, + "learning_rate": 2.9495499459597116e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009181863581761717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004290905967354775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02265463285148144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9597603499889373, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7720, + "weighted_flow_matching_loss": 0.016055378876626492, + "weighted_rkd_loss": 0.021742994338274 + }, + { + "epoch": 2.2490544079138783, + "flow_matching_loss": 0.01598466085270047, + "grad_norm": 0.2813285291194916, + "learning_rate": 2.9493371157908563e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009521472081542016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004701465182006359, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023744409345090388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9596573829650878, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7730, + "weighted_flow_matching_loss": 0.01598466085270047, + "weighted_rkd_loss": 0.02278656456619501 + }, + { + "epoch": 2.2519639220250216, + "flow_matching_loss": 0.018179372884333134, + "grad_norm": 0.2446318119764328, + "learning_rate": 2.9491238453516524e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008290393697097897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038732039276510475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02045399136841297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9595542848110199, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7740, + "weighted_flow_matching_loss": 0.018179372884333134, + "weighted_rkd_loss": 0.019626734778285027 + }, + { + "epoch": 2.2548734361361653, + "flow_matching_loss": 0.016092872712761164, + "grad_norm": 0.25856855511665344, + "learning_rate": 2.9489101347068868e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008965011965483428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004107832233421505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02203785628080368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9594510912895202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7750, + "weighted_flow_matching_loss": 0.016092872712761164, + "weighted_rkd_loss": 0.021144217625260354 + }, + { + "epoch": 2.2577829502473086, + "flow_matching_loss": 0.014664200972765684, + "grad_norm": 0.2059895545244217, + "learning_rate": 2.948695983921478e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008848439948633314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041773219127207994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021874201856553553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9593477547168732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7760, + "weighted_flow_matching_loss": 0.014664200972765684, + "weighted_rkd_loss": 0.020984932594001294 + }, + { + "epoch": 2.2606924643584523, + "flow_matching_loss": 0.014923835918307305, + "grad_norm": 0.2708016335964203, + "learning_rate": 2.9484813930604804e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008757278881967067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004157889983616769, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021672447584569453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9592442810535431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7770, + "weighted_flow_matching_loss": 0.014923835918307305, + "weighted_rkd_loss": 0.02078916281461716 + }, + { + "epoch": 2.2636019784695955, + "flow_matching_loss": 0.013984374515712262, + "grad_norm": 0.24109652638435364, + "learning_rate": 2.9482663621890804e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0086060656234622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004004097171127796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021216228045523165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9591406941413879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7780, + "weighted_flow_matching_loss": 0.013984374515712262, + "weighted_rkd_loss": 0.02034933362156153 + }, + { + "epoch": 2.2665114925807393, + "flow_matching_loss": 0.015636178944259883, + "grad_norm": 0.2577589452266693, + "learning_rate": 2.9480508913725986e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009504885971546173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042683171108365055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02327808886766434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9590369760990143, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7790, + "weighted_flow_matching_loss": 0.015636178944259883, + "weighted_rkd_loss": 0.022324532642960547 + }, + { + "epoch": 2.2694210066918825, + "flow_matching_loss": 0.01550313513725996, + "grad_norm": 0.33216592669487, + "learning_rate": 2.9478349806764895e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008640100015327334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003991016373038292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021271216310560705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9589331209659576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7800, + "weighted_flow_matching_loss": 0.01550313513725996, + "weighted_rkd_loss": 0.020397666096687316 + }, + { + "epoch": 2.2723305208030258, + "flow_matching_loss": 0.014763866178691387, + "grad_norm": 0.23714742064476013, + "learning_rate": 2.9476186301663414e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00881059062667191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004216515575535595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021837696991860868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.958829152584076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7810, + "weighted_flow_matching_loss": 0.014763866178691387, + "weighted_rkd_loss": 0.02093864819034934 + }, + { + "epoch": 2.2752400349141695, + "flow_matching_loss": 0.015470107644796371, + "grad_norm": 0.3301858901977539, + "learning_rate": 2.9474018399078752e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008216387825086713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003805091907270253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020237867534160615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9587250649929047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7820, + "weighted_flow_matching_loss": 0.015470107644796371, + "weighted_rkd_loss": 0.01940253544598818 + }, + { + "epoch": 2.2781495490253127, + "flow_matching_loss": 0.01486289780586958, + "grad_norm": 0.31579500436782837, + "learning_rate": 2.947184609966947e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009547119587659835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004641129961237311, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02373536955565214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9586208164691925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7830, + "weighted_flow_matching_loss": 0.01486289780586958, + "weighted_rkd_loss": 0.022753131575882435 + }, + { + "epoch": 2.281059063136456, + "flow_matching_loss": 0.014846089482307433, + "grad_norm": 0.23272296786308289, + "learning_rate": 2.9469669404095447e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008916052570566536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003956190263852477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02178829498589039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9585164844989776, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7840, + "weighted_flow_matching_loss": 0.014846089482307433, + "weighted_rkd_loss": 0.020884452760219573 + }, + { + "epoch": 2.2839685772475997, + "flow_matching_loss": 0.01653488976880908, + "grad_norm": 0.2832123935222626, + "learning_rate": 2.9467488313017908e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008445829572156071, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039050572086125612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02079671612009406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9584120154380799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7850, + "weighted_flow_matching_loss": 0.01653488976880908, + "weighted_rkd_loss": 0.019931753817945718 + }, + { + "epoch": 2.286878091358743, + "flow_matching_loss": 0.015413065068423748, + "grad_norm": 0.25100699067115784, + "learning_rate": 2.9465302827099412e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009060630947351456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041330954758450385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022254357300698756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.958307421207428, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7860, + "weighted_flow_matching_loss": 0.015413065068423748, + "weighted_rkd_loss": 0.021326481364667416 + }, + { + "epoch": 2.2897876054698867, + "flow_matching_loss": 0.015348783787339925, + "grad_norm": 0.1966957449913025, + "learning_rate": 2.9463112947003854e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008425802504643799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004104101611301303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02095570657402277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9582026958465576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7870, + "weighted_flow_matching_loss": 0.015348783787339925, + "weighted_rkd_loss": 0.020079810079187153 + }, + { + "epoch": 2.29269711958103, + "flow_matching_loss": 0.01523688081651926, + "grad_norm": 0.23445406556129456, + "learning_rate": 2.9460918673396455e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009623824153095484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004782714159227907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02403036253526807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9580978393554688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7880, + "weighted_flow_matching_loss": 0.01523688081651926, + "weighted_rkd_loss": 0.02302347021177411 + }, + { + "epoch": 2.295606633692173, + "flow_matching_loss": 0.01663055093958974, + "grad_norm": 0.2337886095046997, + "learning_rate": 2.9458720006943784e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009184317197650671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00435278017539531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02272141482681036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9579928636550903, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7890, + "weighted_flow_matching_loss": 0.01663055093958974, + "weighted_rkd_loss": 0.021766975708305835 + }, + { + "epoch": 2.298516147803317, + "flow_matching_loss": 0.0161094781011343, + "grad_norm": 0.223036527633667, + "learning_rate": 2.9456516948313736e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009467316837981344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004386860900558532, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023321494273841382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9578877627849579, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7900, + "weighted_flow_matching_loss": 0.0161094781011343, + "weighted_rkd_loss": 0.02233944311738014 + }, + { + "epoch": 2.30142566191446, + "flow_matching_loss": 0.015991129633039235, + "grad_norm": 0.2684442698955536, + "learning_rate": 2.9454309498175538e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008100483473390341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037754349643364547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01997640198096633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9577825367450714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7910, + "weighted_flow_matching_loss": 0.015991129633039235, + "weighted_rkd_loss": 0.019133012741804123 + }, + { + "epoch": 2.304335176025604, + "flow_matching_loss": 0.015175069496035577, + "grad_norm": 0.2585475444793701, + "learning_rate": 2.9452097657199762e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009018248366191983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004335521114990115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022372017800807952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9576771914958954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7920, + "weighted_flow_matching_loss": 0.015175069496035577, + "weighted_rkd_loss": 0.021425138227641582 + }, + { + "epoch": 2.307244690136747, + "flow_matching_loss": 0.014674345403909684, + "grad_norm": 0.23371198773384094, + "learning_rate": 2.9449881426058303e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008235172368586063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003864456550218165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02033480126410723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9575717091560364, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7930, + "weighted_flow_matching_loss": 0.014674345403909684, + "weighted_rkd_loss": 0.019472075905650855 + }, + { + "epoch": 2.3101542042478904, + "flow_matching_loss": 0.01493929261341691, + "grad_norm": 0.263891339302063, + "learning_rate": 2.9447660805424388e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009203627519309521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040766448248177765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02248390018939972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9574661016464233, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7940, + "weighted_flow_matching_loss": 0.01493929261341691, + "weighted_rkd_loss": 0.02152759376913309 + }, + { + "epoch": 2.313063718359034, + "flow_matching_loss": 0.014719218108803033, + "grad_norm": 0.29799774289131165, + "learning_rate": 2.94454357959726e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008844441827386617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003997154301032424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021686038002371788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9573603808879853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7950, + "weighted_flow_matching_loss": 0.014719218108803033, + "weighted_rkd_loss": 0.0207613336853683 + }, + { + "epoch": 2.3159732324701774, + "flow_matching_loss": 0.016144862584769725, + "grad_norm": 0.22715966403484344, + "learning_rate": 2.9443206398378815e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008713041944429279, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004092780081555248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021518863923847675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9572545170783997, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7960, + "weighted_flow_matching_loss": 0.016144862584769725, + "weighted_rkd_loss": 0.0205990307033062 + }, + { + "epoch": 2.318882746581321, + "flow_matching_loss": 0.014146113023161887, + "grad_norm": 0.27304142713546753, + "learning_rate": 2.9440972613320282e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009340982930734754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004657530132681132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02333949599415064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9571485400199891, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7970, + "weighted_flow_matching_loss": 0.014146113023161887, + "weighted_rkd_loss": 0.0223393265157938 + }, + { + "epoch": 2.3217922606924644, + "flow_matching_loss": 0.015573231130838394, + "grad_norm": 0.22038701176643372, + "learning_rate": 2.943873444147556e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008604924753308296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004283531196415424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021493380703032017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9570424377918243, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7980, + "weighted_flow_matching_loss": 0.015573231130838394, + "weighted_rkd_loss": 0.02057007374241948 + }, + { + "epoch": 2.3247017748036076, + "flow_matching_loss": 0.014024742133915424, + "grad_norm": 0.2377471625804901, + "learning_rate": 2.9436491883524547e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009430693276226521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004603341780602932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023464728333055972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9569362103939056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7990, + "weighted_flow_matching_loss": 0.014024742133915424, + "weighted_rkd_loss": 0.022454204596579076 + }, + { + "epoch": 2.3276112889147513, + "flow_matching_loss": 0.0161852503195405, + "grad_norm": 0.22634398937225342, + "learning_rate": 2.9434244940148472e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008407256752252578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003848261176608503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020662774704396723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.956829845905304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8000, + "weighted_flow_matching_loss": 0.0161852503195405, + "weighted_rkd_loss": 0.01977078914642334 + }, + { + "epoch": 2.3305208030258946, + "flow_matching_loss": 0.017124445550143717, + "grad_norm": 0.2017577886581421, + "learning_rate": 2.9431993612029903e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008750488609075546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003759991843253374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021260969154536725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9567233741283416, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8010, + "weighted_flow_matching_loss": 0.017124445550143717, + "weighted_rkd_loss": 0.020340894907712938 + }, + { + "epoch": 2.3334303171370383, + "flow_matching_loss": 0.013086509704589844, + "grad_norm": 0.22734710574150085, + "learning_rate": 2.9429737899852728e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008289810549467803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003861211589537561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020440832804888488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9566167771816254, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8020, + "weighted_flow_matching_loss": 0.013086509704589844, + "weighted_rkd_loss": 0.019554053992033006 + }, + { + "epoch": 2.3363398312481816, + "flow_matching_loss": 0.015671677980571987, + "grad_norm": 0.2482413351535797, + "learning_rate": 2.9427477804302174e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008473712299019099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004148098081350327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021095522679388523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9565100371837616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8030, + "weighted_flow_matching_loss": 0.015671677980571987, + "weighted_rkd_loss": 0.020178054086863996 + }, + { + "epoch": 2.339249345359325, + "flow_matching_loss": 0.01593680903315544, + "grad_norm": 0.25515541434288025, + "learning_rate": 2.9425213326064797e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008549771877005697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004058054299093783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02115759812295437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9564031779766082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8040, + "weighted_flow_matching_loss": 0.01593680903315544, + "weighted_rkd_loss": 0.020235219225287437 + }, + { + "epoch": 2.3421588594704685, + "flow_matching_loss": 0.014887205325067043, + "grad_norm": 0.20991799235343933, + "learning_rate": 2.942294446582849e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008372657420113682, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003744651121087372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02048996603116393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9562961995601654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8050, + "weighted_flow_matching_loss": 0.014887205325067043, + "weighted_rkd_loss": 0.019594506546854972 + }, + { + "epoch": 2.345068373581612, + "flow_matching_loss": 0.015603471640497446, + "grad_norm": 0.2940078377723694, + "learning_rate": 2.9420671224282468e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008647974301129579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003924869676120579, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02122081834822893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.956189090013504, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8060, + "weighted_flow_matching_loss": 0.015603471640497446, + "weighted_rkd_loss": 0.020291089825332165 + }, + { + "epoch": 2.347977887692755, + "flow_matching_loss": 0.016752268001437188, + "grad_norm": 0.2695174217224121, + "learning_rate": 2.9418393602117284e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008592606848105788, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003841935959644616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021027149446308614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9560818672180176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8070, + "weighted_flow_matching_loss": 0.016752268001437188, + "weighted_rkd_loss": 0.020103709399700166 + }, + { + "epoch": 2.3508874018038988, + "flow_matching_loss": 0.016026651486754417, + "grad_norm": 0.27691134810447693, + "learning_rate": 2.9416111600024816e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008369216462597251, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037272868445143103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020465719513595104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9559745013713836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8080, + "weighted_flow_matching_loss": 0.016026651486754417, + "weighted_rkd_loss": 0.01956465933471918 + }, + { + "epoch": 2.353796915915042, + "flow_matching_loss": 0.015462386794388295, + "grad_norm": 0.25210943818092346, + "learning_rate": 2.9413825218698284e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008223977778106928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038235927233472465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020271548256278037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9558670163154602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8090, + "weighted_flow_matching_loss": 0.015462386794388295, + "weighted_rkd_loss": 0.019376876950263976 + }, + { + "epoch": 2.3567064300261857, + "flow_matching_loss": 0.015843940898776054, + "grad_norm": 0.2651820182800293, + "learning_rate": 2.9411534458832222e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0082309041172266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037401375360786916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02020194586366415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9557594180107116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8100, + "weighted_flow_matching_loss": 0.015843940898776054, + "weighted_rkd_loss": 0.01930818734690547 + }, + { + "epoch": 2.359615944137329, + "flow_matching_loss": 0.018115952145308256, + "grad_norm": 0.2848023474216461, + "learning_rate": 2.9409239321122513e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007806095108389854, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003762248414568603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01937443846836686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9556516885757447, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8110, + "weighted_flow_matching_loss": 0.018115952145308256, + "weighted_rkd_loss": 0.01851524170488119 + }, + { + "epoch": 2.3625254582484727, + "flow_matching_loss": 0.015335238259285688, + "grad_norm": 0.2572993338108063, + "learning_rate": 2.940693980626634e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00838986849412322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038187331054359676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020598470140248536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9555438339710236, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8120, + "weighted_flow_matching_loss": 0.015335238259285688, + "weighted_rkd_loss": 0.01968272039666772 + }, + { + "epoch": 2.365434972359616, + "flow_matching_loss": 0.015319690480828286, + "grad_norm": 0.2025359570980072, + "learning_rate": 2.9404635914962258e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00877520777285099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003808083035983145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021358498744666575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.955435848236084, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8130, + "weighted_flow_matching_loss": 0.015319690480828286, + "weighted_rkd_loss": 0.02040670458227396 + }, + { + "epoch": 2.3683444864707592, + "flow_matching_loss": 0.014635163079947233, + "grad_norm": 0.22091025114059448, + "learning_rate": 2.9402327647910113e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008543480979278683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039239812176674604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021010943315923215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9553277432918549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8140, + "weighted_flow_matching_loss": 0.014635163079947233, + "weighted_rkd_loss": 0.02007230706512928 + }, + { + "epoch": 2.371254000581903, + "flow_matching_loss": 0.014779493492096663, + "grad_norm": 0.23014512658119202, + "learning_rate": 2.94000150058111e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008312268601730465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003795518772676587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020420055743306874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9552195072174072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8150, + "weighted_flow_matching_loss": 0.014779493492096663, + "weighted_rkd_loss": 0.019505614694207907 + }, + { + "epoch": 2.374163514693046, + "flow_matching_loss": 0.015326057001948357, + "grad_norm": 0.19435596466064453, + "learning_rate": 2.939769798936774e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008205290324985981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003593977959826589, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02000455856323242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9551111459732056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8160, + "weighted_flow_matching_loss": 0.015326057001948357, + "weighted_rkd_loss": 0.019106604903936387 + }, + { + "epoch": 2.3770730288041895, + "flow_matching_loss": 0.014763627666980028, + "grad_norm": 0.2426217496395111, + "learning_rate": 2.9395376599283878e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008949105953797698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004116524662822485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022014736756682397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9550026655197144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8170, + "weighted_flow_matching_loss": 0.014763627666980028, + "weighted_rkd_loss": 0.021024123392999174 + }, + { + "epoch": 2.379982542915333, + "flow_matching_loss": 0.014742093905806542, + "grad_norm": 0.22862565517425537, + "learning_rate": 2.9393050836264697e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008419578475877643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003881355095654726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020720512233674528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9548940658569336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8180, + "weighted_flow_matching_loss": 0.014742093905806542, + "weighted_rkd_loss": 0.019785900972783564 + }, + { + "epoch": 2.3828920570264764, + "flow_matching_loss": 0.015369070507586003, + "grad_norm": 0.2902276813983917, + "learning_rate": 2.9390720701016693e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008402928849682212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004011268517933786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020817126240581275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9547853410243988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8190, + "weighted_flow_matching_loss": 0.015369070507586003, + "weighted_rkd_loss": 0.019875829014927147 + }, + { + "epoch": 2.38580157113762, + "flow_matching_loss": 0.014459119085222483, + "grad_norm": 0.23203088343143463, + "learning_rate": 2.938838619424771e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007821971038356423, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038803195813670753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019524261727929114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9546764850616455, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8200, + "weighted_flow_matching_loss": 0.014459119085222483, + "weighted_rkd_loss": 0.018639358598738907 + }, + { + "epoch": 2.3887110852487634, + "flow_matching_loss": 0.0158506041392684, + "grad_norm": 0.2187754213809967, + "learning_rate": 2.93860473166669e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008962715929374098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00439774573314935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022323177754879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9545675098896027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8210, + "weighted_flow_matching_loss": 0.0158506041392684, + "weighted_rkd_loss": 0.02130893263965845 + }, + { + "epoch": 2.391620599359907, + "flow_matching_loss": 0.01336903553456068, + "grad_norm": 0.17981165647506714, + "learning_rate": 2.938370406898476e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007232347223907709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003509056381881237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017973750457167627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9544584035873414, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8220, + "weighted_flow_matching_loss": 0.01336903553456068, + "weighted_rkd_loss": 0.017155234050005675 + }, + { + "epoch": 2.3945301134710504, + "flow_matching_loss": 0.014917998760938644, + "grad_norm": 0.21924249827861786, + "learning_rate": 2.9381356451913103e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00908108027651906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004050824721343815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022212985344231127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9543491721153259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8230, + "weighted_flow_matching_loss": 0.014917998760938644, + "weighted_rkd_loss": 0.021199028193950652 + }, + { + "epoch": 2.3974396275821936, + "flow_matching_loss": 0.014851746149361134, + "grad_norm": 0.2536202371120453, + "learning_rate": 2.9379004466165072e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008290267502889036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003593728342093527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020174263417720793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9542398095130921, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8240, + "weighted_flow_matching_loss": 0.014851746149361134, + "weighted_rkd_loss": 0.01925103571265936 + }, + { + "epoch": 2.4003491416933374, + "flow_matching_loss": 0.01454908326268196, + "grad_norm": 0.20396147668361664, + "learning_rate": 2.937664811245514e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008257783995941282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003773430758155882, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020288999006152154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9541303455829621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8250, + "weighted_flow_matching_loss": 0.01454908326268196, + "weighted_rkd_loss": 0.019358360022306443 + }, + { + "epoch": 2.4032586558044806, + "flow_matching_loss": 0.016714068502187727, + "grad_norm": 0.21182747185230255, + "learning_rate": 2.9374287391499108e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008963495772331954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004152456275187432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022079447470605373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9540207386016846, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8260, + "weighted_flow_matching_loss": 0.016714068502187727, + "weighted_rkd_loss": 0.021064282953739168 + }, + { + "epoch": 2.406168169915624, + "flow_matching_loss": 0.01627591336145997, + "grad_norm": 0.23597030341625214, + "learning_rate": 2.9371922304014093e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008201542263850569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037121428176760674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02011522725224495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9539110124111175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8270, + "weighted_flow_matching_loss": 0.01627591336145997, + "weighted_rkd_loss": 0.01918814843520522 + }, + { + "epoch": 2.4090776840267676, + "flow_matching_loss": 0.014926653448492288, + "grad_norm": 0.22982650995254517, + "learning_rate": 2.9369552850718557e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008739299094304442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004096013586968183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02157461177557707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9538011610507965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8280, + "weighted_flow_matching_loss": 0.014926653448492288, + "weighted_rkd_loss": 0.020577952824532987 + }, + { + "epoch": 2.411987198137911, + "flow_matching_loss": 0.015089463535696268, + "grad_norm": 0.2274298369884491, + "learning_rate": 2.9367179032332262e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008820603741332889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004105595219880342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021746802888810635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9536911845207214, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8290, + "weighted_flow_matching_loss": 0.015089463535696268, + "weighted_rkd_loss": 0.02073971536010504 + }, + { + "epoch": 2.4148967122490546, + "flow_matching_loss": 0.01707094321027398, + "grad_norm": 0.2649328112602234, + "learning_rate": 2.936480084957633e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00869838111102581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004109648242592812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02150641046464443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9535810887813568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8300, + "weighted_flow_matching_loss": 0.01707094321027398, + "weighted_rkd_loss": 0.020508110616356136 + }, + { + "epoch": 2.417806226360198, + "flow_matching_loss": 0.016048634983599187, + "grad_norm": 0.22488701343536377, + "learning_rate": 2.9362418303173174e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008431538753211499, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004211463104002178, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021074540819972752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9534708559513092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8310, + "weighted_flow_matching_loss": 0.016048634983599187, + "weighted_rkd_loss": 0.020094044227153062 + }, + { + "epoch": 2.4207157404713415, + "flow_matching_loss": 0.014560166001319885, + "grad_norm": 0.21787837147712708, + "learning_rate": 2.9360031393846558e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075069907121360305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003663845919072628, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018677827343344687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.953360515832901, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8320, + "weighted_flow_matching_loss": 0.014560166001319885, + "weighted_rkd_loss": 0.017806678265333175 + }, + { + "epoch": 2.423625254582485, + "flow_matching_loss": 0.014136118162423373, + "grad_norm": 0.24160392582416534, + "learning_rate": 2.9357640122321557e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007257458427920938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033111119410023093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017826028726994993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9532500386238099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8330, + "weighted_flow_matching_loss": 0.014136118162423373, + "weighted_rkd_loss": 0.016992672998458146 + }, + { + "epoch": 2.426534768693628, + "flow_matching_loss": 0.01585536990314722, + "grad_norm": 0.22460874915122986, + "learning_rate": 2.935524448932458e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007951532723382115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037501540733501314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019653219543397427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9531394422054291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8340, + "weighted_flow_matching_loss": 0.01585536990314722, + "weighted_rkd_loss": 0.018732274789363144 + }, + { + "epoch": 2.4294442828047718, + "flow_matching_loss": 0.014316465985029937, + "grad_norm": 0.20030535757541656, + "learning_rate": 2.9352844495583356e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008002173155546188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003793410141952336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01979775633662939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9530287206172943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8350, + "weighted_flow_matching_loss": 0.014316465985029937, + "weighted_rkd_loss": 0.018867819383740426 + }, + { + "epoch": 2.432353796915915, + "flow_matching_loss": 0.014873544685542583, + "grad_norm": 0.27162662148475647, + "learning_rate": 2.9350440141826938e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009451935393735766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004376459401100874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023280330188572406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9529178738594055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8360, + "weighted_flow_matching_loss": 0.014873544685542583, + "weighted_rkd_loss": 0.022184223867952824 + }, + { + "epoch": 2.4352633110270583, + "flow_matching_loss": 0.01559155061841011, + "grad_norm": 0.24760288000106812, + "learning_rate": 2.9348031428785705e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008325446723029017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036285867914557456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020279480423778296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9528069019317627, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8370, + "weighted_flow_matching_loss": 0.01559155061841011, + "weighted_rkd_loss": 0.019322494976222515 + }, + { + "epoch": 2.438172825138202, + "flow_matching_loss": 0.01652508657425642, + "grad_norm": 0.26135531067848206, + "learning_rate": 2.9345618357191363e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008307206211611629, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038037985330447556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020418211072683334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9526958107948303, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8380, + "weighted_flow_matching_loss": 0.01652508657425642, + "weighted_rkd_loss": 0.01945231845602393 + }, + { + "epoch": 2.4410823392493453, + "flow_matching_loss": 0.014093879703432321, + "grad_norm": 0.2113395631313324, + "learning_rate": 2.934320092777694e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008344253897666931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036746546858921645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020363162457942962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9525845944881439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8390, + "weighted_flow_matching_loss": 0.014093879703432321, + "weighted_rkd_loss": 0.019397646095603703 + }, + { + "epoch": 2.443991853360489, + "flow_matching_loss": 0.013925729226320982, + "grad_norm": 0.24570512771606445, + "learning_rate": 2.9340779141276782e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008136802585795522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037220496218651534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019995655119419097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.952473247051239, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8400, + "weighted_flow_matching_loss": 0.013925729226320982, + "weighted_rkd_loss": 0.019045311212539672 + }, + { + "epoch": 2.4469013674716322, + "flow_matching_loss": 0.015198691561818124, + "grad_norm": 0.23223863542079926, + "learning_rate": 2.933835299842657e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008642646437510847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004029909544624388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021315202675759794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9523617804050446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8410, + "weighted_flow_matching_loss": 0.015198691561818124, + "weighted_rkd_loss": 0.0202998461201787 + }, + { + "epoch": 2.4498108815827755, + "flow_matching_loss": 0.015868385042995214, + "grad_norm": 0.2717077434062958, + "learning_rate": 2.93359224999633e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008924887795001269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003691808064468205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021541583444923164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9522502005100251, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8420, + "weighted_flow_matching_loss": 0.015868385042995214, + "weighted_rkd_loss": 0.020512977987527846 + }, + { + "epoch": 2.452720395693919, + "flow_matching_loss": 0.014716164208948612, + "grad_norm": 0.2041371911764145, + "learning_rate": 2.933348764662529e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008583201793953776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003977529751136899, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02114393366500735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9521384835243225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8430, + "weighted_flow_matching_loss": 0.014716164208948612, + "weighted_rkd_loss": 0.020131965540349482 + }, + { + "epoch": 2.4556299098050625, + "flow_matching_loss": 0.013862016890197993, + "grad_norm": 0.21489237248897552, + "learning_rate": 2.933104843915219e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008807617239654064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003885495895519853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021500730141997337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9520266473293304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8440, + "weighted_flow_matching_loss": 0.013862016890197993, + "weighted_rkd_loss": 0.020469230227172376 + }, + { + "epoch": 2.458539423916206, + "flow_matching_loss": 0.014058343693614006, + "grad_norm": 0.2079968899488449, + "learning_rate": 2.9328604878284963e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009016227442771196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004088855301961303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022121310140937565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9519146680831909, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8450, + "weighted_flow_matching_loss": 0.014058343693614006, + "weighted_rkd_loss": 0.02105760257691145 + }, + { + "epoch": 2.4614489380273494, + "flow_matching_loss": 0.01551224458962679, + "grad_norm": 0.24969804286956787, + "learning_rate": 2.9326156964765892e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00864880126900971, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003966702590696513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021264304965734483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9518025934696197, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8460, + "weighted_flow_matching_loss": 0.01551224458962679, + "weighted_rkd_loss": 0.020239396952092648 + }, + { + "epoch": 2.4643584521384927, + "flow_matching_loss": 0.01530473679304123, + "grad_norm": 0.2479257732629776, + "learning_rate": 2.9323704699338603e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008270885655656457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038375217700377107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020379293151199816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9516903817653656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8470, + "weighted_flow_matching_loss": 0.01530473679304123, + "weighted_rkd_loss": 0.019394753128290178 + }, + { + "epoch": 2.4672679662496364, + "flow_matching_loss": 0.01515371659770608, + "grad_norm": 0.22535434365272522, + "learning_rate": 2.932124808274802e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008633472118526697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004007556661963463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021274501085281373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9515780448913574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8480, + "weighted_flow_matching_loss": 0.01515371659770608, + "weighted_rkd_loss": 0.0202443796209991 + }, + { + "epoch": 2.4701774803607797, + "flow_matching_loss": 0.015117610711604357, + "grad_norm": 0.22022870182991028, + "learning_rate": 2.93187871157404e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008392690820619464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003958668932318688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02074405029416084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9514655888080596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8490, + "weighted_flow_matching_loss": 0.015117610711604357, + "weighted_rkd_loss": 0.01973728956654668 + }, + { + "epoch": 2.4730869944719234, + "flow_matching_loss": 0.015337605588138103, + "grad_norm": 0.24350833892822266, + "learning_rate": 2.931632179906332e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008910495368763805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004285375354811549, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02210636604577303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9513530194759369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8500, + "weighted_flow_matching_loss": 0.015337605588138103, + "weighted_rkd_loss": 0.021030963771045207 + }, + { + "epoch": 2.4759965085830666, + "flow_matching_loss": 0.016003498714417218, + "grad_norm": 0.25292617082595825, + "learning_rate": 2.931385213346568e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009909651149064303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004459076141938567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0242783784866333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9512403070926666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8510, + "weighted_flow_matching_loss": 0.016003498714417218, + "weighted_rkd_loss": 0.02309458814561367 + }, + { + "epoch": 2.47890602269421, + "flow_matching_loss": 0.016974058747291566, + "grad_norm": 0.30240559577941895, + "learning_rate": 2.9311378119697694e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008313699625432491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003873981605283916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020501381158828734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9511274814605712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8520, + "weighted_flow_matching_loss": 0.016974058747291566, + "weighted_rkd_loss": 0.01949941348284483 + }, + { + "epoch": 2.4818155368053536, + "flow_matching_loss": 0.014472543448209762, + "grad_norm": 0.26581913232803345, + "learning_rate": 2.9308899758510903e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008704659761860967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003974770358763635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02138408962637186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9510145246982574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8530, + "weighted_flow_matching_loss": 0.014472543448209762, + "weighted_rkd_loss": 0.020336439833045006 + }, + { + "epoch": 2.484725050916497, + "flow_matching_loss": 0.016093956865370275, + "grad_norm": 0.2912863492965698, + "learning_rate": 2.9306417050658172e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00878156814724207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00423841648735106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021801552921533584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9509014546871185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8540, + "weighted_flow_matching_loss": 0.016093956865370275, + "weighted_rkd_loss": 0.02073113024234772 + }, + { + "epoch": 2.4876345650276406, + "flow_matching_loss": 0.015051427111029626, + "grad_norm": 0.22134047746658325, + "learning_rate": 2.9303929996893687e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009129703557118774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004055958730168641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022315365634858607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9507882595062256, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8550, + "weighted_flow_matching_loss": 0.015051427111029626, + "weighted_rkd_loss": 0.021217212080955505 + }, + { + "epoch": 2.490544079138784, + "flow_matching_loss": 0.01654309993609786, + "grad_norm": 0.29030925035476685, + "learning_rate": 2.9301438597972932e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009468930307775736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004361129319295287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02329898960888386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9506749272346496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8560, + "weighted_flow_matching_loss": 0.01654309993609786, + "weighted_rkd_loss": 0.022149826027452947 + }, + { + "epoch": 2.493453593249927, + "flow_matching_loss": 0.014818301051855087, + "grad_norm": 0.25485336780548096, + "learning_rate": 2.9298942854652744e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008360855467617511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004070677631534636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02079238835722208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9505614817142487, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8570, + "weighted_flow_matching_loss": 0.014818301051855087, + "weighted_rkd_loss": 0.01976440502330661 + }, + { + "epoch": 2.496363107361071, + "flow_matching_loss": 0.016412827745079996, + "grad_norm": 0.2791491746902466, + "learning_rate": 2.9296442767691257e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008329791808500886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003792760614305735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02045234404504299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9504479169845581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8580, + "weighted_flow_matching_loss": 0.016412827745079996, + "weighted_rkd_loss": 0.019438849948346614 + }, + { + "epoch": 2.499272621472214, + "flow_matching_loss": 0.01456994703039527, + "grad_norm": 0.24617017805576324, + "learning_rate": 2.9293938337847936e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00799533948302269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037674460094422103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019758125022053718, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9503342092037201, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8590, + "weighted_flow_matching_loss": 0.01456994703039527, + "weighted_rkd_loss": 0.018776779249310493 + }, + { + "epoch": 2.5021821355833573, + "flow_matching_loss": 0.014451395254582167, + "grad_norm": 0.2381831407546997, + "learning_rate": 2.9291429565883554e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007551236823201179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003593570273369551, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018696044199168683, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9502204060554504, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8600, + "weighted_flow_matching_loss": 0.014451395254582167, + "weighted_rkd_loss": 0.01776532465592027 + }, + { + "epoch": 2.505091649694501, + "flow_matching_loss": 0.014565786998718976, + "grad_norm": 0.24077565968036652, + "learning_rate": 2.9288916452560214e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00855261473916471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039044188102707266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02100964803248644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9501064538955688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8610, + "weighted_flow_matching_loss": 0.014565786998718976, + "weighted_rkd_loss": 0.019961346313357352 + }, + { + "epoch": 2.5080011638056443, + "flow_matching_loss": 0.016380681563168763, + "grad_norm": 0.2684432864189148, + "learning_rate": 2.9286398998641334e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008113324362784623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035749848233535887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019801633432507515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9499923884868622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8620, + "weighted_flow_matching_loss": 0.016380681563168763, + "weighted_rkd_loss": 0.01881143171340227 + }, + { + "epoch": 2.510910677916788, + "flow_matching_loss": 0.014845523331314326, + "grad_norm": 0.2394900619983673, + "learning_rate": 2.9283877204891645e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007725349348038435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034063365776091812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018857035599648953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.949878203868866, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8630, + "weighted_flow_matching_loss": 0.014845523331314326, + "weighted_rkd_loss": 0.01791187794879079 + }, + { + "epoch": 2.5138201920279313, + "flow_matching_loss": 0.013512992672622204, + "grad_norm": 0.2576269805431366, + "learning_rate": 2.9281351072077208e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008360031060874462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003924058703705669, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020644120685756207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9497638881206513, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8640, + "weighted_flow_matching_loss": 0.013512992672622204, + "weighted_rkd_loss": 0.01960703507065773 + }, + { + "epoch": 2.516729706139075, + "flow_matching_loss": 0.014877188298851252, + "grad_norm": 0.20285294950008392, + "learning_rate": 2.9278820600965393e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007569851726293564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034784841118380427, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01861818740144372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.949649453163147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8650, + "weighted_flow_matching_loss": 0.014877188298851252, + "weighted_rkd_loss": 0.01768078552559018 + }, + { + "epoch": 2.5196392202502182, + "flow_matching_loss": 0.015308463759720326, + "grad_norm": 0.23592643439769745, + "learning_rate": 2.9276285792324887e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008251648535951972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038172745145857332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02032057140022516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9495348989963531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8660, + "weighted_flow_matching_loss": 0.015308463759720326, + "weighted_rkd_loss": 0.01929509099572897 + }, + { + "epoch": 2.5225487343613615, + "flow_matching_loss": 0.01620763847604394, + "grad_norm": 0.2324291467666626, + "learning_rate": 2.9273746646925703e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008949505724012851, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004416793235577643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022315804846584797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9494202196598053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8670, + "weighted_flow_matching_loss": 0.01620763847604394, + "weighted_rkd_loss": 0.02118709161877632 + }, + { + "epoch": 2.525458248472505, + "flow_matching_loss": 0.014371990878134965, + "grad_norm": 0.20049265027046204, + "learning_rate": 2.9271203165539166e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008532516472041608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036762085277587176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020741241704672574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9493054270744323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8680, + "weighted_flow_matching_loss": 0.014371990878134965, + "weighted_rkd_loss": 0.019689824990928174 + }, + { + "epoch": 2.5283677625836485, + "flow_matching_loss": 0.013083752524107695, + "grad_norm": 0.24667306244373322, + "learning_rate": 2.9268655348937913e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008151900582015514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003988669905811548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02029247088357806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9491904973983765, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8690, + "weighted_flow_matching_loss": 0.013083752524107695, + "weighted_rkd_loss": 0.01926144855096936 + }, + { + "epoch": 2.5312772766947917, + "flow_matching_loss": 0.016070236917585135, + "grad_norm": 0.20278692245483398, + "learning_rate": 2.926610319789591e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00837402418255806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036282259970903397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02037627436220646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.949075448513031, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8700, + "weighted_flow_matching_loss": 0.016070236917585135, + "weighted_rkd_loss": 0.019338686671108007 + }, + { + "epoch": 2.5341867908059355, + "flow_matching_loss": 0.015306037291884423, + "grad_norm": 0.235345259308815, + "learning_rate": 2.926354671318843e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007325870310887694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003406726522371173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018058467470109463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9489602625370026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8710, + "weighted_flow_matching_loss": 0.015306037291884423, + "weighted_rkd_loss": 0.01713676583021879 + }, + { + "epoch": 2.5370963049170787, + "flow_matching_loss": 0.013326841779053212, + "grad_norm": 0.21364863216876984, + "learning_rate": 2.926098589559206e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007863793056458235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003525695181451738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01925328131765127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9488449692726135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8720, + "weighted_flow_matching_loss": 0.013326841779053212, + "weighted_rkd_loss": 0.01826832313090563 + }, + { + "epoch": 2.5400058190282224, + "flow_matching_loss": 0.016494847275316714, + "grad_norm": 0.25806471705436707, + "learning_rate": 2.925842074588472e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007590268878266216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034532201010733844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01863375771790743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9487295508384704, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8730, + "weighted_flow_matching_loss": 0.016494847275316714, + "weighted_rkd_loss": 0.017678424436599015 + }, + { + "epoch": 2.5429153331393657, + "flow_matching_loss": 0.017247560992836954, + "grad_norm": 0.24746990203857422, + "learning_rate": 2.925585126484563e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008901507826521993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004169361991807818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02197237769141793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9486140012741089, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8740, + "weighted_flow_matching_loss": 0.017247560992836954, + "weighted_rkd_loss": 0.020843405835330488 + }, + { + "epoch": 2.5458248472505094, + "flow_matching_loss": 0.017191739473491906, + "grad_norm": 0.22316721081733704, + "learning_rate": 2.9253277453255326e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008275559078902006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037404957693070173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020291613694280387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9484983444213867, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8750, + "weighted_flow_matching_loss": 0.017191739473491906, + "weighted_rkd_loss": 0.01924657691270113 + }, + { + "epoch": 2.5487343613616527, + "flow_matching_loss": 0.015863383561372756, + "grad_norm": 0.2429952770471573, + "learning_rate": 2.925069931189567e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007958321180194617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003714376385323703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01963101848959923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9483825623989105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8760, + "weighted_flow_matching_loss": 0.015863383561372756, + "weighted_rkd_loss": 0.018617702648043633 + }, + { + "epoch": 2.551643875472796, + "flow_matching_loss": 0.01570917246863246, + "grad_norm": 0.2691846489906311, + "learning_rate": 2.924811684154983e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008313150517642498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038756073452532293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020501908287405966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9482666492462158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8770, + "weighted_flow_matching_loss": 0.01570917246863246, + "weighted_rkd_loss": 0.019441276043653487 + }, + { + "epoch": 2.5545533895839396, + "flow_matching_loss": 0.0142817672342062, + "grad_norm": 0.23345430195331573, + "learning_rate": 2.9245530043002293e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008698474941775202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039204942993819715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02131744399666786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9481506168842315, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8780, + "weighted_flow_matching_loss": 0.0142817672342062, + "weighted_rkd_loss": 0.02021216005086899 + }, + { + "epoch": 2.557462903695083, + "flow_matching_loss": 0.017413760907948016, + "grad_norm": 0.29965972900390625, + "learning_rate": 2.9242938917038863e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008741768030449747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044513199711218475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021934856101870535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9480344355106354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8790, + "weighted_flow_matching_loss": 0.017413760907948016, + "weighted_rkd_loss": 0.020794961042702197 + }, + { + "epoch": 2.560372417806226, + "flow_matching_loss": 0.015178073849529027, + "grad_norm": 0.2922346591949463, + "learning_rate": 2.9240343464446647e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008339807065203786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004101964202709496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020781578309834004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9479181706905365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8800, + "weighted_flow_matching_loss": 0.015178073849529027, + "weighted_rkd_loss": 0.019699224457144737 + }, + { + "epoch": 2.56328193191737, + "flow_matching_loss": 0.0166058287024498, + "grad_norm": 0.28024202585220337, + "learning_rate": 2.923774368601409e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008645665552467107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004349228274077177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02164055947214365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9478017747402191, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8810, + "weighted_flow_matching_loss": 0.0166058287024498, + "weighted_rkd_loss": 0.020510913338512182 + }, + { + "epoch": 2.566191446028513, + "flow_matching_loss": 0.014232801366597414, + "grad_norm": 0.24404673278331757, + "learning_rate": 2.9235139582530918e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008688291162252426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00406577626708895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02144235894083977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9476852476596832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8820, + "weighted_flow_matching_loss": 0.014232801366597414, + "weighted_rkd_loss": 0.020320662297308445 + }, + { + "epoch": 2.569100960139657, + "flow_matching_loss": 0.013972209673374892, + "grad_norm": 0.30322667956352234, + "learning_rate": 2.9232531154788203e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008633008133620024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037358790636062623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021001895144581794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9475686013698578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8830, + "weighted_flow_matching_loss": 0.013972209673374892, + "weighted_rkd_loss": 0.01990065732970834 + }, + { + "epoch": 2.5720104742508, + "flow_matching_loss": 0.014199352636933327, + "grad_norm": 0.30435481667518616, + "learning_rate": 2.9229918403578313e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008802442764863371, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00423808335326612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021842968929558993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9474518299102783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8840, + "weighted_flow_matching_loss": 0.014199352636933327, + "weighted_rkd_loss": 0.02069514226168394 + }, + { + "epoch": 2.574919988361944, + "flow_matching_loss": 0.013820161297917365, + "grad_norm": 0.21808408200740814, + "learning_rate": 2.922730132969493e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007978199608623981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003884169599041343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019840568862855436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9473349392414093, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8850, + "weighted_flow_matching_loss": 0.013820161297917365, + "weighted_rkd_loss": 0.01879563331604004 + }, + { + "epoch": 2.577829502473087, + "flow_matching_loss": 0.014178892970085144, + "grad_norm": 0.22214294970035553, + "learning_rate": 2.922467993393305e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008170184493064881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003979471861384809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02031984105706215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9472179174423218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8860, + "weighted_flow_matching_loss": 0.014178892970085144, + "weighted_rkd_loss": 0.019247328210622073 + }, + { + "epoch": 2.5807390165842303, + "flow_matching_loss": 0.01527830446138978, + "grad_norm": 0.23622682690620422, + "learning_rate": 2.9222054217088985e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008579654479399323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004140446335077286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02129975501447916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9471007704734802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8870, + "weighted_flow_matching_loss": 0.01527830446138978, + "weighted_rkd_loss": 0.02017295528203249 + }, + { + "epoch": 2.583648530695374, + "flow_matching_loss": 0.015127559751272201, + "grad_norm": 0.24443846940994263, + "learning_rate": 2.921942417996037e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00828603347763419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037400612141937017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020312128122895956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9469835162162781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8880, + "weighted_flow_matching_loss": 0.015127559751272201, + "weighted_rkd_loss": 0.01923523610457778 + }, + { + "epoch": 2.5865580448065173, + "flow_matching_loss": 0.015379179175943137, + "grad_norm": 0.22659814357757568, + "learning_rate": 2.921678982334612e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008024717727676035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00396772576496005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020017161220312118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9468661248683929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8890, + "weighted_flow_matching_loss": 0.015379179175943137, + "weighted_rkd_loss": 0.01895358320325613 + }, + { + "epoch": 2.5894675589176606, + "flow_matching_loss": 0.015631949063390493, + "grad_norm": 0.25310245156288147, + "learning_rate": 2.92141511480465e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008320685010403395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038947614841163157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02053613159805536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9467486262321472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8900, + "weighted_flow_matching_loss": 0.015631949063390493, + "weighted_rkd_loss": 0.019442578312009573 + }, + { + "epoch": 2.5923770730288043, + "flow_matching_loss": 0.01457833917811513, + "grad_norm": 0.318674772977829, + "learning_rate": 2.9211508154863066e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008815258229151367, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004103150428272783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021733666770160198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9466309964656829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8910, + "weighted_flow_matching_loss": 0.01457833917811513, + "weighted_rkd_loss": 0.02057376252487302 + }, + { + "epoch": 2.5952865871399475, + "flow_matching_loss": 0.015059398021548986, + "grad_norm": 0.24200642108917236, + "learning_rate": 2.9208860844598684e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007839785004034639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036519365850836037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019331506453454494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9465132415294647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8920, + "weighted_flow_matching_loss": 0.015059398021548986, + "weighted_rkd_loss": 0.018297467008233072 + }, + { + "epoch": 2.598196101251091, + "flow_matching_loss": 0.016993033047765494, + "grad_norm": 0.20728926360607147, + "learning_rate": 2.9206209218057543e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008357818983495236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038161265198141336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020531764440238475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9463953733444214, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8930, + "weighted_flow_matching_loss": 0.016993033047765494, + "weighted_rkd_loss": 0.019431250263005496 + }, + { + "epoch": 2.6011056153622345, + "flow_matching_loss": 0.013659597374498844, + "grad_norm": 0.24363648891448975, + "learning_rate": 2.9203553276045135e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008181119011715055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037253716960549354, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02008760962635279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9462773740291596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8940, + "weighted_flow_matching_loss": 0.013659597374498844, + "weighted_rkd_loss": 0.01900847153738141 + }, + { + "epoch": 2.604015129473378, + "flow_matching_loss": 0.015477843768894672, + "grad_norm": 0.26317235827445984, + "learning_rate": 2.9200893019368263e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008500710409134627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004014146327972412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021015567053109408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9461592555046081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8950, + "weighted_flow_matching_loss": 0.015477843768894672, + "weighted_rkd_loss": 0.019884124770760535 + }, + { + "epoch": 2.6069246435845215, + "flow_matching_loss": 0.013752040918916464, + "grad_norm": 0.22855038940906525, + "learning_rate": 2.9198228448835045e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008815296180546283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004322023247368634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021952616050839424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9460410118103028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8960, + "weighted_flow_matching_loss": 0.013752040918916464, + "weighted_rkd_loss": 0.020768060628324748 + }, + { + "epoch": 2.6098341576956647, + "flow_matching_loss": 0.014209198579192162, + "grad_norm": 0.2281920462846756, + "learning_rate": 2.9195559565254908e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00923303929157555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004345078906044364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022811157256364824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9459226489067077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8970, + "weighted_flow_matching_loss": 0.014209198579192162, + "weighted_rkd_loss": 0.021577613335102795 + }, + { + "epoch": 2.6127436718068084, + "flow_matching_loss": 0.015283089596778154, + "grad_norm": 0.22868452966213226, + "learning_rate": 2.919288636943858e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009301549335941673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044511786429211496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02305427733808756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9458041727542877, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8980, + "weighted_flow_matching_loss": 0.015283089596778154, + "weighted_rkd_loss": 0.021804817765951157 + }, + { + "epoch": 2.6156531859179517, + "flow_matching_loss": 0.017145647387951612, + "grad_norm": 0.36228275299072266, + "learning_rate": 2.9190208862198126e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007799727004021406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003633242379873991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01923269610852003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9456855595111847, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8990, + "weighted_flow_matching_loss": 0.017145647387951612, + "weighted_rkd_loss": 0.01818809797987342 + }, + { + "epoch": 2.618562700029095, + "flow_matching_loss": 0.01437198705971241, + "grad_norm": 0.23377877473831177, + "learning_rate": 2.9187527044346883e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007824657671153545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035962054040282964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019245520792901517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9455668270587921, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9000, + "weighted_flow_matching_loss": 0.01437198705971241, + "weighted_rkd_loss": 0.01819795435294509 + }, + { + "epoch": 2.6214722141402387, + "flow_matching_loss": 0.015619212202727795, + "grad_norm": 0.30025625228881836, + "learning_rate": 2.918484091669953e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00874984273687005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003996611828915775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0214962974190712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9454479813575745, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9010, + "weighted_flow_matching_loss": 0.015619212202727795, + "weighted_rkd_loss": 0.020323636196553706 + }, + { + "epoch": 2.624381728251382, + "flow_matching_loss": 0.013459864724427461, + "grad_norm": 0.21011577546596527, + "learning_rate": 2.9182150480072026e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007106699422001839, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031782113946974277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017391610238701106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9453289985656739, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9020, + "weighted_flow_matching_loss": 0.013459864724427461, + "weighted_rkd_loss": 0.016440781485289335 + }, + { + "epoch": 2.627291242362525, + "flow_matching_loss": 0.014514406584203244, + "grad_norm": 0.26440173387527466, + "learning_rate": 2.917945573528167e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008059403393417597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035357710905373095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01965457759797573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9452099084854126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9030, + "weighted_flow_matching_loss": 0.014514406584203244, + "weighted_rkd_loss": 0.018577634263783692 + }, + { + "epoch": 2.630200756473669, + "flow_matching_loss": 0.017982880212366582, + "grad_norm": 0.21129253506660461, + "learning_rate": 2.9176756683147052e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008512065885588527, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038592419121414425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020883373729884625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9450906872749328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9040, + "weighted_flow_matching_loss": 0.017982880212366582, + "weighted_rkd_loss": 0.019736695941537617 + }, + { + "epoch": 2.633110270584812, + "flow_matching_loss": 0.015535193029791117, + "grad_norm": 0.2602657377719879, + "learning_rate": 2.9174053324488063e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008282575896009803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003750531608238816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02031568344682455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9449713468551636, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9050, + "weighted_flow_matching_loss": 0.015535193029791117, + "weighted_rkd_loss": 0.019197709485888483 + }, + { + "epoch": 2.636019784695956, + "flow_matching_loss": 0.015399298910051585, + "grad_norm": 0.23663298785686493, + "learning_rate": 2.917134566012592e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008367490069940687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037106462521478535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020445626601576804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9448518753051758, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9060, + "weighted_flow_matching_loss": 0.015399298910051585, + "weighted_rkd_loss": 0.019318060018122196 + }, + { + "epoch": 2.638929298807099, + "flow_matching_loss": 0.013914082292467355, + "grad_norm": 0.21304374933242798, + "learning_rate": 2.916863369088314e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008531785290688276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003740287246182561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020803857780992984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9447322964668274, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9070, + "weighted_flow_matching_loss": 0.013914082292467355, + "weighted_rkd_loss": 0.019654053170233964 + }, + { + "epoch": 2.641838812918243, + "flow_matching_loss": 0.014840823598206043, + "grad_norm": 0.22884659469127655, + "learning_rate": 2.9165917417583544e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008274697605520487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003679567691870034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020228962786495684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.944612592458725, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9080, + "weighted_flow_matching_loss": 0.014840823598206043, + "weighted_rkd_loss": 0.01910854671150446 + }, + { + "epoch": 2.644748327029386, + "flow_matching_loss": 0.015130392834544181, + "grad_norm": 0.2054426670074463, + "learning_rate": 2.916319684105227e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008811316639184951, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042085405206307765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021831173822283743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9444927632808685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9090, + "weighted_flow_matching_loss": 0.015130392834544181, + "weighted_rkd_loss": 0.020619387179613112 + }, + { + "epoch": 2.6476578411405294, + "flow_matching_loss": 0.014891911391168832, + "grad_norm": 0.27548840641975403, + "learning_rate": 2.916047196211575e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008251840248703957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038504726253449917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020354153402149678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9443728148937225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9100, + "weighted_flow_matching_loss": 0.014891911391168832, + "weighted_rkd_loss": 0.019221926014870404 + }, + { + "epoch": 2.650567355251673, + "flow_matching_loss": 0.015084930136799813, + "grad_norm": 0.21482008695602417, + "learning_rate": 2.915774278160173e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00820436947979033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036918865283951164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020100625697523355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.944252747297287, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9110, + "weighted_flow_matching_loss": 0.015084930136799813, + "weighted_rkd_loss": 0.01898005548864603 + }, + { + "epoch": 2.6534768693628163, + "flow_matching_loss": 0.01620988016948104, + "grad_norm": 0.2313072681427002, + "learning_rate": 2.9155009300339268e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008719706302508711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041494680335745215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021588880755007266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9441325485706329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9120, + "weighted_flow_matching_loss": 0.01620988016948104, + "weighted_rkd_loss": 0.020382715761661528 + }, + { + "epoch": 2.6563863834739596, + "flow_matching_loss": 0.015042982622981071, + "grad_norm": 0.2620067000389099, + "learning_rate": 2.9152271519158724e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00827081180177629, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004007213865406811, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020548837445676326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9440122246742249, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9130, + "weighted_flow_matching_loss": 0.015042982622981071, + "weighted_rkd_loss": 0.019398327078670265 + }, + { + "epoch": 2.6592958975851033, + "flow_matching_loss": 0.01496717631816864, + "grad_norm": 0.23844996094703674, + "learning_rate": 2.9149529438891764e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007962524984031915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035401996225118636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01946524940431118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9438917994499206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9140, + "weighted_flow_matching_loss": 0.01496717631816864, + "weighted_rkd_loss": 0.01837308770045638 + }, + { + "epoch": 2.6622054116962466, + "flow_matching_loss": 0.014837906882166862, + "grad_norm": 0.24529820680618286, + "learning_rate": 2.9146783060371352e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076677457895129916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038852741941809653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019220765866339206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9437712371349335, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9150, + "weighted_flow_matching_loss": 0.014837906882166862, + "weighted_rkd_loss": 0.01814003400504589 + }, + { + "epoch": 2.6651149258073903, + "flow_matching_loss": 0.016199540812522173, + "grad_norm": 0.24908754229545593, + "learning_rate": 2.914403238443177e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008510210178792476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040441953344270585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021064615808427334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9436505615711213, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9160, + "weighted_flow_matching_loss": 0.016199540812522173, + "weighted_rkd_loss": 0.019877601135522126 + }, + { + "epoch": 2.6680244399185336, + "flow_matching_loss": 0.01657465463504195, + "grad_norm": 0.27630624175071716, + "learning_rate": 2.9141277411908607e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008197750709950925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040825383039191365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020478039421141146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9435297548770905, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9170, + "weighted_flow_matching_loss": 0.01657465463504195, + "weighted_rkd_loss": 0.019321649614721537 + }, + { + "epoch": 2.6709339540296773, + "flow_matching_loss": 0.015665275137871505, + "grad_norm": 0.2300502359867096, + "learning_rate": 2.9138518143638744e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008028953475877643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038284144131466745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019886321388185024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9434088170528412, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9180, + "weighted_flow_matching_loss": 0.015665275137871505, + "weighted_rkd_loss": 0.018760892748832702 + }, + { + "epoch": 2.6738434681408205, + "flow_matching_loss": 0.015011941455304623, + "grad_norm": 0.27800849080085754, + "learning_rate": 2.9135754580460373e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008430678211152554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003990580048412085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020851936284452678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9432877838611603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9190, + "weighted_flow_matching_loss": 0.015011941455304623, + "weighted_rkd_loss": 0.01966936718672514 + }, + { + "epoch": 2.676752982251964, + "flow_matching_loss": 0.015766989812254904, + "grad_norm": 0.20076970756053925, + "learning_rate": 2.9132986723212992e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072659187018871306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003289127186872065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017820964474231003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9431666195392608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9200, + "weighted_flow_matching_loss": 0.015766989812254904, + "weighted_rkd_loss": 0.01680817026644945 + }, + { + "epoch": 2.6796624963631075, + "flow_matching_loss": 0.014388030208647252, + "grad_norm": 0.19753186404705048, + "learning_rate": 2.9130214572737403e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008454830991104246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00354709152597934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020456753484904767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9430453300476074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9210, + "weighted_flow_matching_loss": 0.014388030208647252, + "weighted_rkd_loss": 0.019291667453944682 + }, + { + "epoch": 2.6825720104742508, + "flow_matching_loss": 0.015520013589411973, + "grad_norm": 0.21655677258968353, + "learning_rate": 2.9127438129875717e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008229007897898554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003711790801025927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020169806852936744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9429239213466645, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9220, + "weighted_flow_matching_loss": 0.015520013589411973, + "weighted_rkd_loss": 0.019018590077757837 + }, + { + "epoch": 2.685481524585394, + "flow_matching_loss": 0.015164948906749487, + "grad_norm": 0.22921989858150482, + "learning_rate": 2.912465739547134e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007211965974420309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032991045620292423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017723036464303732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.942802369594574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9230, + "weighted_flow_matching_loss": 0.015164948906749487, + "weighted_rkd_loss": 0.016709368396550416 + }, + { + "epoch": 2.6883910386965377, + "flow_matching_loss": 0.015661293361335992, + "grad_norm": 0.25227877497673035, + "learning_rate": 2.9121872370368988e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008448320953175426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003916961909271777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020813603512942792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9426807403564453, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9240, + "weighted_flow_matching_loss": 0.015661293361335992, + "weighted_rkd_loss": 0.01962057054042816 + }, + { + "epoch": 2.691300552807681, + "flow_matching_loss": 0.01614554189145565, + "grad_norm": 0.23036210238933563, + "learning_rate": 2.911908305541467e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007297018123790622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034020239720121024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01799606028944254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9425589621067048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9250, + "weighted_flow_matching_loss": 0.01614554189145565, + "weighted_rkd_loss": 0.01696231318637729 + }, + { + "epoch": 2.6942100669188247, + "flow_matching_loss": 0.014432500395923854, + "grad_norm": 0.2138209044933319, + "learning_rate": 2.9116289451455717e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007325656432658434, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032519335160031914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01790324654430151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9424370646476745, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9260, + "weighted_flow_matching_loss": 0.014432500395923854, + "weighted_rkd_loss": 0.01687266193330288 + }, + { + "epoch": 2.697119581029968, + "flow_matching_loss": 0.014818779844790696, + "grad_norm": 0.23980484902858734, + "learning_rate": 2.911349155934075e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076122012920677665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003587480983696878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018811883591115474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9423150479793548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9270, + "weighted_flow_matching_loss": 0.014818779844790696, + "weighted_rkd_loss": 0.017726745828986167 + }, + { + "epoch": 2.7000290951411117, + "flow_matching_loss": 0.013592899218201638, + "grad_norm": 0.22609958052635193, + "learning_rate": 2.9110689379919687e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00862080198712647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038326639449223878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021074267849326134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9421929180622101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9280, + "weighted_flow_matching_loss": 0.013592899218201638, + "weighted_rkd_loss": 0.019856098759919404 + }, + { + "epoch": 2.702938609252255, + "flow_matching_loss": 0.013958702236413956, + "grad_norm": 0.18753845989704132, + "learning_rate": 2.9107882914043767e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00801010588183999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003637346159666777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019657557643949986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9420706570148468, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9290, + "weighted_flow_matching_loss": 0.013958702236413956, + "weighted_rkd_loss": 0.01851885160431266 + }, + { + "epoch": 2.705848123363398, + "flow_matching_loss": 0.013738995790481568, + "grad_norm": 0.22903956472873688, + "learning_rate": 2.910507216256551e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00808725943788886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003942901431582868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02011742042377591, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9419482767581939, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9300, + "weighted_flow_matching_loss": 0.013738995790481568, + "weighted_rkd_loss": 0.018949614465236665 + }, + { + "epoch": 2.708757637474542, + "flow_matching_loss": 0.015992534067481756, + "grad_norm": 0.27418237924575806, + "learning_rate": 2.910225712633876e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008282861579209566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003836009302176535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0204017324373126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9418257713317871, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9310, + "weighted_flow_matching_loss": 0.015992534067481756, + "weighted_rkd_loss": 0.019214857276529072 + }, + { + "epoch": 2.711667151585685, + "flow_matching_loss": 0.014281109906733036, + "grad_norm": 0.21403096616268158, + "learning_rate": 2.9099437806218637e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008072452479973435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00354241905733943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019687323831021786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9417031466960907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9320, + "weighted_flow_matching_loss": 0.014281109906733036, + "weighted_rkd_loss": 0.018539638817310335 + }, + { + "epoch": 2.7145766656968284, + "flow_matching_loss": 0.014333433378487826, + "grad_norm": 0.24100208282470703, + "learning_rate": 2.9096614203061584e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007912558084353805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034085958730429413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01923371199518442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9415804147720337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9330, + "weighted_flow_matching_loss": 0.014333433378487826, + "weighted_rkd_loss": 0.018110141064971684 + }, + { + "epoch": 2.717486179807972, + "flow_matching_loss": 0.016660079732537268, + "grad_norm": 0.22306570410728455, + "learning_rate": 2.9093786317725336e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007650272315368056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037178166676312685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019018361158668996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9414575338363648, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9340, + "weighted_flow_matching_loss": 0.016660079732537268, + "weighted_rkd_loss": 0.017904977407306434 + }, + { + "epoch": 2.7203956939191154, + "flow_matching_loss": 0.014854325354099274, + "grad_norm": 0.22893179953098297, + "learning_rate": 2.909095415106893e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008347246935591102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004115014639683068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020809508487582208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9413345575332641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9350, + "weighted_flow_matching_loss": 0.014854325354099274, + "weighted_rkd_loss": 0.019588682614266873 + }, + { + "epoch": 2.723305208030259, + "flow_matching_loss": 0.015025437250733376, + "grad_norm": 0.23129506409168243, + "learning_rate": 2.9088117703952698e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00831406395882368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039268650347366926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02055499292910099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9412114381790161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9360, + "weighted_flow_matching_loss": 0.015025437250733376, + "weighted_rkd_loss": 0.01934668542817235 + }, + { + "epoch": 2.7262147221414024, + "flow_matching_loss": 0.015347407665103674, + "grad_norm": 0.22810569405555725, + "learning_rate": 2.908527697723829e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008320920588448643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003663448756560683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020305289793759583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9410882115364074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9370, + "weighted_flow_matching_loss": 0.015347407665103674, + "weighted_rkd_loss": 0.01910912673920393 + }, + { + "epoch": 2.729124236252546, + "flow_matching_loss": 0.015272013284265995, + "grad_norm": 0.1927763819694519, + "learning_rate": 2.9082431971788635e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008146708458662033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037362601608037947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02002967717126012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9409648478031158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9380, + "weighted_flow_matching_loss": 0.015272013284265995, + "weighted_rkd_loss": 0.01884725969284773 + }, + { + "epoch": 2.7320337503636893, + "flow_matching_loss": 0.015142657700926066, + "grad_norm": 0.20551402866840363, + "learning_rate": 2.9079582688467972e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008715598285198212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039160103304311635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021347206830978394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.940841406583786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9390, + "weighted_flow_matching_loss": 0.015142657700926066, + "weighted_rkd_loss": 0.02008434049785137 + }, + { + "epoch": 2.7349432644748326, + "flow_matching_loss": 0.016444596741348506, + "grad_norm": 0.31235313415527344, + "learning_rate": 2.9076729128141833e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008833350474014878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004046810674481094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02171351183205843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9407178103923798, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9400, + "weighted_flow_matching_loss": 0.016444596741348506, + "weighted_rkd_loss": 0.02042615953832865 + }, + { + "epoch": 2.7378527785859763, + "flow_matching_loss": 0.013167983246967196, + "grad_norm": 0.20418760180473328, + "learning_rate": 2.9073871291677063e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008152990788221359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003709800494834781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020015781745314598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9405941069126129, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9410, + "weighted_flow_matching_loss": 0.013167983246967196, + "weighted_rkd_loss": 0.018826774507761 + }, + { + "epoch": 2.7407622926971196, + "flow_matching_loss": 0.013592079188674689, + "grad_norm": 0.20067353546619415, + "learning_rate": 2.9071009179941797e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007885638019070029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035203545819967987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019291630666702985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9404702603816986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9420, + "weighted_flow_matching_loss": 0.013592079188674689, + "weighted_rkd_loss": 0.0181432468816638 + }, + { + "epoch": 2.743671806808263, + "flow_matching_loss": 0.015034094545990228, + "grad_norm": 0.21264208853244781, + "learning_rate": 2.906814279380546e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007736909063532948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037592239677906036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0192330420948565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9403463006019592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9430, + "weighted_flow_matching_loss": 0.015034094545990228, + "weighted_rkd_loss": 0.018085692171007396 + }, + { + "epoch": 2.7465813209194065, + "flow_matching_loss": 0.013325569592416287, + "grad_norm": 0.21721145510673523, + "learning_rate": 2.9065272134138795e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076026688795536755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032663010293617844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018471638672053814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9402222394943237, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9440, + "weighted_flow_matching_loss": 0.013325569592416287, + "weighted_rkd_loss": 0.017367437854409216 + }, + { + "epoch": 2.74949083503055, + "flow_matching_loss": 0.01460236357524991, + "grad_norm": 0.18797679245471954, + "learning_rate": 2.9062397201813823e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006996125914156437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031626621494069695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017154913768172263, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9400980472564697, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9450, + "weighted_flow_matching_loss": 0.01460236357524991, + "weighted_rkd_loss": 0.01612725676968694 + }, + { + "epoch": 2.752400349141693, + "flow_matching_loss": 0.015747878514230252, + "grad_norm": 0.27228355407714844, + "learning_rate": 2.905951799770388e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008225376624614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035175876691937447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019968340918421746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9399737298488617, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9460, + "weighted_flow_matching_loss": 0.015747878514230252, + "weighted_rkd_loss": 0.01876968629658222 + }, + { + "epoch": 2.7553098632528368, + "flow_matching_loss": 0.016031747218221426, + "grad_norm": 0.1911093145608902, + "learning_rate": 2.9056634522683586e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009866007650271059, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004404771188274026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024136786814779043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9398492932319641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9470, + "weighted_flow_matching_loss": 0.016031747218221426, + "weighted_rkd_loss": 0.022684939019382 + }, + { + "epoch": 2.7582193773639805, + "flow_matching_loss": 0.01556932833045721, + "grad_norm": 0.20237937569618225, + "learning_rate": 2.9053746777628872e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007997837848961353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035578377079218625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019553513545542956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9397247433662415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9480, + "weighted_flow_matching_loss": 0.01556932833045721, + "weighted_rkd_loss": 0.018374953605234622 + }, + { + "epoch": 2.7611288914751237, + "flow_matching_loss": 0.01678880862891674, + "grad_norm": 0.26211586594581604, + "learning_rate": 2.9050854763416952e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008013063902035355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035981557331979273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019624283630400897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9396000623703002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9490, + "weighted_flow_matching_loss": 0.01678880862891674, + "weighted_rkd_loss": 0.01843898808583617 + }, + { + "epoch": 2.764038405586267, + "flow_matching_loss": 0.016672473400831223, + "grad_norm": 0.24618268013000488, + "learning_rate": 2.9047958480926346e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008390569966286421, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003945787763223052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020726927556097507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9394752681255341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9500, + "weighted_flow_matching_loss": 0.016672473400831223, + "weighted_rkd_loss": 0.019472398050129413 + }, + { + "epoch": 2.7669479196974107, + "flow_matching_loss": 0.014267408289015293, + "grad_norm": 0.2298816293478012, + "learning_rate": 2.9045057931036874e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008219308778643608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036174432840198278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02005606098100543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9393503427505493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9510, + "weighted_flow_matching_loss": 0.014267408289015293, + "weighted_rkd_loss": 0.0188397197984159 + }, + { + "epoch": 2.769857433808554, + "flow_matching_loss": 0.017839008755981923, + "grad_norm": 0.23267383873462677, + "learning_rate": 2.9042153114629636e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00779453688301146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003672933182679117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019262006878852843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9392253041267395, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9520, + "weighted_flow_matching_loss": 0.017839008755981923, + "weighted_rkd_loss": 0.018091386090964078 + }, + { + "epoch": 2.7727669479196972, + "flow_matching_loss": 0.01593355620279908, + "grad_norm": 0.246006578207016, + "learning_rate": 2.9039244032587043e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00828381348401308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037392926635220645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020306919515132905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9391001462936401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9530, + "weighted_flow_matching_loss": 0.01593355620279908, + "weighted_rkd_loss": 0.01907018106430769 + }, + { + "epoch": 2.775676462030841, + "flow_matching_loss": 0.017344502732157707, + "grad_norm": 0.23433050513267517, + "learning_rate": 2.90363306857928e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008266690606251358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037109819473698734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020244363136589526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9389748692512512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9540, + "weighted_flow_matching_loss": 0.017344502732157707, + "weighted_rkd_loss": 0.019008946418762208 + }, + { + "epoch": 2.778585976141984, + "flow_matching_loss": 0.013869765400886535, + "grad_norm": 0.2076636552810669, + "learning_rate": 2.90334130751319e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00790948630310595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036480163224041464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01946698911488056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9388494729995728, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9550, + "weighted_flow_matching_loss": 0.013869765400886535, + "weighted_rkd_loss": 0.01827652733772993 + }, + { + "epoch": 2.7814954902531275, + "flow_matching_loss": 0.014340422581881285, + "grad_norm": 0.18428067862987518, + "learning_rate": 2.903049120149064e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007579413242638111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003605600423179567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018764426931738853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9387239515781403, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9560, + "weighted_flow_matching_loss": 0.014340422581881285, + "weighted_rkd_loss": 0.017614634335041048 + }, + { + "epoch": 2.784405004364271, + "flow_matching_loss": 0.01617085551843047, + "grad_norm": 0.21870747208595276, + "learning_rate": 2.9027565065756608e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007777992868795991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035386798437684773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01909466590732336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9385982990264893, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9570, + "weighted_flow_matching_loss": 0.01617085551843047, + "weighted_rkd_loss": 0.01792220016941428 + }, + { + "epoch": 2.7873145184754144, + "flow_matching_loss": 0.016103522572666406, + "grad_norm": 0.1952124685049057, + "learning_rate": 2.9024634668818682e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008311959449201822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003628575336188078, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02025249432772398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9384725391864777, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9580, + "weighted_flow_matching_loss": 0.016103522572666406, + "weighted_rkd_loss": 0.019006471149623394 + }, + { + "epoch": 2.790224032586558, + "flow_matching_loss": 0.015080259833484887, + "grad_norm": 0.23044684529304504, + "learning_rate": 2.9021700011567045e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00785434851422906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035622023744508622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019270899519324303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.938346666097641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9590, + "weighted_flow_matching_loss": 0.015080259833484887, + "weighted_rkd_loss": 0.018082737922668457 + }, + { + "epoch": 2.7931335466977014, + "flow_matching_loss": 0.015642821788787842, + "grad_norm": 0.19806620478630066, + "learning_rate": 2.9018761094893166e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007948312675580382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003522805287502706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019419430755078793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9382206737995148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9600, + "weighted_flow_matching_loss": 0.015642821788787842, + "weighted_rkd_loss": 0.0182197330519557 + }, + { + "epoch": 2.796043060808845, + "flow_matching_loss": 0.01649662097916007, + "grad_norm": 0.29630348086357117, + "learning_rate": 2.9015817919689808e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00834451401606202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003868663753382862, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020557691901922227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9380945444107056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9610, + "weighted_flow_matching_loss": 0.01649662097916007, + "weighted_rkd_loss": 0.019285026006400585 + }, + { + "epoch": 2.7989525749199884, + "flow_matching_loss": 0.01536101959645748, + "grad_norm": 0.21250897645950317, + "learning_rate": 2.9012870486851034e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00699658589437604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033533809706568717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01734655275940895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9379683017730713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9620, + "weighted_flow_matching_loss": 0.01536101959645748, + "weighted_rkd_loss": 0.016270489990711214 + }, + { + "epoch": 2.8018620890311317, + "flow_matching_loss": 0.01599358571693301, + "grad_norm": 0.21378540992736816, + "learning_rate": 2.9009918797272195e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008063251664862037, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036634313175454734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019789934623986483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.937841945886612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9630, + "weighted_flow_matching_loss": 0.01599358571693301, + "weighted_rkd_loss": 0.018559772241860627 + }, + { + "epoch": 2.8047716031422754, + "flow_matching_loss": 0.015157362073659897, + "grad_norm": 0.22307763993740082, + "learning_rate": 2.9006962851849933e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007762358337640762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035350482445210218, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019059764966368676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9377154648303986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9640, + "weighted_flow_matching_loss": 0.015157362073659897, + "weighted_rkd_loss": 0.017872688919305803 + }, + { + "epoch": 2.8076811172534186, + "flow_matching_loss": 0.01564261307939887, + "grad_norm": 0.21388986706733704, + "learning_rate": 2.9004002651482192e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075075964909046885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034430508967489004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01845824355259538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9375888705253601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9650, + "weighted_flow_matching_loss": 0.01564261307939887, + "weighted_rkd_loss": 0.017306268401443958 + }, + { + "epoch": 2.810590631364562, + "flow_matching_loss": 0.0149748170748353, + "grad_norm": 0.2471323013305664, + "learning_rate": 2.9001038197068198e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00814752415753901, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038702564081177114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020165304746478797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9374621450901032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9660, + "weighted_flow_matching_loss": 0.0149748170748353, + "weighted_rkd_loss": 0.018904173839837314 + }, + { + "epoch": 2.8135001454757056, + "flow_matching_loss": 0.01677692113444209, + "grad_norm": 0.21910269558429718, + "learning_rate": 2.899806948950848e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007338035479187965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034142226446419953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018090293556451798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9373353064060211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9670, + "weighted_flow_matching_loss": 0.01677692113444209, + "weighted_rkd_loss": 0.016956608463078737 + }, + { + "epoch": 2.816409659586849, + "flow_matching_loss": 0.014949539862573146, + "grad_norm": 0.21082803606987, + "learning_rate": 2.8995096529704846e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008032201929017902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037830870831385257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019847490731626748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9372083425521851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9680, + "weighted_flow_matching_loss": 0.014949539862573146, + "weighted_rkd_loss": 0.01860123025253415 + }, + { + "epoch": 2.8193191736979926, + "flow_matching_loss": 0.015377624798566103, + "grad_norm": 0.2280951887369156, + "learning_rate": 2.8992119318560403e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007278277352452278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003460975387133658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018017529882490634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.937081253528595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9690, + "weighted_flow_matching_loss": 0.015377624798566103, + "weighted_rkd_loss": 0.016883928794413805 + }, + { + "epoch": 2.822228687809136, + "flow_matching_loss": 0.01412342321127653, + "grad_norm": 0.1922813504934311, + "learning_rate": 2.8989137856979555e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007623725570738316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032479271059855817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018495378270745277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9369540512561798, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9700, + "weighted_flow_matching_loss": 0.01412342321127653, + "weighted_rkd_loss": 0.017329417075961827 + }, + { + "epoch": 2.8251382019202795, + "flow_matching_loss": 0.014642448257654905, + "grad_norm": 0.20671610534191132, + "learning_rate": 2.8986152145867983e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008065386768430471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036222388735041022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01975301243364811, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9368267357349396, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9710, + "weighted_flow_matching_loss": 0.014642448257654905, + "weighted_rkd_loss": 0.018505183421075343 + }, + { + "epoch": 2.828047716031423, + "flow_matching_loss": 0.01601398317143321, + "grad_norm": 0.24043279886245728, + "learning_rate": 2.8983162186132678e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007699431478977203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035396337043493986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018938496708869934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9366993069648742, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9720, + "weighted_flow_matching_loss": 0.01601398317143321, + "weighted_rkd_loss": 0.017739659734070302 + }, + { + "epoch": 2.830957230142566, + "flow_matching_loss": 0.014518107566982508, + "grad_norm": 0.21695435047149658, + "learning_rate": 2.8980167978681905e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007829098682850599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033336152555420995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01899181241169572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9365717470645905, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9730, + "weighted_flow_matching_loss": 0.014518107566982508, + "weighted_rkd_loss": 0.01778721297159791 + }, + { + "epoch": 2.8338667442537098, + "flow_matching_loss": 0.014781202748417854, + "grad_norm": 0.23946966230869293, + "learning_rate": 2.897716952442522e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007695823768153786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036150490399450065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01900669652968645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9364440619945527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9740, + "weighted_flow_matching_loss": 0.014781202748417854, + "weighted_rkd_loss": 0.017798675689846277 + }, + { + "epoch": 2.836776258364853, + "flow_matching_loss": 0.015365883056074382, + "grad_norm": 0.2756076455116272, + "learning_rate": 2.8974166824273482e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007408373430371285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034040240570902824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018220771010965107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9363162517547607, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9750, + "weighted_flow_matching_loss": 0.015365883056074382, + "weighted_rkd_loss": 0.01706041982397437 + }, + { + "epoch": 2.8396857724759963, + "flow_matching_loss": 0.014219910744577647, + "grad_norm": 0.21856307983398438, + "learning_rate": 2.897115987913883e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00776025727391243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003469783114269376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01899029780179262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9361883342266083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9760, + "weighted_flow_matching_loss": 0.014219910744577647, + "weighted_rkd_loss": 0.017778514325618743 + }, + { + "epoch": 2.84259528658714, + "flow_matching_loss": 0.014869870524853468, + "grad_norm": 0.22742605209350586, + "learning_rate": 2.896814868993469e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008267214288935065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037528524175286294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020287280902266502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9360603094100952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9770, + "weighted_flow_matching_loss": 0.014869870524853468, + "weighted_rkd_loss": 0.018990178778767585 + }, + { + "epoch": 2.8455048006982833, + "flow_matching_loss": 0.01585245169699192, + "grad_norm": 0.21946528553962708, + "learning_rate": 2.896513325757579e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008589947083964943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038581694476306438, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021038063429296018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9359321475028992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9780, + "weighted_flow_matching_loss": 0.01585245169699192, + "weighted_rkd_loss": 0.01969020487740636 + }, + { + "epoch": 2.848414314809427, + "flow_matching_loss": 0.014879389386624097, + "grad_norm": 0.21811561286449432, + "learning_rate": 2.896211358297813e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008134236140176655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003462136210873723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01973060853779316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9358038663864136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9790, + "weighted_flow_matching_loss": 0.014879389386624097, + "weighted_rkd_loss": 0.018464006297290326 + }, + { + "epoch": 2.8513238289205702, + "flow_matching_loss": 0.017414710484445095, + "grad_norm": 0.23978672921657562, + "learning_rate": 2.8959089667059014e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008063228195533156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035421869717538355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019668643549084663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9356754720211029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9800, + "weighted_flow_matching_loss": 0.017414710484445095, + "weighted_rkd_loss": 0.01840343726798892 + }, + { + "epoch": 2.854233343031714, + "flow_matching_loss": 0.01322843162342906, + "grad_norm": 0.30957332253456116, + "learning_rate": 2.8956061510737027e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007289334433153272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033702240791171787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017948892805725335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9355469524860383, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9810, + "weighted_flow_matching_loss": 0.01322843162342906, + "weighted_rkd_loss": 0.016791993379592897 + }, + { + "epoch": 2.857142857142857, + "flow_matching_loss": 0.014379788376390933, + "grad_norm": 0.2801324725151062, + "learning_rate": 2.8953029114932042e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007984832767397166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004083911911584437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020053577329963445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9354183077812195, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9820, + "weighted_flow_matching_loss": 0.014379788376390933, + "weighted_rkd_loss": 0.018758385628461837 + }, + { + "epoch": 2.8600523712540005, + "flow_matching_loss": 0.014597674179822207, + "grad_norm": 0.20922580361366272, + "learning_rate": 2.8949992480565214e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007622909778729081, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036016473546624183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018847466632723808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9352895557880402, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9830, + "weighted_flow_matching_loss": 0.014597674179822207, + "weighted_rkd_loss": 0.01762785781174898 + }, + { + "epoch": 2.862961885365144, + "flow_matching_loss": 0.014811067841947079, + "grad_norm": 0.1870385855436325, + "learning_rate": 2.894695160855901e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007941404404118658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003393741603940725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019276550225913524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9351606845855713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9840, + "weighted_flow_matching_loss": 0.014811067841947079, + "weighted_rkd_loss": 0.01802673377096653 + }, + { + "epoch": 2.8658713994762874, + "flow_matching_loss": 0.015826540905982255, + "grad_norm": 0.24846601486206055, + "learning_rate": 2.8943906499837146e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007165428623557091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003393374290317297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017724231537431478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9350316941738128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9850, + "weighted_flow_matching_loss": 0.015826540905982255, + "weighted_rkd_loss": 0.016572674456983804 + }, + { + "epoch": 2.8687809135874307, + "flow_matching_loss": 0.01536856135353446, + "grad_norm": 0.2182813435792923, + "learning_rate": 2.8940857155324655e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00786486784927547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034342426341027023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019163978286087513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9349025845527649, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9860, + "weighted_flow_matching_loss": 0.01536856135353446, + "weighted_rkd_loss": 0.017916475981473924 + }, + { + "epoch": 2.8716904276985744, + "flow_matching_loss": 0.014562891703099013, + "grad_norm": 0.19299501180648804, + "learning_rate": 2.893780357594785e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007334815990179777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032643319806084036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017933963891118766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9347733497619629, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9870, + "weighted_flow_matching_loss": 0.014562891703099013, + "weighted_rkd_loss": 0.016764188278466464 + }, + { + "epoch": 2.8745999418097177, + "flow_matching_loss": 0.014959083497524261, + "grad_norm": 0.22358208894729614, + "learning_rate": 2.8934745762634326e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008122295001521706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003418012522161007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01966260252520442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9346439957618713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9880, + "weighted_flow_matching_loss": 0.014959083497524261, + "weighted_rkd_loss": 0.018377518747001886 + }, + { + "epoch": 2.8775094559208614, + "flow_matching_loss": 0.014819532912224532, + "grad_norm": 0.23624064028263092, + "learning_rate": 2.8931683716312964e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00793524831533432, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036571234464645385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01952762035652995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9345145225524902, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9890, + "weighted_flow_matching_loss": 0.014819532912224532, + "weighted_rkd_loss": 0.018248849920928477 + }, + { + "epoch": 2.8804189700320046, + "flow_matching_loss": 0.015243404265493154, + "grad_norm": 0.23041583597660065, + "learning_rate": 2.8928617437913932e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007753891265019774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003750990401022136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019258773326873778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9343849420547485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9900, + "weighted_flow_matching_loss": 0.015243404265493154, + "weighted_rkd_loss": 0.017995110992342234 + }, + { + "epoch": 2.8833284841431484, + "flow_matching_loss": 0.014511201903223992, + "grad_norm": 0.2515402138233185, + "learning_rate": 2.8925546928368688e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007748604007065296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003624518564902246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019121726416051386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9342552244663238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9910, + "weighted_flow_matching_loss": 0.014511201903223992, + "weighted_rkd_loss": 0.01786461370065808 + }, + { + "epoch": 2.8862379982542916, + "flow_matching_loss": 0.015911524277180435, + "grad_norm": 0.16839179396629333, + "learning_rate": 2.8922472188609968e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007020513014867902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032360109966248275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017277036886662246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.934125405550003, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9920, + "weighted_flow_matching_loss": 0.015911524277180435, + "weighted_rkd_loss": 0.016138933040201663 + }, + { + "epoch": 2.889147512365435, + "flow_matching_loss": 0.016420793160796164, + "grad_norm": 0.22074830532073975, + "learning_rate": 2.8919393219571805e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008096515852957963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036834257654845713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019876457378268243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9339954555034637, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9930, + "weighted_flow_matching_loss": 0.016420793160796164, + "weighted_rkd_loss": 0.01856462499126792 + }, + { + "epoch": 2.8920570264765786, + "flow_matching_loss": 0.015231913514435291, + "grad_norm": 0.22391292452812195, + "learning_rate": 2.8916310022189495e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00800196877680719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003702694596722722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0197066318243742, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9338653922080994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9940, + "weighted_flow_matching_loss": 0.015231913514435291, + "weighted_rkd_loss": 0.018403303902596237 + }, + { + "epoch": 2.894966540587722, + "flow_matching_loss": 0.016375051531940698, + "grad_norm": 0.2089516818523407, + "learning_rate": 2.891322259739964e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007082261703908443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003278364473953843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01744288792833686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9337351977825165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9950, + "weighted_flow_matching_loss": 0.016375051531940698, + "weighted_rkd_loss": 0.016287090629339217 + }, + { + "epoch": 2.897876054698865, + "flow_matching_loss": 0.015200420189648867, + "grad_norm": 0.25784358382225037, + "learning_rate": 2.891013094614012e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007912945793941617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00358927717898041, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019415169022977352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.933604896068573, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9960, + "weighted_flow_matching_loss": 0.015200420189648867, + "weighted_rkd_loss": 0.018126087542623282 + }, + { + "epoch": 2.900785568810009, + "flow_matching_loss": 0.014673666562885046, + "grad_norm": 0.20460133254528046, + "learning_rate": 2.8907035069350093e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007626034645363688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036657480988651512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018917817249894144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9334744811058044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9970, + "weighted_flow_matching_loss": 0.014673666562885046, + "weighted_rkd_loss": 0.017659291811287402 + }, + { + "epoch": 2.903695082921152, + "flow_matching_loss": 0.01417699856683612, + "grad_norm": 0.21440213918685913, + "learning_rate": 2.8903934967970007e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070712222717702385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031893322011455894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017331776954233647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9333439409732819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9980, + "weighted_flow_matching_loss": 0.01417699856683612, + "weighted_rkd_loss": 0.01617650669068098 + }, + { + "epoch": 2.9066045970322953, + "flow_matching_loss": 0.013523007836192846, + "grad_norm": 0.20582231879234314, + "learning_rate": 2.8900830642941583e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00839051795192063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038876595906913282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020668695122003554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9332132697105407, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9990, + "weighted_flow_matching_loss": 0.013523007836192846, + "weighted_rkd_loss": 0.01928827427327633 + }, + { + "epoch": 2.909514111143439, + "flow_matching_loss": 0.014064501225948333, + "grad_norm": 0.23039330542087555, + "learning_rate": 2.889772209520785e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00782481781207025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003387631895020604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019037267565727232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9330824911594391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10000, + "weighted_flow_matching_loss": 0.014064501225948333, + "weighted_rkd_loss": 0.017763339914381504 + }, + { + "epoch": 2.9124236252545828, + "flow_matching_loss": 0.015460360515862703, + "grad_norm": 0.2903507351875305, + "learning_rate": 2.8894609325713087e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008029443863779306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003879008209332824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019937895890325307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9329515933990479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10010, + "weighted_flow_matching_loss": 0.015460360515862703, + "weighted_rkd_loss": 0.018601001892238855 + }, + { + "epoch": 2.915333139365726, + "flow_matching_loss": 0.01343058617785573, + "grad_norm": 0.20325030386447906, + "learning_rate": 2.889149233540288e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007370431022718549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033915201667696236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01813238225877285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9328205823898316, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10020, + "weighted_flow_matching_loss": 0.01343058617785573, + "weighted_rkd_loss": 0.01691416148096323 + }, + { + "epoch": 2.9182426534768693, + "flow_matching_loss": 0.016176325734704733, + "grad_norm": 0.24790018796920776, + "learning_rate": 2.8888371125224084e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076463714241981505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037250891560688614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0190178319811821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9326894462108613, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10030, + "weighted_flow_matching_loss": 0.016176325734704733, + "weighted_rkd_loss": 0.01773771857842803 + }, + { + "epoch": 2.921152167588013, + "flow_matching_loss": 0.013675963599234819, + "grad_norm": 0.2245282232761383, + "learning_rate": 2.8885245696124835e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007671580417081714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035143200075253844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01885748077183962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9325581848621368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10040, + "weighted_flow_matching_loss": 0.013675963599234819, + "weighted_rkd_loss": 0.01758565353229642 + }, + { + "epoch": 2.9240616816991563, + "flow_matching_loss": 0.015966220386326314, + "grad_norm": 0.23813609778881073, + "learning_rate": 2.888211604905457e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007491297088563442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003307277918793261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01828987244516611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9324268102645874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10050, + "weighted_flow_matching_loss": 0.015966220386326314, + "weighted_rkd_loss": 0.01705395970493555 + }, + { + "epoch": 2.9269711958102995, + "flow_matching_loss": 0.014171970728784799, + "grad_norm": 0.20590238273143768, + "learning_rate": 2.8878982184963986e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006942248810082674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030188834527507423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016903381142765284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9322953343391418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10060, + "weighted_flow_matching_loss": 0.014171970728784799, + "weighted_rkd_loss": 0.01575897550210357 + }, + { + "epoch": 2.9298807099214432, + "flow_matching_loss": 0.014288220833986998, + "grad_norm": 0.20403002202510834, + "learning_rate": 2.8875844104805066e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007484693638980388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034334169700741767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01840280434116721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9321637213230133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10070, + "weighted_flow_matching_loss": 0.014288220833986998, + "weighted_rkd_loss": 0.01715450659394264 + }, + { + "epoch": 2.9327902240325865, + "flow_matching_loss": 0.016787505708634853, + "grad_norm": 0.19406042993068695, + "learning_rate": 2.887270180953107e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007620991813018918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003537210519425571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018779193982481956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9320319950580597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10080, + "weighted_flow_matching_loss": 0.016787505708634853, + "weighted_rkd_loss": 0.017502854578197002 + }, + { + "epoch": 2.9356997381437298, + "flow_matching_loss": 0.014446630887687207, + "grad_norm": 0.2428513914346695, + "learning_rate": 2.886955530009656e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007888428028672933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037477074656635523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019524563755840062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.931900143623352, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10090, + "weighted_flow_matching_loss": 0.014446630887687207, + "weighted_rkd_loss": 0.0181949564255774 + }, + { + "epoch": 2.9386092522548735, + "flow_matching_loss": 0.015712245274335145, + "grad_norm": 0.20598365366458893, + "learning_rate": 2.8866404577457352e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007639260776340961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003424747660756111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01870326902717352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9317681789398193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10100, + "weighted_flow_matching_loss": 0.015712245274335145, + "weighted_rkd_loss": 0.017427079472690822 + }, + { + "epoch": 2.9415187663660167, + "flow_matching_loss": 0.013969757966697216, + "grad_norm": 0.24756675958633423, + "learning_rate": 2.8863249642570557e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00855886903591454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004107405710965395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021225143782794477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9316361010074615, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10110, + "weighted_flow_matching_loss": 0.013969757966697216, + "weighted_rkd_loss": 0.019774146564304828 + }, + { + "epoch": 2.9444282804771604, + "flow_matching_loss": 0.016534673608839513, + "grad_norm": 0.22015061974525452, + "learning_rate": 2.8860090496394552e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007037984626367688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003187196282669902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017263165302574636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9315038859844208, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10120, + "weighted_flow_matching_loss": 0.016534673608839513, + "weighted_rkd_loss": 0.01608071895316243 + }, + { + "epoch": 2.9473377945883037, + "flow_matching_loss": 0.014224795065820217, + "grad_norm": 0.19963233172893524, + "learning_rate": 2.8856927139889015e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007557943742722273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033944402122870086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01851032776758075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9313715696334839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10130, + "weighted_flow_matching_loss": 0.014224795065820217, + "weighted_rkd_loss": 0.017239952366799115 + }, + { + "epoch": 2.9502473086994474, + "flow_matching_loss": 0.014846880454570055, + "grad_norm": 0.22266459465026855, + "learning_rate": 2.8853759574014878e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007894882746040822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035954315681010485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019385197199881077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9312391340732574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10140, + "weighted_flow_matching_loss": 0.014846880454570055, + "weighted_rkd_loss": 0.018052215222269297 + }, + { + "epoch": 2.9531568228105907, + "flow_matching_loss": 0.014957437478005885, + "grad_norm": 0.23551571369171143, + "learning_rate": 2.8850587799734372e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077801413834095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003585649118758738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019145931769162416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9311065793037414, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10150, + "weighted_flow_matching_loss": 0.014957437478005885, + "weighted_rkd_loss": 0.017826863937079905 + }, + { + "epoch": 2.956066336921734, + "flow_matching_loss": 0.016104225907474755, + "grad_norm": 0.23661009967327118, + "learning_rate": 2.8847411818011e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00788858481682837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003532628575339913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019309798162430523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.930973905324936, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10160, + "weighted_flow_matching_loss": 0.016104225907474755, + "weighted_rkd_loss": 0.017976958025246858 + }, + { + "epoch": 2.9589758510328776, + "flow_matching_loss": 0.014740936178714036, + "grad_norm": 0.21159201860427856, + "learning_rate": 2.8844231629809534e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007176153035834432, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033479692647233605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017700275219976903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9308411061763764, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10170, + "weighted_flow_matching_loss": 0.014740936178714036, + "weighted_rkd_loss": 0.016476151160895826 + }, + { + "epoch": 2.961885365144021, + "flow_matching_loss": 0.013007902912795544, + "grad_norm": 0.23479673266410828, + "learning_rate": 2.8841047236096037e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007552476646378636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003626925335265696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01873187879100442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9307081997394562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10180, + "weighted_flow_matching_loss": 0.013007902912795544, + "weighted_rkd_loss": 0.017433917429298163 + }, + { + "epoch": 2.964794879255164, + "flow_matching_loss": 0.01495766844600439, + "grad_norm": 0.22207094728946686, + "learning_rate": 2.8837858637837838e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00821170611307025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00369259063154459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020116003043949605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.930575168132782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10190, + "weighted_flow_matching_loss": 0.01495766844600439, + "weighted_rkd_loss": 0.018719525448977946 + }, + { + "epoch": 2.967704393366308, + "flow_matching_loss": 0.013652693899348377, + "grad_norm": 0.22335486114025116, + "learning_rate": 2.8834665836003556e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008151225373148918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00409401988144964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020396470557898283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9304420292377472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10200, + "weighted_flow_matching_loss": 0.013652693899348377, + "weighted_rkd_loss": 0.018977691140025853 + }, + { + "epoch": 2.970613907477451, + "flow_matching_loss": 0.01547720842063427, + "grad_norm": 0.21403664350509644, + "learning_rate": 2.8831468831563075e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006896028062328696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032626813976094128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01705473754554987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9303087592124939, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10210, + "weighted_flow_matching_loss": 0.01547720842063427, + "weighted_rkd_loss": 0.015866193547844888 + }, + { + "epoch": 2.973523421588595, + "flow_matching_loss": 0.01340879751369357, + "grad_norm": 0.25018906593322754, + "learning_rate": 2.882826762548756e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007869289536029101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037469716044142842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019485550560057165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.93017538189888, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10220, + "weighted_flow_matching_loss": 0.01340879751369357, + "weighted_rkd_loss": 0.018124955240637064 + }, + { + "epoch": 2.976432935699738, + "flow_matching_loss": 0.01477071549743414, + "grad_norm": 0.2380269169807434, + "learning_rate": 2.8825062218749456e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007989100553095341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036673529073596002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019645554199814796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9300418794155121, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10230, + "weighted_flow_matching_loss": 0.01477071549743414, + "weighted_rkd_loss": 0.01827116422355175 + }, + { + "epoch": 2.979342449810882, + "flow_matching_loss": 0.015608388744294644, + "grad_norm": 0.22928616404533386, + "learning_rate": 2.8821852612322477e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00724967191927135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003241249336861074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017740593198686837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9299082577228546, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10240, + "weighted_flow_matching_loss": 0.015608388744294644, + "weighted_rkd_loss": 0.016497102845460177 + }, + { + "epoch": 2.982251963922025, + "flow_matching_loss": 0.013912384584546088, + "grad_norm": 0.17862960696220398, + "learning_rate": 2.8818638807181622e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007349449628964067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003301133471541107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01800003247335553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9297745227813721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10250, + "weighted_flow_matching_loss": 0.013912384584546088, + "weighted_rkd_loss": 0.01673594620078802 + }, + { + "epoch": 2.9851614780331683, + "flow_matching_loss": 0.016536344401538373, + "grad_norm": 0.22892983257770538, + "learning_rate": 2.8815420804303154e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008011108124628663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038276614155620336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019849877804517746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9296406686306, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10260, + "weighted_flow_matching_loss": 0.016536344401538373, + "weighted_rkd_loss": 0.018453284725546838 + }, + { + "epoch": 2.988070992144312, + "flow_matching_loss": 0.01501545999199152, + "grad_norm": 0.1916409134864807, + "learning_rate": 2.8812198604664627e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007406242005527019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003465371276251972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018277855403721333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9295066893100739, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10270, + "weighted_flow_matching_loss": 0.01501545999199152, + "weighted_rkd_loss": 0.016989410668611527 + }, + { + "epoch": 2.9909805062554553, + "flow_matching_loss": 0.014263064321130513, + "grad_norm": 0.19674168527126312, + "learning_rate": 2.8808972209244848e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008071169350296259, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035228228429332376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01966516152024269, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9293726086616516, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10280, + "weighted_flow_matching_loss": 0.014263064321130513, + "weighted_rkd_loss": 0.01827626321464777 + }, + { + "epoch": 2.9938900203665986, + "flow_matching_loss": 0.013385856803506613, + "grad_norm": 0.3011946678161621, + "learning_rate": 2.8805741619023922e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00782667095772922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035378670319914816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019191209133714436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9292383909225463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10290, + "weighted_flow_matching_loss": 0.013385856803506613, + "weighted_rkd_loss": 0.017833223566412926 + }, + { + "epoch": 2.9967995344777423, + "flow_matching_loss": 0.014297038782387972, + "grad_norm": 0.19030797481536865, + "learning_rate": 2.880250683498321e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006854776479303837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031886443961411715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016898197308182716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.929104071855545, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10300, + "weighted_flow_matching_loss": 0.014297038782387972, + "weighted_rkd_loss": 0.015700231958180665 + }, + { + "epoch": 2.9997090485888855, + "flow_matching_loss": 0.015505137853324413, + "grad_norm": 0.20447690784931183, + "learning_rate": 2.879926785810535e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007709083566442132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036329496651887894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01905111689120531, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9289696216583252, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10310, + "weighted_flow_matching_loss": 0.015505137853324413, + "weighted_rkd_loss": 0.017697873711585998 + }, + { + "epoch": 3.0026185627000292, + "flow_matching_loss": 0.015398352593183517, + "grad_norm": 0.28162211179733276, + "learning_rate": 2.8796024689374267e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006624346226453781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030397803289815783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01628847299143672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9288350701332092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10320, + "weighted_flow_matching_loss": 0.015398352593183517, + "weighted_rkd_loss": 0.015129228588193655 + }, + { + "epoch": 3.0055280768111725, + "flow_matching_loss": 0.016033040825277567, + "grad_norm": 0.20967522263526917, + "learning_rate": 2.8792777329775148e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071121346671134235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031018798472359776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0173261490650475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9287003874778748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10330, + "weighted_flow_matching_loss": 0.016033040825277567, + "weighted_rkd_loss": 0.016090812534093855 + }, + { + "epoch": 3.0084375909223158, + "flow_matching_loss": 0.0145457923412323, + "grad_norm": 0.2533435821533203, + "learning_rate": 2.878952578029445e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074909654911607506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003359537897631526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01834146920591593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9285655856132508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10340, + "weighted_flow_matching_loss": 0.0145457923412323, + "weighted_rkd_loss": 0.0170312718488276 + }, + { + "epoch": 3.0113471050334595, + "flow_matching_loss": 0.01377388769760728, + "grad_norm": 0.19325479865074158, + "learning_rate": 2.8786270041919915e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007511448627337813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035305189434438945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018553416058421136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9284306764602661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10350, + "weighted_flow_matching_loss": 0.01377388769760728, + "weighted_rkd_loss": 0.01722552301362157 + }, + { + "epoch": 3.0142566191446027, + "flow_matching_loss": 0.015113085974007845, + "grad_norm": 0.2058713138103485, + "learning_rate": 2.8783010115640546e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007149826874956489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003321396536193788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017621050402522088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9282956480979919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10360, + "weighted_flow_matching_loss": 0.015113085974007845, + "weighted_rkd_loss": 0.016357524320483208 + }, + { + "epoch": 3.0171661332557465, + "flow_matching_loss": 0.013951197545975447, + "grad_norm": 0.2096242755651474, + "learning_rate": 2.877974600244662e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007377457851544022, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031766040017828347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01793151944875717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9281604945659637, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10370, + "weighted_flow_matching_loss": 0.013951197545975447, + "weighted_rkd_loss": 0.016643407475203275 + }, + { + "epoch": 3.0200756473668897, + "flow_matching_loss": 0.014320273045450449, + "grad_norm": 0.2292679250240326, + "learning_rate": 2.8776477703329696e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007273390749469399, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033672323683276773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017914013750851155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9280252158641815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10380, + "weighted_flow_matching_loss": 0.014320273045450449, + "weighted_rkd_loss": 0.016624672804027796 + }, + { + "epoch": 3.022985161478033, + "flow_matching_loss": 0.01480367723852396, + "grad_norm": 0.21498580276966095, + "learning_rate": 2.877320521928259e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008037821296602488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035145351430401206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019590178038924932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9278898537158966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10390, + "weighted_flow_matching_loss": 0.01480367723852396, + "weighted_rkd_loss": 0.018177489191293715 + }, + { + "epoch": 3.0258946755891767, + "flow_matching_loss": 0.015533934906125068, + "grad_norm": 0.2636665403842926, + "learning_rate": 2.87699285512994e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007540001953020692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034260064363479613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018506010342389346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9277543485164642, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10400, + "weighted_flow_matching_loss": 0.015533934906125068, + "weighted_rkd_loss": 0.017169007752090692 + }, + { + "epoch": 3.02880418970032, + "flow_matching_loss": 0.013987446390092372, + "grad_norm": 0.19263668358325958, + "learning_rate": 2.876664770037549e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007454197620972991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033139703795313836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018222365714609622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9276187241077423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10410, + "weighted_flow_matching_loss": 0.013987446390092372, + "weighted_rkd_loss": 0.016903432551771402 + }, + { + "epoch": 3.0317137038114637, + "flow_matching_loss": 0.016492162831127644, + "grad_norm": 0.20938870310783386, + "learning_rate": 2.8763362667507495e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007378212222829461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032290078699588774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017985432501882316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9274829983711242, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10420, + "weighted_flow_matching_loss": 0.016492162831127644, + "weighted_rkd_loss": 0.01668121023103595 + }, + { + "epoch": 3.034623217922607, + "flow_matching_loss": 0.016101290937513113, + "grad_norm": 0.30102410912513733, + "learning_rate": 2.8760073453693322e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007843682821840049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035782311344519256, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01926559694111347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9273471236228943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10430, + "weighted_flow_matching_loss": 0.016101290937513113, + "weighted_rkd_loss": 0.01786579778417945 + }, + { + "epoch": 3.03753273203375, + "flow_matching_loss": 0.014608075190335513, + "grad_norm": 0.1964079886674881, + "learning_rate": 2.8756780059932146e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007013282598927617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003285843227058649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017312408611178397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9272111773490905, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10440, + "weighted_flow_matching_loss": 0.014608075190335513, + "weighted_rkd_loss": 0.016052269842475653 + }, + { + "epoch": 3.040442246144894, + "flow_matching_loss": 0.01520037343725562, + "grad_norm": 0.24983634054660797, + "learning_rate": 2.875348248722442e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007525503216311335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003541279165074229, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01859228564426303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9270750880241394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10450, + "weighted_flow_matching_loss": 0.01520037343725562, + "weighted_rkd_loss": 0.01723643271252513 + }, + { + "epoch": 3.043351760256037, + "flow_matching_loss": 0.016489285975694656, + "grad_norm": 0.22866322100162506, + "learning_rate": 2.8750180736571848e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007399959024041891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032893150113523007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018089233338832854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9269388854503632, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10460, + "weighted_flow_matching_loss": 0.016489285975694656, + "weighted_rkd_loss": 0.016767595801502465 + }, + { + "epoch": 3.046261274367181, + "flow_matching_loss": 0.014275345858186484, + "grad_norm": 0.2077227681875229, + "learning_rate": 2.8746874808977424e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007099489029496908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003234222938772291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01743320133537054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9268025696277619, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10470, + "weighted_flow_matching_loss": 0.014275345858186484, + "weighted_rkd_loss": 0.016157117020338774 + }, + { + "epoch": 3.049170788478324, + "flow_matching_loss": 0.01479467675089836, + "grad_norm": 0.23673737049102783, + "learning_rate": 2.87435647054454e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00721456459723413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003371049417182803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017800178751349448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9266661107540131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10480, + "weighted_flow_matching_loss": 0.01479467675089836, + "weighted_rkd_loss": 0.01649473952129483 + }, + { + "epoch": 3.0520803025894674, + "flow_matching_loss": 0.015754256583750247, + "grad_norm": 0.23005138337612152, + "learning_rate": 2.8740250426981292e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008231741236522793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038138578645884992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020277340523898603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9265295743942261, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10490, + "weighted_flow_matching_loss": 0.015754256583750247, + "weighted_rkd_loss": 0.01878761574625969 + }, + { + "epoch": 3.054989816700611, + "flow_matching_loss": 0.013697139266878367, + "grad_norm": 0.20442211627960205, + "learning_rate": 2.8736931974591895e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00721358391456306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034180676331743596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017845235485583543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9263929009437561, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10500, + "weighted_flow_matching_loss": 0.013697139266878367, + "weighted_rkd_loss": 0.016531653329730035 + }, + { + "epoch": 3.0578993308117544, + "flow_matching_loss": 0.01459164870902896, + "grad_norm": 0.24113169312477112, + "learning_rate": 2.8733609349285268e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007495668204501271, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003430816577747464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018422153033316135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9262561142444611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10510, + "weighted_flow_matching_loss": 0.01459164870902896, + "weighted_rkd_loss": 0.017063617054373027 + }, + { + "epoch": 3.060808844922898, + "flow_matching_loss": 0.014499835763126611, + "grad_norm": 0.2168397307395935, + "learning_rate": 2.8730282552070734e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007906045764684677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003596593951806426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01940868590027094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9261192083358765, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10520, + "weighted_flow_matching_loss": 0.014499835763126611, + "weighted_rkd_loss": 0.01797476038336754 + }, + { + "epoch": 3.0637183590340413, + "flow_matching_loss": 0.014500521682202816, + "grad_norm": 0.24558590352535248, + "learning_rate": 2.872695158395889e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00782772242091596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003807175951078534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019462620839476585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9259821772575378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10530, + "weighted_flow_matching_loss": 0.014500521682202816, + "weighted_rkd_loss": 0.018022021371871234 + }, + { + "epoch": 3.0666278731451846, + "flow_matching_loss": 0.015062385238707066, + "grad_norm": 0.2242431491613388, + "learning_rate": 2.872361644596159e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007539010746404529, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034086154075339435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018486637249588968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9258450508117676, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10540, + "weighted_flow_matching_loss": 0.015062385238707066, + "weighted_rkd_loss": 0.017115754075348377 + }, + { + "epoch": 3.0695373872563283, + "flow_matching_loss": 0.014224156178534031, + "grad_norm": 0.2109292596578598, + "learning_rate": 2.8720277139091972e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069937854539603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032988560386002065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01728642676025629, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9257077932357788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10550, + "weighted_flow_matching_loss": 0.014224156178534031, + "weighted_rkd_loss": 0.016002225410193204 + }, + { + "epoch": 3.0724469013674716, + "flow_matching_loss": 0.014663287531584501, + "grad_norm": 0.2351847141981125, + "learning_rate": 2.8716933664364417e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007586460001766682, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003329475992359221, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01850239569321275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9255704164505005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10560, + "weighted_flow_matching_loss": 0.014663287531584501, + "weighted_rkd_loss": 0.017125282250344754 + }, + { + "epoch": 3.0753564154786153, + "flow_matching_loss": 0.015121576376259327, + "grad_norm": 0.18630662560462952, + "learning_rate": 2.871358602279459e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007701525837182999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031750972382724287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01857814909890294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9254329323768615, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10570, + "weighted_flow_matching_loss": 0.015121576376259327, + "weighted_rkd_loss": 0.01719287605956197 + }, + { + "epoch": 3.0782659295897585, + "flow_matching_loss": 0.014276502281427383, + "grad_norm": 0.19307461380958557, + "learning_rate": 2.871023421539942e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076565482188016174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032257744809612633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018538870848715305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9252953231334686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10580, + "weighted_flow_matching_loss": 0.014276502281427383, + "weighted_rkd_loss": 0.017153940070420502 + }, + { + "epoch": 3.081175443700902, + "flow_matching_loss": 0.015432846080511809, + "grad_norm": 0.2725880742073059, + "learning_rate": 2.8706878243197085e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008205953845754265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036224638344720004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02003437150269747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9251576066017151, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10590, + "weighted_flow_matching_loss": 0.015432846080511809, + "weighted_rkd_loss": 0.0185348610393703 + }, + { + "epoch": 3.0840849578120455, + "flow_matching_loss": 0.013594762422144413, + "grad_norm": 0.20571322739124298, + "learning_rate": 2.8703518107207052e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007634405884891749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003552570263855159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01882138177752495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9250197589397431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10600, + "weighted_flow_matching_loss": 0.013594762422144413, + "weighted_rkd_loss": 0.01741018993780017 + }, + { + "epoch": 3.0869944719231888, + "flow_matching_loss": 0.01387378079816699, + "grad_norm": 0.21344690024852753, + "learning_rate": 2.870015380845004e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007713738363236189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003660961030982435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019088437873870136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9248818099498749, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10610, + "weighted_flow_matching_loss": 0.01387378079816699, + "weighted_rkd_loss": 0.01765450770035386 + }, + { + "epoch": 3.0899039860343325, + "flow_matching_loss": 0.014736206829547882, + "grad_norm": 0.21645615994930267, + "learning_rate": 2.8696785347948035e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007847130065783858, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036916591925546527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019385919068008663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9247437357902527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10620, + "weighted_flow_matching_loss": 0.014736206829547882, + "weighted_rkd_loss": 0.017927029728889467 + }, + { + "epoch": 3.0928135001454757, + "flow_matching_loss": 0.014732387196272612, + "grad_norm": 0.21856437623500824, + "learning_rate": 2.8693412726724282e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007578343385830522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034723646473139524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018629051558673382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9246055483818054, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10630, + "weighted_flow_matching_loss": 0.014732387196272612, + "weighted_rkd_loss": 0.017224540561437608 + }, + { + "epoch": 3.095723014256619, + "flow_matching_loss": 0.01534735718742013, + "grad_norm": 0.27124693989753723, + "learning_rate": 2.869003594580329e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007988575519993902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036161680473014712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019593319017440082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.924467247724533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10640, + "weighted_flow_matching_loss": 0.01534735718742013, + "weighted_rkd_loss": 0.018113281205296518 + }, + { + "epoch": 3.0986325283677627, + "flow_matching_loss": 0.015361978765577078, + "grad_norm": 0.21949391067028046, + "learning_rate": 2.868665500621085e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007189670670777559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031867942307144403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017566135991364716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9243288278579712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10650, + "weighted_flow_matching_loss": 0.015361978765577078, + "weighted_rkd_loss": 0.01623686999082565 + }, + { + "epoch": 3.101542042478906, + "flow_matching_loss": 0.013762258645147084, + "grad_norm": 0.23600023984909058, + "learning_rate": 2.8683269908973992e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007635099859908223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003368446696549654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01863864650949836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9241902828216553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10660, + "weighted_flow_matching_loss": 0.013762258645147084, + "weighted_rkd_loss": 0.017225679475814104 + }, + { + "epoch": 3.1044515565900497, + "flow_matching_loss": 0.015606945566833019, + "grad_norm": 0.18725994229316711, + "learning_rate": 2.867988065512102e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007625917857512832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034861312713474037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018737966660410166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9240516304969788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10670, + "weighted_flow_matching_loss": 0.015606945566833019, + "weighted_rkd_loss": 0.017314942460507153 + }, + { + "epoch": 3.107361070701193, + "flow_matching_loss": 0.01514013335108757, + "grad_norm": 0.22808495163917542, + "learning_rate": 2.8676487245681507e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007534646615386009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003436575550585985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01850586859509349, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9239128530025482, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10680, + "weighted_flow_matching_loss": 0.01514013335108757, + "weighted_rkd_loss": 0.017097795289009808 + }, + { + "epoch": 3.110270584812336, + "flow_matching_loss": 0.014368578605353833, + "grad_norm": 0.18849776685237885, + "learning_rate": 2.867308968168627e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007521182717755437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003500871988944709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01854323772713542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9237739682197571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10690, + "weighted_flow_matching_loss": 0.014368578605353833, + "weighted_rkd_loss": 0.017129731830209494 + }, + { + "epoch": 3.11318009892348, + "flow_matching_loss": 0.015288838930428029, + "grad_norm": 0.2674383223056793, + "learning_rate": 2.8669687964167404e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007909531332552433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034792959224432705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019298358540982007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9236349821090698, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10700, + "weighted_flow_matching_loss": 0.015288838930428029, + "weighted_rkd_loss": 0.017824490647763015 + }, + { + "epoch": 3.116089613034623, + "flow_matching_loss": 0.015673859603703023, + "grad_norm": 0.2604334056377411, + "learning_rate": 2.8666282094158264e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008526726253330707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039191191317513585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020972571708261965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9234958589076996, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10710, + "weighted_flow_matching_loss": 0.015673859603703023, + "weighted_rkd_loss": 0.01936813574284315 + }, + { + "epoch": 3.1189991271457664, + "flow_matching_loss": 0.014391174167394638, + "grad_norm": 0.17511068284511566, + "learning_rate": 2.866287207269346e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006904358696192503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031667873030528427, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01697550481185317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9233566224575043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10720, + "weighted_flow_matching_loss": 0.014391174167394638, + "weighted_rkd_loss": 0.015674504451453686 + }, + { + "epoch": 3.12190864125691, + "flow_matching_loss": 0.014690472930669784, + "grad_norm": 0.24377933144569397, + "learning_rate": 2.8659457900808868e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00709769306704402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032022783998399973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017397664487361908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9232172727584839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10730, + "weighted_flow_matching_loss": 0.014690472930669784, + "weighted_rkd_loss": 0.01606179531663656 + }, + { + "epoch": 3.1248181553680534, + "flow_matching_loss": 0.01551890354603529, + "grad_norm": 0.20122607052326202, + "learning_rate": 2.8656039579541628e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007320046285167336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003561871126294136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018201963510364294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9230778098106385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10740, + "weighted_flow_matching_loss": 0.01551890354603529, + "weighted_rkd_loss": 0.016801900509744883 + }, + { + "epoch": 3.127727669479197, + "flow_matching_loss": 0.014478020742535591, + "grad_norm": 0.23393560945987701, + "learning_rate": 2.865261710993012e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006748444074764848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002974564163014293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016471452452242373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9229382276535034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10750, + "weighted_flow_matching_loss": 0.014478020742535591, + "weighted_rkd_loss": 0.015202120319008828 + }, + { + "epoch": 3.1306371835903404, + "flow_matching_loss": 0.014628520421683788, + "grad_norm": 0.20732903480529785, + "learning_rate": 2.864919049301402e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006511355517432094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031211723806336524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016143883392214776, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9227985203266144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10760, + "weighted_flow_matching_loss": 0.014628520421683788, + "weighted_rkd_loss": 0.014897478464990854 + }, + { + "epoch": 3.133546697701484, + "flow_matching_loss": 0.013312563579529525, + "grad_norm": 0.1816171258687973, + "learning_rate": 2.8645759729834227e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007188534317538142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031541730277240276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01753124166280031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9226587116718292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10770, + "weighted_flow_matching_loss": 0.013312563579529525, + "weighted_rkd_loss": 0.016175377089530228 + }, + { + "epoch": 3.1364562118126273, + "flow_matching_loss": 0.015263212472200393, + "grad_norm": 0.23754894733428955, + "learning_rate": 2.864232482143293e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007771107461303472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00345866521820426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019000880233943463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9225187838077545, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10780, + "weighted_flow_matching_loss": 0.015263212472200393, + "weighted_rkd_loss": 0.017528623901307584 + }, + { + "epoch": 3.1393657259237706, + "flow_matching_loss": 0.014113807957619429, + "grad_norm": 0.22441647946834564, + "learning_rate": 2.8638885768853548e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006511048553511501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003109297249466181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016131394263356923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9223787426948548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10790, + "weighted_flow_matching_loss": 0.014113807957619429, + "weighted_rkd_loss": 0.014879242330789567 + }, + { + "epoch": 3.1422752400349143, + "flow_matching_loss": 0.01429645661264658, + "grad_norm": 0.20726238191127777, + "learning_rate": 2.8635442573140793e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007823243644088507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003771366644650698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019417854119092227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9222385704517364, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10800, + "weighted_flow_matching_loss": 0.01429645661264658, + "weighted_rkd_loss": 0.017907929699867965 + }, + { + "epoch": 3.1451847541460576, + "flow_matching_loss": 0.014047944173216819, + "grad_norm": 0.24828924238681793, + "learning_rate": 2.8631995235340606e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006844477355480194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029231094289571048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016612064093351364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9220982968807221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10810, + "weighted_flow_matching_loss": 0.014047944173216819, + "weighted_rkd_loss": 0.015317987184971571 + }, + { + "epoch": 3.148094268257201, + "flow_matching_loss": 0.014920998178422452, + "grad_norm": 0.23501792550086975, + "learning_rate": 2.8628543756500192e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070590999443084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003380179964005947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017498379945755003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9219578981399537, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10820, + "weighted_flow_matching_loss": 0.014920998178422452, + "weighted_rkd_loss": 0.0161327444948256 + }, + { + "epoch": 3.1510037823683446, + "flow_matching_loss": 0.01577661940827966, + "grad_norm": 0.2140168994665146, + "learning_rate": 2.862508813766803e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069745142944157125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031781384255737065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017127167154103517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9218173921108246, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10830, + "weighted_flow_matching_loss": 0.01577661940827966, + "weighted_rkd_loss": 0.015788140427321195 + }, + { + "epoch": 3.153913296479488, + "flow_matching_loss": 0.016007569897919894, + "grad_norm": 0.22587500512599945, + "learning_rate": 2.8621628379893837e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007153784204274416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003287316346541047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01759488470852375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.921676754951477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10840, + "weighted_flow_matching_loss": 0.016007569897919894, + "weighted_rkd_loss": 0.016216711699962617 + }, + { + "epoch": 3.1568228105906315, + "flow_matching_loss": 0.014832467958331107, + "grad_norm": 0.24239353835582733, + "learning_rate": 2.861816448422861e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007235358934849501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003263173019513488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017733890656381847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9215360283851624, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10850, + "weighted_flow_matching_loss": 0.014832467958331107, + "weighted_rkd_loss": 0.016342391166836022 + }, + { + "epoch": 3.159732324701775, + "flow_matching_loss": 0.013327892124652862, + "grad_norm": 0.19770318269729614, + "learning_rate": 2.8614696451724573e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007078240206465125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032445019576698543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017400982882827522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9213951706886292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10860, + "weighted_flow_matching_loss": 0.013327892124652862, + "weighted_rkd_loss": 0.0160332215949893 + }, + { + "epoch": 3.162641838812918, + "flow_matching_loss": 0.015299403946846724, + "grad_norm": 0.21267399191856384, + "learning_rate": 2.861122428343523e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072276915423572065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003388894721865654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017844277992844583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9212541997432708, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10870, + "weighted_flow_matching_loss": 0.015299403946846724, + "weighted_rkd_loss": 0.016439193580299615 + }, + { + "epoch": 3.1655513529240618, + "flow_matching_loss": 0.014543513767421246, + "grad_norm": 0.20450879633426666, + "learning_rate": 2.8607747980415333e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007947727665305137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003449673508293927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019345128815621136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.921113109588623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10880, + "weighted_flow_matching_loss": 0.014543513767421246, + "weighted_rkd_loss": 0.017819068860262632 + }, + { + "epoch": 3.168460867035205, + "flow_matching_loss": 0.014277296606451273, + "grad_norm": 0.24238084256649017, + "learning_rate": 2.860426754372089e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007486462034285069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034429731080308557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01841589715331793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9209719002246857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10890, + "weighted_flow_matching_loss": 0.014277296606451273, + "weighted_rkd_loss": 0.016960466187447308 + }, + { + "epoch": 3.1713703811463487, + "flow_matching_loss": 0.01587151251733303, + "grad_norm": 0.1755771040916443, + "learning_rate": 2.860078297440917e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007249239599332213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034559058491140603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017954385094344615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9208305954933167, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10900, + "weighted_flow_matching_loss": 0.01587151251733303, + "weighted_rkd_loss": 0.0165329628624022 + }, + { + "epoch": 3.174279895257492, + "flow_matching_loss": 0.01365711260586977, + "grad_norm": 0.22479598224163055, + "learning_rate": 2.8597294273538687e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007140809670090675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003411000734195113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017692620120942592, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9206891596317291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10910, + "weighted_flow_matching_loss": 0.01365711260586977, + "weighted_rkd_loss": 0.01628942359238863 + }, + { + "epoch": 3.1771894093686353, + "flow_matching_loss": 0.01440425356850028, + "grad_norm": 0.18032534420490265, + "learning_rate": 2.8593801442169223e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006481497315689921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002948395162820816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015911389701068403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9205476105213165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10920, + "weighted_flow_matching_loss": 0.01440425356850028, + "weighted_rkd_loss": 0.014647206570953131 + }, + { + "epoch": 3.180098923479779, + "flow_matching_loss": 0.015980301424860954, + "grad_norm": 0.21535472571849823, + "learning_rate": 2.85903044813618e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007975256815552712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038579119136556983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0198084257543087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9204059481620789, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10930, + "weighted_flow_matching_loss": 0.015980301424860954, + "weighted_rkd_loss": 0.018231727369129657 + }, + { + "epoch": 3.1830084375909222, + "flow_matching_loss": 0.014712117239832878, + "grad_norm": 0.24294863641262054, + "learning_rate": 2.8586803392178715e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007946172077208758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003983944025821984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01987628834322095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9202641785144806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10940, + "weighted_flow_matching_loss": 0.014712117239832878, + "weighted_rkd_loss": 0.018291334528476 + }, + { + "epoch": 3.185917951702066, + "flow_matching_loss": 0.014553305227309465, + "grad_norm": 0.2981439530849457, + "learning_rate": 2.858329817568349e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007458046730607748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034734247950837017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018389518093317746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9201222777366638, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10950, + "weighted_flow_matching_loss": 0.014553305227309465, + "weighted_rkd_loss": 0.01692055268213153 + }, + { + "epoch": 3.188827465813209, + "flow_matching_loss": 0.01447321819141507, + "grad_norm": 0.21330846846103668, + "learning_rate": 2.8579788832940925e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073851412627846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032022805884480477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01797256339341402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9199802577495575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10960, + "weighted_flow_matching_loss": 0.01447321819141507, + "weighted_rkd_loss": 0.016534370277076958 + }, + { + "epoch": 3.1917369799243525, + "flow_matching_loss": 0.013860910758376121, + "grad_norm": 0.22861121594905853, + "learning_rate": 2.8576275365017073e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007849412551149727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035300146555528045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019228839688003062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9198381304740906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10970, + "weighted_flow_matching_loss": 0.013860910758376121, + "weighted_rkd_loss": 0.01768738329410553 + }, + { + "epoch": 3.194646494035496, + "flow_matching_loss": 0.014678348228335381, + "grad_norm": 0.23293431103229523, + "learning_rate": 2.857275777297922e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007115548057481646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031142819207161665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017345378268510105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.919695895910263, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10980, + "weighted_flow_matching_loss": 0.014678348228335381, + "weighted_rkd_loss": 0.015952506475150584 + }, + { + "epoch": 3.1975560081466394, + "flow_matching_loss": 0.015107856877148152, + "grad_norm": 0.25453367829322815, + "learning_rate": 2.8569236057895924e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007699569733813405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037385702133178713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019137709587812423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.919553542137146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10990, + "weighted_flow_matching_loss": 0.015107856877148152, + "weighted_rkd_loss": 0.017598134092986585 + }, + { + "epoch": 3.200465522257783, + "flow_matching_loss": 0.013176716770976782, + "grad_norm": 0.21983268857002258, + "learning_rate": 2.8565710220836994e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007754876418039203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034848386188969017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018994591478258372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9194110751152038, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11000, + "weighted_flow_matching_loss": 0.013176716770976782, + "weighted_rkd_loss": 0.017463830299675465 + }, + { + "epoch": 3.2033750363689264, + "flow_matching_loss": 0.01455358499661088, + "grad_norm": 0.181795135140419, + "learning_rate": 2.8562180262873474e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006762059452012181, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003121888544410467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016646007355302573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9192684888839722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11010, + "weighted_flow_matching_loss": 0.01455358499661088, + "weighted_rkd_loss": 0.015302130579948425 + }, + { + "epoch": 3.2062845504800697, + "flow_matching_loss": 0.014695927873253823, + "grad_norm": 0.25285786390304565, + "learning_rate": 2.8558646185077682e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007250744709745049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00315327572170645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017654765304177998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9191257953643799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11020, + "weighted_flow_matching_loss": 0.014695927873253823, + "weighted_rkd_loss": 0.016226867213845253 + }, + { + "epoch": 3.2091940645912134, + "flow_matching_loss": 0.013811147958040237, + "grad_norm": 0.1643800139427185, + "learning_rate": 2.8555107988523177e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007066005328670144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031969514908269047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017328962124884127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9189829707145691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11030, + "weighted_flow_matching_loss": 0.013811147958040237, + "weighted_rkd_loss": 0.015925039257854223 + }, + { + "epoch": 3.2121035787023566, + "flow_matching_loss": 0.014311324059963226, + "grad_norm": 0.2051883041858673, + "learning_rate": 2.8551565674284762e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007477527810260653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003433638904243708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018388694524765013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9188400387763977, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11040, + "weighted_flow_matching_loss": 0.014311324059963226, + "weighted_rkd_loss": 0.01689632870256901 + }, + { + "epoch": 3.2150130928135003, + "flow_matching_loss": 0.015723720658570528, + "grad_norm": 0.1990324705839157, + "learning_rate": 2.8548019243438503e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007866507582366466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003477438841946423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019210454262793063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9186969935894013, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11050, + "weighted_flow_matching_loss": 0.015723720658570528, + "weighted_rkd_loss": 0.017648565396666527 + }, + { + "epoch": 3.2179226069246436, + "flow_matching_loss": 0.014738620724529028, + "grad_norm": 0.22131282091140747, + "learning_rate": 2.8544468697061715e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007132582180202007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030888640787452458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017354028206318615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9185538411140441, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11060, + "weighted_flow_matching_loss": 0.014738620724529028, + "weighted_rkd_loss": 0.015940634347498418 + }, + { + "epoch": 3.220832121035787, + "flow_matching_loss": 0.014832382090389729, + "grad_norm": 0.19306470453739166, + "learning_rate": 2.854091403623296e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00767580927349627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033109941286966205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018662612605839967, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9184105634689331, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11070, + "weighted_flow_matching_loss": 0.014832382090389729, + "weighted_rkd_loss": 0.01713993549346924 + }, + { + "epoch": 3.2237416351469306, + "flow_matching_loss": 0.014282103534787894, + "grad_norm": 0.2250852882862091, + "learning_rate": 2.8537355262032047e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007086994173005223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003366783098317683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017540771421045064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9182671785354615, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11080, + "weighted_flow_matching_loss": 0.014282103534787894, + "weighted_rkd_loss": 0.016107086651027204 + }, + { + "epoch": 3.226651149258074, + "flow_matching_loss": 0.01373063987120986, + "grad_norm": 0.2216377705335617, + "learning_rate": 2.853379237554004e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007113196421414614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034305180888623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017656910978257657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9181236743927002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11090, + "weighted_flow_matching_loss": 0.01373063987120986, + "weighted_rkd_loss": 0.01621119249612093 + }, + { + "epoch": 3.2295606633692175, + "flow_matching_loss": 0.014799044746905566, + "grad_norm": 0.2153840959072113, + "learning_rate": 2.853022537783924e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006769310357049108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032422133488580586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01678083436563611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9179800450801849, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11100, + "weighted_flow_matching_loss": 0.014799044746905566, + "weighted_rkd_loss": 0.01540450369939208 + }, + { + "epoch": 3.232470177480361, + "flow_matching_loss": 0.014140425249934196, + "grad_norm": 0.2224186509847641, + "learning_rate": 2.8526654270013223e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007006816938519478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003006794326938689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017020428366959096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9178363263607026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11110, + "weighted_flow_matching_loss": 0.014140425249934196, + "weighted_rkd_loss": 0.01562198856845498 + }, + { + "epoch": 3.235379691591504, + "flow_matching_loss": 0.014503298234194517, + "grad_norm": 0.23322781920433044, + "learning_rate": 2.8523079053146785e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007707822509109974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003643424320034683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019059069454669952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9176924705505372, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11120, + "weighted_flow_matching_loss": 0.014503298234194517, + "weighted_rkd_loss": 0.017490303516387938 + }, + { + "epoch": 3.2382892057026478, + "flow_matching_loss": 0.015130548179149628, + "grad_norm": 0.2013099193572998, + "learning_rate": 2.851949972832599e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067314979620277885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003024100721813738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016487096529453994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.91754851937294, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11130, + "weighted_flow_matching_loss": 0.015130548179149628, + "weighted_rkd_loss": 0.01512766033411026 + }, + { + "epoch": 3.241198719813791, + "flow_matching_loss": 0.014689091127365828, + "grad_norm": 0.20197433233261108, + "learning_rate": 2.851591629663814e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007383730867877603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033750781556591393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018142539914697407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9174044370651245, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11140, + "weighted_flow_matching_loss": 0.014689091127365828, + "weighted_rkd_loss": 0.016643996629863978 + }, + { + "epoch": 3.2441082339249343, + "flow_matching_loss": 0.014957542810589076, + "grad_norm": 0.24778927862644196, + "learning_rate": 2.8512328759171783e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007473668362945318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003278883150778711, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018226219899952413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9172602295875549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11150, + "weighted_flow_matching_loss": 0.014957542810589076, + "weighted_rkd_loss": 0.016718193236738444 + }, + { + "epoch": 3.247017748036078, + "flow_matching_loss": 0.014844965096563101, + "grad_norm": 0.21176739037036896, + "learning_rate": 2.8508737117016722e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00704786335118115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032112653600051997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017306992225348948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9171159446239472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11160, + "weighted_flow_matching_loss": 0.014844965096563101, + "weighted_rkd_loss": 0.015872520115226506 + }, + { + "epoch": 3.2499272621472213, + "flow_matching_loss": 0.01591931050643325, + "grad_norm": 0.2321012169122696, + "learning_rate": 2.8505141371264005e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00771814682520926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003573673963546753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01900996742770076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9169715225696564, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11170, + "weighted_flow_matching_loss": 0.01591931050643325, + "weighted_rkd_loss": 0.017431524489074944 + }, + { + "epoch": 3.252836776258365, + "flow_matching_loss": 0.016601699125021695, + "grad_norm": 0.2163742333650589, + "learning_rate": 2.8501541523005918e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076833283994346855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035737565718591212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018940413277596234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9168269872665405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11180, + "weighted_flow_matching_loss": 0.016601699125021695, + "weighted_rkd_loss": 0.017365088406950235 + }, + { + "epoch": 3.2557462903695082, + "flow_matching_loss": 0.01627154117450118, + "grad_norm": 0.2334444224834442, + "learning_rate": 2.8497937573336007e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007038313755765557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032534203492105007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017330047953873873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.916682344675064, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11190, + "weighted_flow_matching_loss": 0.01627154117450118, + "weighted_rkd_loss": 0.015886124037206174 + }, + { + "epoch": 3.258655804480652, + "flow_matching_loss": 0.014873140305280686, + "grad_norm": 0.2008829414844513, + "learning_rate": 2.8494329523349053e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007178025832399726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032250528456643224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017581104580312968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9165375769138336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11200, + "weighted_flow_matching_loss": 0.014873140305280686, + "weighted_rkd_loss": 0.016113787330687045 + }, + { + "epoch": 3.261565318591795, + "flow_matching_loss": 0.013114936742931604, + "grad_norm": 0.24754105508327484, + "learning_rate": 2.8490717374141088e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007047635084018111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032646886771544813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017359958868473767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.916392695903778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11210, + "weighted_flow_matching_loss": 0.013114936742931604, + "weighted_rkd_loss": 0.015908472146838902 + }, + { + "epoch": 3.2644748327029385, + "flow_matching_loss": 0.013973320554941892, + "grad_norm": 0.19968147575855255, + "learning_rate": 2.8487101126809384e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007478607026860118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003269232390448451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018226446770131587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9162477135658265, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11220, + "weighted_flow_matching_loss": 0.013973320554941892, + "weighted_rkd_loss": 0.01669994881376624 + }, + { + "epoch": 3.267384346814082, + "flow_matching_loss": 0.014956929255276918, + "grad_norm": 0.2787853479385376, + "learning_rate": 2.8483480782452463e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007004524720832706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030506378505378963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017059687338769437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9161026060581208, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11230, + "weighted_flow_matching_loss": 0.014956929255276918, + "weighted_rkd_loss": 0.015628425870090723 + }, + { + "epoch": 3.2702938609252254, + "flow_matching_loss": 0.0144475937820971, + "grad_norm": 0.20255732536315918, + "learning_rate": 2.8479856342170097e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006823563156649471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030100114177912473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016657137870788576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9159573912620544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11240, + "weighted_flow_matching_loss": 0.0144475937820971, + "weighted_rkd_loss": 0.015257211215794087 + }, + { + "epoch": 3.2732033750363687, + "flow_matching_loss": 0.0132446164265275, + "grad_norm": 0.2442372739315033, + "learning_rate": 2.8476227807063283e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006793335219845176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031499369535595177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016736607439815997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9158120632171631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11250, + "weighted_flow_matching_loss": 0.0132446164265275, + "weighted_rkd_loss": 0.015327581856399774 + }, + { + "epoch": 3.2761128891475124, + "flow_matching_loss": 0.015024709515273571, + "grad_norm": 0.18588949739933014, + "learning_rate": 2.8472595178234284e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007721230108290911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036019520135596394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019044412206858395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9156666159629822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11260, + "weighted_flow_matching_loss": 0.015024709515273571, + "weighted_rkd_loss": 0.017438360210508107 + }, + { + "epoch": 3.2790224032586557, + "flow_matching_loss": 0.014891583938151598, + "grad_norm": 0.21649125218391418, + "learning_rate": 2.846895845678659e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068251184653490785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003141555283218622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016791792307049035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9155210494995117, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11270, + "weighted_flow_matching_loss": 0.014891583938151598, + "weighted_rkd_loss": 0.015373150166124106 + }, + { + "epoch": 3.2819319173697994, + "flow_matching_loss": 0.016185378562659024, + "grad_norm": 0.24171729385852814, + "learning_rate": 2.8465317643824945e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007759601669386029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003468309482559562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018987512867897748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9153753757476807, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11280, + "weighted_flow_matching_loss": 0.016185378562659024, + "weighted_rkd_loss": 0.017380734626203775 + }, + { + "epoch": 3.2848414314809427, + "flow_matching_loss": 0.014182719308882952, + "grad_norm": 0.21798473596572876, + "learning_rate": 2.8461672740455334e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007343442086130381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032127311220392584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017899615224450827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9152295827865601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11290, + "weighted_flow_matching_loss": 0.014182719308882952, + "weighted_rkd_loss": 0.016382271144539117 + }, + { + "epoch": 3.2877509455920864, + "flow_matching_loss": 0.01675314586609602, + "grad_norm": 0.209937185049057, + "learning_rate": 2.8458023747784976e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007919110311195254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003820570046082139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019658790715038777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9150836765766144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11300, + "weighted_flow_matching_loss": 0.01675314586609602, + "weighted_rkd_loss": 0.017989475280046463 + }, + { + "epoch": 3.2906604597032296, + "flow_matching_loss": 0.014855963550508023, + "grad_norm": 0.19910508394241333, + "learning_rate": 2.8454370666922343e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00716461893171072, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032966704457066953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017625908274203538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9149376690387726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11310, + "weighted_flow_matching_loss": 0.014855963550508023, + "weighted_rkd_loss": 0.016126531921327115 + }, + { + "epoch": 3.293569973814373, + "flow_matching_loss": 0.016349697578698398, + "grad_norm": 0.24333372712135315, + "learning_rate": 2.8450713498977145e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008199334749951959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004091602587141097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020490271970629692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9147915363311767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11320, + "weighted_flow_matching_loss": 0.016349697578698398, + "weighted_rkd_loss": 0.018744243029505014 + }, + { + "epoch": 3.2964794879255166, + "flow_matching_loss": 0.014654278475791215, + "grad_norm": 0.23125393688678741, + "learning_rate": 2.844705224506033e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008128976542502641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037799483630806207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02003790158778429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9146452963352203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11330, + "weighted_flow_matching_loss": 0.014654278475791215, + "weighted_rkd_loss": 0.018327628448605537 + }, + { + "epoch": 3.29938900203666, + "flow_matching_loss": 0.015138539485633373, + "grad_norm": 0.1784098595380783, + "learning_rate": 2.844338690628409e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007616922305896878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035824352642521264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018816279526799918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9144989311695099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11340, + "weighted_flow_matching_loss": 0.015138539485633373, + "weighted_rkd_loss": 0.01720748357474804 + }, + { + "epoch": 3.302298516147803, + "flow_matching_loss": 0.015175977256149054, + "grad_norm": 0.18347817659378052, + "learning_rate": 2.8439717483761863e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00782571816816926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003617289662361145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019268725998699665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9143524587154388, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11350, + "weighted_flow_matching_loss": 0.015175977256149054, + "weighted_rkd_loss": 0.017618561536073683 + }, + { + "epoch": 3.305208030258947, + "flow_matching_loss": 0.014961083792150021, + "grad_norm": 0.22412371635437012, + "learning_rate": 2.843604397860832e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007730190875008702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003760229889303446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01922061201184988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9142058908939361, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11360, + "weighted_flow_matching_loss": 0.014961083792150021, + "weighted_rkd_loss": 0.017571611609309912 + }, + { + "epoch": 3.30811754437009, + "flow_matching_loss": 0.014726995769888163, + "grad_norm": 0.21699151396751404, + "learning_rate": 2.8432366391939373e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007185393245890737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00333673357963562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017707519792020322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9140591859817505, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11370, + "weighted_flow_matching_loss": 0.014726995769888163, + "weighted_rkd_loss": 0.016185689810663463 + }, + { + "epoch": 3.311027058481234, + "flow_matching_loss": 0.012586216814815998, + "grad_norm": 0.2872790992259979, + "learning_rate": 2.8428684724872176e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006858671782538295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030072828521952033, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016724626626819374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9139123737812043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11380, + "weighted_flow_matching_loss": 0.012586216814815998, + "weighted_rkd_loss": 0.01528487028554082 + }, + { + "epoch": 3.313936572592377, + "flow_matching_loss": 0.012498632445931435, + "grad_norm": 0.2298833578824997, + "learning_rate": 2.8424998978525123e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007157240016385913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031562797259539364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01747075980529189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9137654423713684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11390, + "weighted_flow_matching_loss": 0.012498632445931435, + "weighted_rkd_loss": 0.015964173525571824 + }, + { + "epoch": 3.3168460867035203, + "flow_matching_loss": 0.014369339309632778, + "grad_norm": 0.2588931620121002, + "learning_rate": 2.842130915401785e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006971903005614876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00328967678360641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017233482841402293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9136184096336365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11400, + "weighted_flow_matching_loss": 0.014369339309632778, + "weighted_rkd_loss": 0.01574482098221779 + }, + { + "epoch": 3.319755600814664, + "flow_matching_loss": 0.015282498206943274, + "grad_norm": 0.20780718326568604, + "learning_rate": 2.8417615252471226e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007939930679276586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037238693563267588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01960373092442751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9134712517261505, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11410, + "weighted_flow_matching_loss": 0.015282498206943274, + "weighted_rkd_loss": 0.017907467763870956 + }, + { + "epoch": 3.3226651149258073, + "flow_matching_loss": 0.013904676586389542, + "grad_norm": 0.20044992864131927, + "learning_rate": 2.841391727500735e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007330849859863519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032364563085138798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017898155935108663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.913323986530304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11420, + "weighted_flow_matching_loss": 0.013904676586389542, + "weighted_rkd_loss": 0.016346785146743058 + }, + { + "epoch": 3.325574629036951, + "flow_matching_loss": 0.015233304910361766, + "grad_norm": 0.23984627425670624, + "learning_rate": 2.8410215222749593e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007193641271442175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003228896181099117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01761617874726653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9131766080856323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11430, + "weighted_flow_matching_loss": 0.015233304910361766, + "weighted_rkd_loss": 0.01608666470274329 + }, + { + "epoch": 3.3284841431480943, + "flow_matching_loss": 0.014369389787316323, + "grad_norm": 0.20965422689914703, + "learning_rate": 2.840650909682252e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006873784773051739, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003006524802185595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01675409423187375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9130291223526001, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11440, + "weighted_flow_matching_loss": 0.014369389787316323, + "weighted_rkd_loss": 0.01529697198420763 + }, + { + "epoch": 3.3313936572592375, + "flow_matching_loss": 0.014794006291776896, + "grad_norm": 0.22008852660655975, + "learning_rate": 2.8402798898351963e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007646576734259725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032825154718011617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018575668893754484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9128815114498139, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11450, + "weighted_flow_matching_loss": 0.014794006291776896, + "weighted_rkd_loss": 0.016957391053438187 + }, + { + "epoch": 3.3343031713703812, + "flow_matching_loss": 0.014181283023208379, + "grad_norm": 0.2094646692276001, + "learning_rate": 2.8399084628464986e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006128412717953324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027691359631717206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015025961399078368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.912733793258667, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11460, + "weighted_flow_matching_loss": 0.014181283023208379, + "weighted_rkd_loss": 0.013714720401912928 + }, + { + "epoch": 3.3372126854815245, + "flow_matching_loss": 0.014129619020968676, + "grad_norm": 0.1924479454755783, + "learning_rate": 2.8395366288289876e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006538392836228013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003112519741989672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016189305298030376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9125859618186951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11470, + "weighted_flow_matching_loss": 0.014129619020968676, + "weighted_rkd_loss": 0.014774167072027921 + }, + { + "epoch": 3.340122199592668, + "flow_matching_loss": 0.014559551700949669, + "grad_norm": 0.20172299444675446, + "learning_rate": 2.839164387895617e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007512368867173791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003261864045634866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01828660173341632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.912438017129898, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11480, + "weighted_flow_matching_loss": 0.014559551700949669, + "weighted_rkd_loss": 0.016685414034873246 + }, + { + "epoch": 3.3430317137038115, + "flow_matching_loss": 0.014820733293890952, + "grad_norm": 0.20316044986248016, + "learning_rate": 2.8387917401594644e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068106955848634245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00301285102032125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016634242329746484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9122899651527405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11490, + "weighted_flow_matching_loss": 0.014820733293890952, + "weighted_rkd_loss": 0.015175240952521563 + }, + { + "epoch": 3.3459412278149547, + "flow_matching_loss": 0.013821905106306076, + "grad_norm": 0.22447369992733002, + "learning_rate": 2.8384186857337296e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006587216397747397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030634485883638263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016237881407141684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9121417939662934, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11500, + "weighted_flow_matching_loss": 0.013821905106306076, + "weighted_rkd_loss": 0.014811246655881405 + }, + { + "epoch": 3.3488507419260984, + "flow_matching_loss": 0.014600786101073026, + "grad_norm": 0.21639297902584076, + "learning_rate": 2.838045224731737e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008252041041851043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003376452042721212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01988053396344185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9119935095310211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11510, + "weighted_flow_matching_loss": 0.014600786101073026, + "weighted_rkd_loss": 0.018130959756672383 + }, + { + "epoch": 3.3517602560372417, + "flow_matching_loss": 0.01606525545939803, + "grad_norm": 0.255057156085968, + "learning_rate": 2.8376713572669348e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007218185439705848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032839067978784444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017720278073102235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9118451118469239, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11520, + "weighted_flow_matching_loss": 0.01606525545939803, + "weighted_rkd_loss": 0.01615817742422223 + }, + { + "epoch": 3.3546697701483854, + "flow_matching_loss": 0.015275660622864961, + "grad_norm": 0.2283281683921814, + "learning_rate": 2.8372970834528924e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006299031618982554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027453284012153743, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015343391615897418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9116965949535369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11530, + "weighted_flow_matching_loss": 0.015275660622864961, + "weighted_rkd_loss": 0.013988515734672547 + }, + { + "epoch": 3.3575792842595287, + "flow_matching_loss": 0.014472455251961946, + "grad_norm": 0.19807696342468262, + "learning_rate": 2.836922403403306e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008439646195620298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004067130666226148, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02094642324373126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9115479826927185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11540, + "weighted_flow_matching_loss": 0.014472455251961946, + "weighted_rkd_loss": 0.01909377733245492 + }, + { + "epoch": 3.360488798370672, + "flow_matching_loss": 0.015976568032056094, + "grad_norm": 0.19813665747642517, + "learning_rate": 2.8365473172319926e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077948813326656815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003355854074470699, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01894561666995287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9113992393016815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11550, + "weighted_flow_matching_loss": 0.015976568032056094, + "weighted_rkd_loss": 0.017266992572695018 + }, + { + "epoch": 3.3633983124818156, + "flow_matching_loss": 0.01484482865780592, + "grad_norm": 0.21508637070655823, + "learning_rate": 2.8361718250528936e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007398199895396829, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032811528537422417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018077552691102027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9112503886222839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11560, + "weighted_flow_matching_loss": 0.01484482865780592, + "weighted_rkd_loss": 0.016473203804343937 + }, + { + "epoch": 3.366307826592959, + "flow_matching_loss": 0.013994233589619398, + "grad_norm": 0.2091463804244995, + "learning_rate": 2.835795926980074e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007015206757932902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032176000298932196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01724801380187273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9111014425754547, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11570, + "weighted_flow_matching_loss": 0.013994233589619398, + "weighted_rkd_loss": 0.015714657213538884 + }, + { + "epoch": 3.369217340704102, + "flow_matching_loss": 0.01376188062131405, + "grad_norm": 0.18776246905326843, + "learning_rate": 2.8354196231277213e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007376802293583751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033615913474932314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01811519591137767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.910952353477478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11580, + "weighted_flow_matching_loss": 0.01376188062131405, + "weighted_rkd_loss": 0.016502087656408548 + }, + { + "epoch": 3.372126854815246, + "flow_matching_loss": 0.01514534056186676, + "grad_norm": 0.21518339216709137, + "learning_rate": 2.835042913610147e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00692808236926794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031609042314812543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017017069179564713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9108031749725342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11590, + "weighted_flow_matching_loss": 0.01514534056186676, + "weighted_rkd_loss": 0.015499207563698292 + }, + { + "epoch": 3.375036368926389, + "flow_matching_loss": 0.014641114510595798, + "grad_norm": 0.21176953613758087, + "learning_rate": 2.8346657985417846e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006802086113020777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003060495713725686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01666466798633337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9106538772583008, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11600, + "weighted_flow_matching_loss": 0.014641114510595798, + "weighted_rkd_loss": 0.015175723284482957 + }, + { + "epoch": 3.377945883037533, + "flow_matching_loss": 0.013806559424847365, + "grad_norm": 0.18471543490886688, + "learning_rate": 2.8342882780371932e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006500656856223941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002909737918525934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015911051630973817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9105044543743134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11610, + "weighted_flow_matching_loss": 0.013806559424847365, + "weighted_rkd_loss": 0.014487114548683167 + }, + { + "epoch": 3.380855397148676, + "flow_matching_loss": 0.015115432906895876, + "grad_norm": 0.18986740708351135, + "learning_rate": 2.833910352211053e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006944243842735886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003159751824568957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017048239707946777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9103549420833588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11620, + "weighted_flow_matching_loss": 0.015115432906895876, + "weighted_rkd_loss": 0.015519897174090146 + }, + { + "epoch": 3.38376491125982, + "flow_matching_loss": 0.014427615702152253, + "grad_norm": 0.21232405304908752, + "learning_rate": 2.8335320211781667e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007457575388252735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032749440986663105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018190094828605653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9102053046226501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11630, + "weighted_flow_matching_loss": 0.014427615702152253, + "weighted_rkd_loss": 0.01655673049390316 + }, + { + "epoch": 3.386674425370963, + "flow_matching_loss": 0.015923662483692168, + "grad_norm": 0.192884162068367, + "learning_rate": 2.8331532850534628e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006707805953919887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030601060716435313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01647571809589863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.910055547952652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11640, + "weighted_flow_matching_loss": 0.015923662483692168, + "weighted_rkd_loss": 0.01499382285401225 + }, + { + "epoch": 3.3895839394821063, + "flow_matching_loss": 0.014695008005946874, + "grad_norm": 0.2211928814649582, + "learning_rate": 2.8327741439519906e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007482485985383392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003237072681076825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0182020447216928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9099056959152222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11650, + "weighted_flow_matching_loss": 0.014695008005946874, + "weighted_rkd_loss": 0.01656217211857438 + }, + { + "epoch": 3.39249345359325, + "flow_matching_loss": 0.01567819844931364, + "grad_norm": 0.2099854201078415, + "learning_rate": 2.8323945979889233e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006681238580495119, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003091234969906509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01645371187478304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9097557127475738, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11660, + "weighted_flow_matching_loss": 0.01567819844931364, + "weighted_rkd_loss": 0.014968888182193042 + }, + { + "epoch": 3.3954029677043933, + "flow_matching_loss": 0.014151744078844786, + "grad_norm": 0.18237780034542084, + "learning_rate": 2.8320146472795572e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006901645939797163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030049720080569386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016808263957500458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9096056222915649, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11670, + "weighted_flow_matching_loss": 0.014151744078844786, + "weighted_rkd_loss": 0.015288935974240303 + }, + { + "epoch": 3.3983124818155366, + "flow_matching_loss": 0.015899523813277482, + "grad_norm": 0.20908842980861664, + "learning_rate": 2.831634291939311e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007148368749767542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031340542249381543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017430791817605495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9094554245471954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11680, + "weighted_flow_matching_loss": 0.015899523813277482, + "weighted_rkd_loss": 0.01585247563198209 + }, + { + "epoch": 3.4012219959266803, + "flow_matching_loss": 0.01458446504548192, + "grad_norm": 0.22642996907234192, + "learning_rate": 2.8312535320837266e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006925668800249696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003026202181354165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016877539828419684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9093051195144654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11690, + "weighted_flow_matching_loss": 0.01458446504548192, + "weighted_rkd_loss": 0.01534679690375924 + }, + { + "epoch": 3.4041315100378235, + "flow_matching_loss": 0.016057569161057474, + "grad_norm": 0.1975649893283844, + "learning_rate": 2.8308723678284687e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006683483486995101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029992274707183243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016366194654256105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9091547071933747, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11700, + "weighted_flow_matching_loss": 0.016057569161057474, + "weighted_rkd_loss": 0.014879356138408184 + }, + { + "epoch": 3.4070410241489673, + "flow_matching_loss": 0.014229085203260184, + "grad_norm": 0.19783048331737518, + "learning_rate": 2.8304907992893254e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006702661933377385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030048614600673316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01641018530353904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9090041637420654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11710, + "weighted_flow_matching_loss": 0.014229085203260184, + "weighted_rkd_loss": 0.014916923362761736 + }, + { + "epoch": 3.4099505382601105, + "flow_matching_loss": 0.015321763791143894, + "grad_norm": 0.16023887693881989, + "learning_rate": 2.8301088265822066e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006578403571620583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027331790653988717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015889986138790845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9088535189628602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11720, + "weighted_flow_matching_loss": 0.015321763791143894, + "weighted_rkd_loss": 0.014441667869687081 + }, + { + "epoch": 3.4128600523712542, + "flow_matching_loss": 0.014808115269988775, + "grad_norm": 0.1884067952632904, + "learning_rate": 2.829726449823146e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007374860905110836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033111185999587177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01806084057316184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9087027609348297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11730, + "weighted_flow_matching_loss": 0.014808115269988775, + "weighted_rkd_loss": 0.016411928366869687 + }, + { + "epoch": 3.4157695664823975, + "flow_matching_loss": 0.016439976636320353, + "grad_norm": 0.19896820187568665, + "learning_rate": 2.8293436691282993e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007073874585330486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003036362770944834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017184111941605806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9085518836975097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11740, + "weighted_flow_matching_loss": 0.016439976636320353, + "weighted_rkd_loss": 0.015612683910876512 + }, + { + "epoch": 3.4186790805935408, + "flow_matching_loss": 0.0147885175421834, + "grad_norm": 0.24769191443920135, + "learning_rate": 2.828960484613945e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007480741711333394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003141885413788259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018103368673473598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9084009051322937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11750, + "weighted_flow_matching_loss": 0.0147885175421834, + "weighted_rkd_loss": 0.016445174533873796 + }, + { + "epoch": 3.4215885947046845, + "flow_matching_loss": 0.014168235659599304, + "grad_norm": 0.1818542629480362, + "learning_rate": 2.828576896396485e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007128208782523871, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002973881363868713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01723029902204871, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.908249807357788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11760, + "weighted_flow_matching_loss": 0.014168235659599304, + "weighted_rkd_loss": 0.0156494552269578 + }, + { + "epoch": 3.4244981088158277, + "flow_matching_loss": 0.014901799894869328, + "grad_norm": 0.23325631022453308, + "learning_rate": 2.828192904592443e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006620688689872622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026993574458174406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015940734837204217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9080986022949219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11770, + "weighted_flow_matching_loss": 0.014901799894869328, + "weighted_rkd_loss": 0.014475719444453717 + }, + { + "epoch": 3.427407622926971, + "flow_matching_loss": 0.014746754243969918, + "grad_norm": 0.226798415184021, + "learning_rate": 2.8278085093184648e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006568656396120786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030895381700247527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016226850915700197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9079472899436951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11780, + "weighted_flow_matching_loss": 0.014746754243969918, + "weighted_rkd_loss": 0.014733080100268126 + }, + { + "epoch": 3.4303171370381147, + "flow_matching_loss": 0.014838626887649298, + "grad_norm": 0.22057566046714783, + "learning_rate": 2.827423710691321e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006682480126619339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002965814131312072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016330774314701556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9077958464622498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11790, + "weighted_flow_matching_loss": 0.014838626887649298, + "weighted_rkd_loss": 0.014824943523854018 + }, + { + "epoch": 3.433226651149258, + "flow_matching_loss": 0.017320768814533948, + "grad_norm": 0.1829206794500351, + "learning_rate": 2.827038508827902e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006483968533575535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028750701691024005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015843007434159517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9076443016529083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11800, + "weighted_flow_matching_loss": 0.017320768814533948, + "weighted_rkd_loss": 0.014379830565303564 + }, + { + "epoch": 3.4361361652604017, + "flow_matching_loss": 0.013047692086547612, + "grad_norm": 0.2116675227880478, + "learning_rate": 2.8266529038452223e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006733526987954974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003155559953302145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016622613929212094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9074926614761353, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11810, + "weighted_flow_matching_loss": 0.013047692086547612, + "weighted_rkd_loss": 0.015084897261112928 + }, + { + "epoch": 3.439045679371545, + "flow_matching_loss": 0.013799869362264872, + "grad_norm": 0.19899028539657593, + "learning_rate": 2.826266895860419e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072912875097244974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034239916363731026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018006566632539033, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9073408782482147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11820, + "weighted_flow_matching_loss": 0.013799869362264872, + "weighted_rkd_loss": 0.016338160261511803 + }, + { + "epoch": 3.4419551934826886, + "flow_matching_loss": 0.014813972916454077, + "grad_norm": 0.22509130835533142, + "learning_rate": 2.8258804849907508e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007001987798139453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003073478234000504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017077453993260862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9071890115737915, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11830, + "weighted_flow_matching_loss": 0.014813972916454077, + "weighted_rkd_loss": 0.015492430422455072 + }, + { + "epoch": 3.444864707593832, + "flow_matching_loss": 0.015404067374765873, + "grad_norm": 0.25610461831092834, + "learning_rate": 2.825493671353599e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007369403215125203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034279270097613336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01816673344001174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9070370256900787, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11840, + "weighted_flow_matching_loss": 0.015404067374765873, + "weighted_rkd_loss": 0.016477892082184552 + }, + { + "epoch": 3.447774221704975, + "flow_matching_loss": 0.014832794480025769, + "grad_norm": 0.2454914003610611, + "learning_rate": 2.8251064550664673e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006913491198793054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033602578565478324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017187240161001684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9068849205970764, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11850, + "weighted_flow_matching_loss": 0.014832794480025769, + "weighted_rkd_loss": 0.015586839895695448 + }, + { + "epoch": 3.450683735816119, + "flow_matching_loss": 0.014748219959437847, + "grad_norm": 0.22769707441329956, + "learning_rate": 2.8247188362469823e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006661989446729422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003113043564371765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016437022387981413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9067327082157135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11860, + "weighted_flow_matching_loss": 0.014748219959437847, + "weighted_rkd_loss": 0.01490396549925208 + }, + { + "epoch": 3.453593249927262, + "flow_matching_loss": 0.013640550058335065, + "grad_norm": 0.2019193023443222, + "learning_rate": 2.824330815012892e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006891410471871496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030439404887147247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016826761420816183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9065803766250611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11870, + "weighted_flow_matching_loss": 0.013640550058335065, + "weighted_rkd_loss": 0.015254784747958184 + }, + { + "epoch": 3.4565027640384054, + "flow_matching_loss": 0.01568956272676587, + "grad_norm": 0.1840122491121292, + "learning_rate": 2.8239423914820668e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006966063287109137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00322456865105778, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017156695201992988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9064279496669769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11880, + "weighted_flow_matching_loss": 0.01568956272676587, + "weighted_rkd_loss": 0.015551368147134781 + }, + { + "epoch": 3.459412278149549, + "flow_matching_loss": 0.014262735191732646, + "grad_norm": 0.19908641278743744, + "learning_rate": 2.8235535657724997e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007361154491081834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003208625945262611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017930934950709344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9062753915786743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11890, + "weighted_flow_matching_loss": 0.014262735191732646, + "weighted_rkd_loss": 0.016250433679670096 + }, + { + "epoch": 3.4623217922606924, + "flow_matching_loss": 0.01516825882717967, + "grad_norm": 0.2445497065782547, + "learning_rate": 2.823164338002306e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006688764458522201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031098699662834408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016487398836761712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9061227381229401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11900, + "weighted_flow_matching_loss": 0.01516825882717967, + "weighted_rkd_loss": 0.014939624257385731 + }, + { + "epoch": 3.465231306371836, + "flow_matching_loss": 0.013699984364211559, + "grad_norm": 0.21274757385253906, + "learning_rate": 2.8227747082897223e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007414471451193094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033892975887283685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01821824088692665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9059699714183808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11910, + "weighted_flow_matching_loss": 0.013699984364211559, + "weighted_rkd_loss": 0.016505152545869352 + }, + { + "epoch": 3.4681408204829793, + "flow_matching_loss": 0.014697437454015017, + "grad_norm": 0.24067261815071106, + "learning_rate": 2.8223846767531084e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073327166493982075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003327154275029898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0179925873875618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9058170795440674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11920, + "weighted_flow_matching_loss": 0.014697437454015017, + "weighted_rkd_loss": 0.016297929268330336 + }, + { + "epoch": 3.4710503345941226, + "flow_matching_loss": 0.013316183071583509, + "grad_norm": 0.18769174814224243, + "learning_rate": 2.821994243510945e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006288377614691853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003026566153857857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015603321418166161, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9056640982627868, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11930, + "weighted_flow_matching_loss": 0.013316183071583509, + "weighted_rkd_loss": 0.0141313336789608 + }, + { + "epoch": 3.4739598487052663, + "flow_matching_loss": 0.015210018493235111, + "grad_norm": 0.17734865844249725, + "learning_rate": 2.821603408681835e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006801317492499947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003013559151440859, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01661619422957301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9055109918117523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11940, + "weighted_flow_matching_loss": 0.015210018493235111, + "weighted_rkd_loss": 0.015046171192079783 + }, + { + "epoch": 3.4768693628164096, + "flow_matching_loss": 0.01801432967185974, + "grad_norm": 0.22731228172779083, + "learning_rate": 2.8212121723845054e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006394594768062234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002767501852940768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015556691400706767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9053577780723572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11950, + "weighted_flow_matching_loss": 0.01801432967185974, + "weighted_rkd_loss": 0.014084387384355069 + }, + { + "epoch": 3.4797788769275533, + "flow_matching_loss": 0.013649528101086616, + "grad_norm": 0.19029773771762848, + "learning_rate": 2.820820534737802e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067384310998022555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032315373653545977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016708399914205075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9052044451236725, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11960, + "weighted_flow_matching_loss": 0.013649528101086616, + "weighted_rkd_loss": 0.015124547481536865 + }, + { + "epoch": 3.4826883910386965, + "flow_matching_loss": 0.01493634758517146, + "grad_norm": 0.20234660804271698, + "learning_rate": 2.8204284958606946e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067220604047179226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032723123207688332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016716432850807907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9050510168075562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11970, + "weighted_flow_matching_loss": 0.01493634758517146, + "weighted_rkd_loss": 0.015129292476922273 + }, + { + "epoch": 3.48559790514984, + "flow_matching_loss": 0.013633273355662823, + "grad_norm": 0.20619802176952362, + "learning_rate": 2.8200360558722738e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007065441878512502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003160408046096563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01729129170998931, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9048974633216857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11980, + "weighted_flow_matching_loss": 0.013633273355662823, + "weighted_rkd_loss": 0.01564690815284848 + }, + { + "epoch": 3.4885074192609835, + "flow_matching_loss": 0.013561874255537986, + "grad_norm": 0.1972305178642273, + "learning_rate": 2.819643214891753e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006469979416579008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032912304974161088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01623118920251727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9047438085079194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11990, + "weighted_flow_matching_loss": 0.013561874255537986, + "weighted_rkd_loss": 0.014685095846652984 + }, + { + "epoch": 3.4914169333721268, + "flow_matching_loss": 0.014256588462740184, + "grad_norm": 0.1628638356924057, + "learning_rate": 2.8192499730384668e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007667562365531922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035037933615967632, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018838918209075926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9045900404453278, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12000, + "weighted_flow_matching_loss": 0.014256588462740184, + "weighted_rkd_loss": 0.017041567992419005 + }, + { + "epoch": 3.4943264474832705, + "flow_matching_loss": 0.013587688468396664, + "grad_norm": 0.2228589504957199, + "learning_rate": 2.818856330431871e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00631543742492795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028298501623794437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015460725035518407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9044361531734466, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12010, + "weighted_flow_matching_loss": 0.013587688468396664, + "weighted_rkd_loss": 0.013983261585235596 + }, + { + "epoch": 3.4972359615944137, + "flow_matching_loss": 0.015293048322200775, + "grad_norm": 0.218210831284523, + "learning_rate": 2.8184622871915446e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073454650584608315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035740493331104517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018264979403465987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9042821645736694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12020, + "weighted_flow_matching_loss": 0.015293048322200775, + "weighted_rkd_loss": 0.01651663091033697 + }, + { + "epoch": 3.500145475705557, + "flow_matching_loss": 0.013618736248463392, + "grad_norm": 0.21861743927001953, + "learning_rate": 2.8180678434371874e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006307043228298425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027485078200697897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015362594462931155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9041280567646026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12030, + "weighted_flow_matching_loss": 0.013618736248463392, + "weighted_rkd_loss": 0.013889779802411795 + }, + { + "epoch": 3.5030549898167007, + "flow_matching_loss": 0.013827250339090824, + "grad_norm": 0.19503511488437653, + "learning_rate": 2.817672999288621e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007019675523042679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003143547032959759, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017182897962629794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9039738476276398, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12040, + "weighted_flow_matching_loss": 0.013827250339090824, + "weighted_rkd_loss": 0.015532913338392972 + }, + { + "epoch": 3.505964503927844, + "flow_matching_loss": 0.014535707887262106, + "grad_norm": 0.20507925748825073, + "learning_rate": 2.8172777548657886e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006648538121953607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030040632002055643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01630113907158375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9038195312023163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12050, + "weighted_flow_matching_loss": 0.014535707887262106, + "weighted_rkd_loss": 0.014733268599957227 + }, + { + "epoch": 3.5088740180389877, + "flow_matching_loss": 0.014675639104098082, + "grad_norm": 0.1641472429037094, + "learning_rate": 2.8168821102887545e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007057675626128912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031681409338489173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017283492349088193, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9036650836467743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12060, + "weighted_flow_matching_loss": 0.014675639104098082, + "weighted_rkd_loss": 0.015618561208248139 + }, + { + "epoch": 3.511783532150131, + "flow_matching_loss": 0.014833687618374825, + "grad_norm": 0.2365863174200058, + "learning_rate": 2.816486065677706e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006827495619654655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00299566553439945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016650656796991826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9035105407238007, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12070, + "weighted_flow_matching_loss": 0.014833687618374825, + "weighted_rkd_loss": 0.015043985191732645 + }, + { + "epoch": 3.514693046261274, + "flow_matching_loss": 0.01482080789282918, + "grad_norm": 0.17175057530403137, + "learning_rate": 2.8160896211529506e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006812996743246913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003145624394528568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016771617997437715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9033558905124665, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12080, + "weighted_flow_matching_loss": 0.01482080789282918, + "weighted_rkd_loss": 0.015150758065283298 + }, + { + "epoch": 3.517602560372418, + "flow_matching_loss": 0.012808068841695785, + "grad_norm": 0.25120651721954346, + "learning_rate": 2.8156927768349175e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006711187632754445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003210668684914708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016633043996989727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9032011151313781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12090, + "weighted_flow_matching_loss": 0.012808068841695785, + "weighted_rkd_loss": 0.015022920351475478 + }, + { + "epoch": 3.520512074483561, + "flow_matching_loss": 0.014113605860620737, + "grad_norm": 0.18455344438552856, + "learning_rate": 2.815295532844158e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007013179222121834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030272090109065173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01705356752499938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9030462443828583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12100, + "weighted_flow_matching_loss": 0.014113605860620737, + "weighted_rkd_loss": 0.015400142967700958 + }, + { + "epoch": 3.5234215885947044, + "flow_matching_loss": 0.01485270569100976, + "grad_norm": 0.2124325931072235, + "learning_rate": 2.8148978893013434e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074305588379502295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034024188993498684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018263536784797908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9028912544250488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12110, + "weighted_flow_matching_loss": 0.01485270569100976, + "weighted_rkd_loss": 0.016490005422383547 + }, + { + "epoch": 3.526331102705848, + "flow_matching_loss": 0.016440146788954734, + "grad_norm": 0.22929124534130096, + "learning_rate": 2.814499846327268e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007024757564067841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003083528159186244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017133043333888053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9027361512184143, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12120, + "weighted_flow_matching_loss": 0.016440146788954734, + "weighted_rkd_loss": 0.015466661285609007 + }, + { + "epoch": 3.5292406168169914, + "flow_matching_loss": 0.01472059590741992, + "grad_norm": 0.19718916714191437, + "learning_rate": 2.8141014040428468e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007210289593786001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003278453787788749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017699032835662366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9025809466838837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12130, + "weighted_flow_matching_loss": 0.01472059590741992, + "weighted_rkd_loss": 0.015974836237728596 + }, + { + "epoch": 3.532150130928135, + "flow_matching_loss": 0.015612734574824571, + "grad_norm": 0.19289687275886536, + "learning_rate": 2.8137025625691163e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006762317894026637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002989057241939008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016513693146407604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9024256289005279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12140, + "weighted_flow_matching_loss": 0.015612734574824571, + "weighted_rkd_loss": 0.014902372751384974 + }, + { + "epoch": 3.5350596450392784, + "flow_matching_loss": 0.014806746412068606, + "grad_norm": 0.26146239042282104, + "learning_rate": 2.813303322027233e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006485996907576919, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030483007663860916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01602029465138912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9022701978683472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12150, + "weighted_flow_matching_loss": 0.014806746412068606, + "weighted_rkd_loss": 0.014454609435051679 + }, + { + "epoch": 3.537969159150422, + "flow_matching_loss": 0.013253181800246238, + "grad_norm": 0.18454617261886597, + "learning_rate": 2.8129036825384763e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077332403045147656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033781226724386216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018844603467732667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9021146595478058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12160, + "weighted_flow_matching_loss": 0.013253181800246238, + "weighted_rkd_loss": 0.017000093683600427 + }, + { + "epoch": 3.5408786732615654, + "flow_matching_loss": 0.016760855820029973, + "grad_norm": 0.18659597635269165, + "learning_rate": 2.8125036442242466e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006793137965723872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031711897463537753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016757465526461603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9019590079784393, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12170, + "weighted_flow_matching_loss": 0.016760855820029973, + "weighted_rkd_loss": 0.015114487707614898 + }, + { + "epoch": 3.5437881873727086, + "flow_matching_loss": 0.01402095789089799, + "grad_norm": 0.2505309283733368, + "learning_rate": 2.8121032072060635e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007083233445882797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033983264118433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017564793583005667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9018032431602478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12180, + "weighted_flow_matching_loss": 0.01402095789089799, + "weighted_rkd_loss": 0.01584000950679183 + }, + { + "epoch": 3.5466977014838523, + "flow_matching_loss": 0.01638306165114045, + "grad_norm": 0.22212451696395874, + "learning_rate": 2.811702371605571e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007046270556747914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003009939577896148, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017102480679750443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9016473710536956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12190, + "weighted_flow_matching_loss": 0.01638306165114045, + "weighted_rkd_loss": 0.015420423913747072 + }, + { + "epoch": 3.5496072155949956, + "flow_matching_loss": 0.0132806527428329, + "grad_norm": 0.22586815059185028, + "learning_rate": 2.8113011375445308e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006608394533395767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027869433630257844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016003732476383446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9014913976192475, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12200, + "weighted_flow_matching_loss": 0.0132806527428329, + "weighted_rkd_loss": 0.014427193999290466 + }, + { + "epoch": 3.552516729706139, + "flow_matching_loss": 0.014017003029584885, + "grad_norm": 0.21318897604942322, + "learning_rate": 2.8108995051448284e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007235418260097504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032495972234755756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017720433603972196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9013352990150452, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12210, + "weighted_flow_matching_loss": 0.014017003029584885, + "weighted_rkd_loss": 0.015972000919282435 + }, + { + "epoch": 3.5554262438172826, + "flow_matching_loss": 0.015367622766643763, + "grad_norm": 0.18007825314998627, + "learning_rate": 2.8104974745284685e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006268941564485431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028209835523739456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015358866751194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9011791050434113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12220, + "weighted_flow_matching_loss": 0.015367622766643763, + "weighted_rkd_loss": 0.01384108243510127 + }, + { + "epoch": 3.558335757928426, + "flow_matching_loss": 0.014659397304058075, + "grad_norm": 0.22700224816799164, + "learning_rate": 2.8100950458175775e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006422669440507889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027673472883179783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015612686332315207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9010227918624878, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12230, + "weighted_flow_matching_loss": 0.014659397304058075, + "weighted_rkd_loss": 0.014067375101149082 + }, + { + "epoch": 3.5612452720395695, + "flow_matching_loss": 0.013622224051505327, + "grad_norm": 0.20712018013000488, + "learning_rate": 2.8096922191344027e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063879558816552166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030330749694257976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0158089866861701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9008663773536683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12240, + "weighted_flow_matching_loss": 0.013622224051505327, + "weighted_rkd_loss": 0.01424171170219779 + }, + { + "epoch": 3.564154786150713, + "flow_matching_loss": 0.014430945087224246, + "grad_norm": 0.18534380197525024, + "learning_rate": 2.809288994601312e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005972374882549048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028279769932851194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014772726874798536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9007098436355591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12250, + "weighted_flow_matching_loss": 0.014430945087224246, + "weighted_rkd_loss": 0.013305956125259399 + }, + { + "epoch": 3.5670643002618565, + "flow_matching_loss": 0.014553217682987452, + "grad_norm": 0.21733132004737854, + "learning_rate": 2.8088853723407946e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007193062687292695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003266480239108205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017652605846524238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9005531907081604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12260, + "weighted_flow_matching_loss": 0.014553217682987452, + "weighted_rkd_loss": 0.015897065680474044 + }, + { + "epoch": 3.5699738143729998, + "flow_matching_loss": 0.013899161946028471, + "grad_norm": 0.20799283683300018, + "learning_rate": 2.8084813524754602e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006175752123817801, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002684585237875581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015036089438945055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9003964364528656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12270, + "weighted_flow_matching_loss": 0.013899161946028471, + "weighted_rkd_loss": 0.013538415916264058 + }, + { + "epoch": 3.572883328484143, + "flow_matching_loss": 0.014735730458050966, + "grad_norm": 0.21063122153282166, + "learning_rate": 2.8080769351280395e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006754038436338305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030492352903820573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016557312104851007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9002395927906036, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12280, + "weighted_flow_matching_loss": 0.014735730458050966, + "weighted_rkd_loss": 0.014905439782887697 + }, + { + "epoch": 3.5757928425952867, + "flow_matching_loss": 0.013099307287484407, + "grad_norm": 0.20999449491500854, + "learning_rate": 2.8076721204213835e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007287913653999567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031891952501609923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01776502262800932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9000826179981232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12290, + "weighted_flow_matching_loss": 0.013099307287484407, + "weighted_rkd_loss": 0.015989917609840632 + }, + { + "epoch": 3.57870235670643, + "flow_matching_loss": 0.014809746854007244, + "grad_norm": 0.16632147133350372, + "learning_rate": 2.8072669084784642e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006821993365883827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030940297292545438, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016738016437739135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8999255359172821, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12300, + "weighted_flow_matching_loss": 0.014809746854007244, + "weighted_rkd_loss": 0.015063010808080434 + }, + { + "epoch": 3.5816118708175733, + "flow_matching_loss": 0.013076392468065023, + "grad_norm": 0.2369810938835144, + "learning_rate": 2.8068612994223747e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006691428180783987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00294283851981163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016325694974511862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8997683584690094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12310, + "weighted_flow_matching_loss": 0.013076392468065023, + "weighted_rkd_loss": 0.014689378067851066 + }, + { + "epoch": 3.584521384928717, + "flow_matching_loss": 0.013846550323069095, + "grad_norm": 0.24639762938022614, + "learning_rate": 2.8064552933763284e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007122593792155385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003055213508196175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01730040116235614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8996110558509827, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12320, + "weighted_flow_matching_loss": 0.013846550323069095, + "weighted_rkd_loss": 0.015563566610217095 + }, + { + "epoch": 3.5874308990398602, + "flow_matching_loss": 0.013024392537772655, + "grad_norm": 0.16568289697170258, + "learning_rate": 2.8060488904636586e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006296239560469985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002935385098680854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01552786398679018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8994536459445953, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12330, + "weighted_flow_matching_loss": 0.013024392537772655, + "weighted_rkd_loss": 0.013966628070920705 + }, + { + "epoch": 3.590340413151004, + "flow_matching_loss": 0.01521455030888319, + "grad_norm": 0.2139713317155838, + "learning_rate": 2.80564209080782e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00705237677320838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031397989485412838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017244552820920946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8992961347103119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12340, + "weighted_flow_matching_loss": 0.01521455030888319, + "weighted_rkd_loss": 0.015508042648434639 + }, + { + "epoch": 3.593249927262147, + "flow_matching_loss": 0.014526923932135106, + "grad_norm": 0.20614151656627655, + "learning_rate": 2.8052348945323877e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006853769673034549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003409723727963865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01711726300418377, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8991385042667389, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12350, + "weighted_flow_matching_loss": 0.014526923932135106, + "weighted_rkd_loss": 0.01539074694737792 + }, + { + "epoch": 3.596159441373291, + "flow_matching_loss": 0.015138191357254981, + "grad_norm": 0.20011146366596222, + "learning_rate": 2.8048273017610574e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006756541458889842, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030266713118180634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016539754439145328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8989807665348053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12360, + "weighted_flow_matching_loss": 0.015138191357254981, + "weighted_rkd_loss": 0.014868958666920663 + }, + { + "epoch": 3.599068955484434, + "flow_matching_loss": 0.014753567613661289, + "grad_norm": 0.17460960149765015, + "learning_rate": 2.804419312617645e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007121938979253173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002945545315742493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017189423181116582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8988229215145112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12370, + "weighted_flow_matching_loss": 0.014753567613661289, + "weighted_rkd_loss": 0.015450277924537658 + }, + { + "epoch": 3.6019784695955774, + "flow_matching_loss": 0.01598925804719329, + "grad_norm": 0.18942435085773468, + "learning_rate": 2.8040109272260862e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007176578184589744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003139485139399767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017492641415446996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8986649632453918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12380, + "weighted_flow_matching_loss": 0.01598925804719329, + "weighted_rkd_loss": 0.0157200294546783 + }, + { + "epoch": 3.604887983706721, + "flow_matching_loss": 0.014252743870019912, + "grad_norm": 0.27775657176971436, + "learning_rate": 2.8036021457104384e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006331947445869446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029876960325054824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015651590935885907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.898506897687912, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12390, + "weighted_flow_matching_loss": 0.014252743870019912, + "weighted_rkd_loss": 0.014063004963099956 + }, + { + "epoch": 3.6077974978178644, + "flow_matching_loss": 0.01620821626856923, + "grad_norm": 0.21355320513248444, + "learning_rate": 2.803192968194878e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071028884500265125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003298221295699477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017503998428583144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8983487188816071, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12400, + "weighted_flow_matching_loss": 0.01620821626856923, + "weighted_rkd_loss": 0.015724737849086522 + }, + { + "epoch": 3.6107070119290077, + "flow_matching_loss": 0.01564638437703252, + "grad_norm": 0.20941996574401855, + "learning_rate": 2.8027833948037032e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006367207784205675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002889324421994388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01562374010682106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8981904447078705, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12410, + "weighted_flow_matching_loss": 0.01564638437703252, + "weighted_rkd_loss": 0.014033113420009614 + }, + { + "epoch": 3.6136165260401514, + "flow_matching_loss": 0.01521502723917365, + "grad_norm": 0.17877215147018433, + "learning_rate": 2.802373425661331e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007111073471605778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003070685197599232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017292832117527725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8980320453643799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12420, + "weighted_flow_matching_loss": 0.01521502723917365, + "weighted_rkd_loss": 0.015529547818005085 + }, + { + "epoch": 3.6165260401512946, + "flow_matching_loss": 0.015670684445649386, + "grad_norm": 0.19283610582351685, + "learning_rate": 2.8019630608922988e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006983158458024263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032969355350360276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017263252474367618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8978735506534576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12430, + "weighted_flow_matching_loss": 0.015670684445649386, + "weighted_rkd_loss": 0.015500179585069418 + }, + { + "epoch": 3.619435554262438, + "flow_matching_loss": 0.01492502735927701, + "grad_norm": 0.1674049198627472, + "learning_rate": 2.801552300621265e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006797071872279048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031407082453370093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016734851989895105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8977149367332459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12440, + "weighted_flow_matching_loss": 0.01492502735927701, + "weighted_rkd_loss": 0.015023143123835325 + }, + { + "epoch": 3.6223450683735816, + "flow_matching_loss": 0.01572134243324399, + "grad_norm": 0.22078467905521393, + "learning_rate": 2.8011411449730084e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007352801179513336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003137021174188703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017842623777687548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8975562214851379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12450, + "weighted_flow_matching_loss": 0.01572134243324399, + "weighted_rkd_loss": 0.01601471807807684 + }, + { + "epoch": 3.6252545824847253, + "flow_matching_loss": 0.014000855665653945, + "grad_norm": 0.2027195394039154, + "learning_rate": 2.800729594072426e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00692595997825265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031412124400958417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016993132419884205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8973973870277405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12460, + "weighted_flow_matching_loss": 0.014000855665653945, + "weighted_rkd_loss": 0.015249628201127052 + }, + { + "epoch": 3.6281640965958686, + "flow_matching_loss": 0.015531690511852502, + "grad_norm": 0.1891535520553589, + "learning_rate": 2.8003176480445373e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007060723658651113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00313826163765043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017259708791971206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8972384512424469, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12470, + "weighted_flow_matching_loss": 0.015531690511852502, + "weighted_rkd_loss": 0.015486052725464106 + }, + { + "epoch": 3.631073610707012, + "flow_matching_loss": 0.015105167962610721, + "grad_norm": 0.20686748623847961, + "learning_rate": 2.7999053070144793e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006635461375117302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032082176301628353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01647914042696357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8970794141292572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12480, + "weighted_flow_matching_loss": 0.015105167962610721, + "weighted_rkd_loss": 0.014783103205263614 + }, + { + "epoch": 3.6339831248181556, + "flow_matching_loss": 0.014723632438108324, + "grad_norm": 0.1480836421251297, + "learning_rate": 2.7994925711075117e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006926675839349627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031750010326504706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017028352711349726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.896920257806778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12490, + "weighted_flow_matching_loss": 0.014723632438108324, + "weighted_rkd_loss": 0.015273160766810179 + }, + { + "epoch": 3.636892638929299, + "flow_matching_loss": 0.014012271258980035, + "grad_norm": 0.21124055981636047, + "learning_rate": 2.7990794404490118e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007138249278068542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029793819412589073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01725588059052825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8967609822750091, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12500, + "weighted_flow_matching_loss": 0.014012271258980035, + "weighted_rkd_loss": 0.015474391356110572 + }, + { + "epoch": 3.639802153040442, + "flow_matching_loss": 0.015898817870765924, + "grad_norm": 0.17994973063468933, + "learning_rate": 2.798665915164478e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00712134288623929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003257971885614097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017500657495111226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8966016113758087, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12510, + "weighted_flow_matching_loss": 0.015898817870765924, + "weighted_rkd_loss": 0.015691118873655797 + }, + { + "epoch": 3.642711667151586, + "flow_matching_loss": 0.01509970733895898, + "grad_norm": 0.20655544102191925, + "learning_rate": 2.7982519953795284e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007699856627732515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035899908747524024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018989703990519045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8964421331882477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12520, + "weighted_flow_matching_loss": 0.01509970733895898, + "weighted_rkd_loss": 0.017023112066090106 + }, + { + "epoch": 3.645621181262729, + "flow_matching_loss": 0.014006468094885349, + "grad_norm": 0.1931701898574829, + "learning_rate": 2.7978376812199013e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006427088705822826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002941322489641607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01579550011083484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.896282535791397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12530, + "weighted_flow_matching_loss": 0.014006468094885349, + "weighted_rkd_loss": 0.014157266076654195 + }, + { + "epoch": 3.6485306953738723, + "flow_matching_loss": 0.015503687504678964, + "grad_norm": 0.15153782069683075, + "learning_rate": 2.797422972811454e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006978778587654233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030116677284240724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016969224903732538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8961228489875793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12540, + "weighted_flow_matching_loss": 0.015503687504678964, + "weighted_rkd_loss": 0.015206495579332114 + }, + { + "epoch": 3.651440209485016, + "flow_matching_loss": 0.012648158241063356, + "grad_norm": 0.2147565633058548, + "learning_rate": 2.7970078702801635e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007207828667014837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031981120002456008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01761376913636923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8959630370140076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12550, + "weighted_flow_matching_loss": 0.012648158241063356, + "weighted_rkd_loss": 0.015781196672469376 + }, + { + "epoch": 3.6543497235961593, + "flow_matching_loss": 0.015001807827502489, + "grad_norm": 0.28382810950279236, + "learning_rate": 2.7965923737521283e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00785269346088171, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003664660337381065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019370047748088835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8958031177520752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12560, + "weighted_flow_matching_loss": 0.015001807827502489, + "weighted_rkd_loss": 0.01735174423083663 + }, + { + "epoch": 3.657259237707303, + "flow_matching_loss": 0.014778024423867464, + "grad_norm": 0.17719939351081848, + "learning_rate": 2.796176483353564e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006982400780543685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003294749278575182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017259550746530294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8956430971622467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12570, + "weighted_flow_matching_loss": 0.014778024423867464, + "weighted_rkd_loss": 0.015458447206765414 + }, + { + "epoch": 3.6601687518184463, + "flow_matching_loss": 0.0145903036929667, + "grad_norm": 0.2268528938293457, + "learning_rate": 2.795760199210808e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007629162399098277, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035122296307235957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018770554382354022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8954829633235931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12580, + "weighted_flow_matching_loss": 0.0145903036929667, + "weighted_rkd_loss": 0.01680865604430437 + }, + { + "epoch": 3.66307826592959, + "flow_matching_loss": 0.014446493610739708, + "grad_norm": 0.1974223554134369, + "learning_rate": 2.7953435214503157e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007199933752417564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033097188221290707, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017709586396813394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.89532271027565, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12590, + "weighted_flow_matching_loss": 0.014446493610739708, + "weighted_rkd_loss": 0.015855860058218242 + }, + { + "epoch": 3.6659877800407332, + "flow_matching_loss": 0.01477509383112192, + "grad_norm": 0.24516144394874573, + "learning_rate": 2.7949264501986634e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007116246921941638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032006350345909597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017433128878474237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8951623678207398, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12600, + "weighted_flow_matching_loss": 0.01477509383112192, + "weighted_rkd_loss": 0.015605524927377701 + }, + { + "epoch": 3.6688972941518765, + "flow_matching_loss": 0.014089093822985888, + "grad_norm": 0.20639638602733612, + "learning_rate": 2.794508985582546e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062660026829689745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003003336186520755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015535341668874025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8950019121170044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12610, + "weighted_flow_matching_loss": 0.014089093822985888, + "weighted_rkd_loss": 0.013904140330851078 + }, + { + "epoch": 3.67180680826302, + "flow_matching_loss": 0.013471606560051442, + "grad_norm": 0.18937194347381592, + "learning_rate": 2.7940911277287777e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00656204684637487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028507421258836983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01597483577206731, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.894841343164444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12620, + "weighted_flow_matching_loss": 0.013471606560051442, + "weighted_rkd_loss": 0.014294979814440011 + }, + { + "epoch": 3.6747163223741635, + "flow_matching_loss": 0.013866354338824749, + "grad_norm": 0.23391740024089813, + "learning_rate": 2.793672876764294e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007224250910803675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031813062727451325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017629808001220226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.894680666923523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12630, + "weighted_flow_matching_loss": 0.013866354338824749, + "weighted_rkd_loss": 0.01577305719256401 + }, + { + "epoch": 3.6776258364853067, + "flow_matching_loss": 0.015376785770058632, + "grad_norm": 0.2011745274066925, + "learning_rate": 2.7932542328161478e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006559623451903462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002940478967502713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016059725917875765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8945198953151703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12640, + "weighted_flow_matching_loss": 0.015376785770058632, + "weighted_rkd_loss": 0.014365722797811032 + }, + { + "epoch": 3.6805353505964504, + "flow_matching_loss": 0.013819303642958403, + "grad_norm": 0.18915633857250214, + "learning_rate": 2.792835196011512e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006782097043469548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028934926260262726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01645768666639924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8943589925765991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12650, + "weighted_flow_matching_loss": 0.013819303642958403, + "weighted_rkd_loss": 0.014719072822481393 + }, + { + "epoch": 3.6834448647075937, + "flow_matching_loss": 0.013970374874770642, + "grad_norm": 0.20007573068141937, + "learning_rate": 2.792415766477679e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00688701132312417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003100519860163331, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01687454227358103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8941979944705963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12660, + "weighted_flow_matching_loss": 0.013970374874770642, + "weighted_rkd_loss": 0.015089200530201196 + }, + { + "epoch": 3.6863543788187374, + "flow_matching_loss": 0.015139711927622557, + "grad_norm": 0.23048870265483856, + "learning_rate": 2.79199594434206e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006187592027708888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027933551697060465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015168539248406887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8940368950366974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12670, + "weighted_flow_matching_loss": 0.015139711927622557, + "weighted_rkd_loss": 0.01356114549562335 + }, + { + "epoch": 3.6892638929298807, + "flow_matching_loss": 0.015771060436964034, + "grad_norm": 0.21173472702503204, + "learning_rate": 2.7915757297321867e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007546807639300823, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033833409193903207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018476956244558095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.893875676393509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12680, + "weighted_flow_matching_loss": 0.015771060436964034, + "weighted_rkd_loss": 0.016516069974750282 + }, + { + "epoch": 3.6921734070410244, + "flow_matching_loss": 0.014031894225627183, + "grad_norm": 0.19736330211162567, + "learning_rate": 2.7911551227757085e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007249738601967692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032696632901206613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017769140563905238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8937143504619598, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12690, + "weighted_flow_matching_loss": 0.014031894225627183, + "weighted_rkd_loss": 0.015880520734936 + }, + { + "epoch": 3.6950829211521676, + "flow_matching_loss": 0.014974140655249357, + "grad_norm": 0.25325319170951843, + "learning_rate": 2.790734123600395e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007825897075235844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036276464350521566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01927944039925933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8935529232025147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12700, + "weighted_flow_matching_loss": 0.014974140655249357, + "weighted_rkd_loss": 0.01722712777554989 + }, + { + "epoch": 3.697992435263311, + "flow_matching_loss": 0.01598916118964553, + "grad_norm": 0.18462294340133667, + "learning_rate": 2.7903127323341347e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006380591634660959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029031018493697047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01566428514197469, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8933913826942443, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12710, + "weighted_flow_matching_loss": 0.01598916118964553, + "weighted_rkd_loss": 0.013994357455521822 + }, + { + "epoch": 3.7009019493744546, + "flow_matching_loss": 0.01445979755371809, + "grad_norm": 0.17370423674583435, + "learning_rate": 2.7898909491049353e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006277960957959294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027744472259655594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01533036893233657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8932297348976135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12720, + "weighted_flow_matching_loss": 0.01445979755371809, + "weighted_rkd_loss": 0.01369346808642149 + }, + { + "epoch": 3.703811463485598, + "flow_matching_loss": 0.01509397653862834, + "grad_norm": 0.20816963911056519, + "learning_rate": 2.789468774040923e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007311152061447501, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033695069840177895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017991811223328113, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8930679857730865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12730, + "weighted_flow_matching_loss": 0.01509397653862834, + "weighted_rkd_loss": 0.016067924536764623 + }, + { + "epoch": 3.706720977596741, + "flow_matching_loss": 0.015311766043305397, + "grad_norm": 0.16736574470996857, + "learning_rate": 2.7890462072703433e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007371806027367711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00337045902851969, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018114071153104306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8929061174392701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12740, + "weighted_flow_matching_loss": 0.015311766043305397, + "weighted_rkd_loss": 0.016174242924898864 + }, + { + "epoch": 3.709630491707885, + "flow_matching_loss": 0.01612724121659994, + "grad_norm": 0.16219724714756012, + "learning_rate": 2.7886232489215617e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006612887699156999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002834549848921597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01606032522395253, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8927441537380219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12750, + "weighted_flow_matching_loss": 0.01612724121659994, + "weighted_rkd_loss": 0.014337813947349787 + }, + { + "epoch": 3.712540005819028, + "flow_matching_loss": 0.015093466080725194, + "grad_norm": 0.2145688235759735, + "learning_rate": 2.7881998991230606e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007469442393630743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033000185852870346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01823890320956707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8925820589065552, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12760, + "weighted_flow_matching_loss": 0.015093466080725194, + "weighted_rkd_loss": 0.016279625799506904 + }, + { + "epoch": 3.715449519930172, + "flow_matching_loss": 0.014600137248635292, + "grad_norm": 0.17894911766052246, + "learning_rate": 2.7877761580034436e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006845671962946653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003083130205050111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016774473898112773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8924198806285858, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12770, + "weighted_flow_matching_loss": 0.014600137248635292, + "weighted_rkd_loss": 0.01496993163600564 + }, + { + "epoch": 3.718359034041315, + "flow_matching_loss": 0.013457033038139343, + "grad_norm": 0.17680613696575165, + "learning_rate": 2.787352025691431e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007378989597782493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032791289035230876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018037107959389686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8922575950622559, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12780, + "weighted_flow_matching_loss": 0.013457033038139343, + "weighted_rkd_loss": 0.016093811299651862 + }, + { + "epoch": 3.7212685481524588, + "flow_matching_loss": 0.013791078049689532, + "grad_norm": 0.1944909542798996, + "learning_rate": 2.7869275023158642e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063549859914928675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029276185436174274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015637590363621712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8920951902866363, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12790, + "weighted_flow_matching_loss": 0.013791078049689532, + "weighted_rkd_loss": 0.013950176164507866 + }, + { + "epoch": 3.724178062263602, + "flow_matching_loss": 0.013856770377606154, + "grad_norm": 0.24215343594551086, + "learning_rate": 2.7865025880057014e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071139564272016285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003309876122511923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017537788674235343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8919326782226562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12800, + "weighted_flow_matching_loss": 0.013856770377606154, + "weighted_rkd_loss": 0.015642521809786557 + }, + { + "epoch": 3.7270875763747453, + "flow_matching_loss": 0.015379968751221895, + "grad_norm": 0.19637610018253326, + "learning_rate": 2.78607728289002e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006339432392269373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027803145116195084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015459179319441319, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8917700707912445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12810, + "weighted_flow_matching_loss": 0.015379968751221895, + "weighted_rkd_loss": 0.013786005787551403 + }, + { + "epoch": 3.729997090485889, + "flow_matching_loss": 0.01477507445961237, + "grad_norm": 0.20137999951839447, + "learning_rate": 2.7856515870980176e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006570217851549387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029662720626220107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01610670769587159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8916073441505432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12820, + "weighted_flow_matching_loss": 0.01477507445961237, + "weighted_rkd_loss": 0.014360892400145531 + }, + { + "epoch": 3.7329066045970323, + "flow_matching_loss": 0.013558899145573377, + "grad_norm": 0.24421179294586182, + "learning_rate": 2.785225500759008e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006083173956722021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002621419820934534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014787767734378576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8914445161819458, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12830, + "weighted_flow_matching_loss": 0.013558899145573377, + "weighted_rkd_loss": 0.013182452414184809 + }, + { + "epoch": 3.7358161187081755, + "flow_matching_loss": 0.014580099657177926, + "grad_norm": 0.19545677304267883, + "learning_rate": 2.7847990240024263e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006691728113219142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029208941850811243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016304350271821023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8912815749645233, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12840, + "weighted_flow_matching_loss": 0.014580099657177926, + "weighted_rkd_loss": 0.014531740359961987 + }, + { + "epoch": 3.7387256328193192, + "flow_matching_loss": 0.013736703526228666, + "grad_norm": 0.2180783599615097, + "learning_rate": 2.784372156957824e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069654866121709345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034094084054231643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017340381722897293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8911185324192047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12850, + "weighted_flow_matching_loss": 0.013736703526228666, + "weighted_rkd_loss": 0.015452332980930805 + }, + { + "epoch": 3.7416351469304625, + "flow_matching_loss": 0.01517219226807356, + "grad_norm": 0.2458879053592682, + "learning_rate": 2.7839448997548728e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006376453582197428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003073413181118667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01582632027566433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8909553766250611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12860, + "weighted_flow_matching_loss": 0.01517219226807356, + "weighted_rkd_loss": 0.014100482501089574 + }, + { + "epoch": 3.744544661041606, + "flow_matching_loss": 0.015131704602390527, + "grad_norm": 0.20314331352710724, + "learning_rate": 2.783517252523361e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006645856564864517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028561899205669762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01614790316671133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8907921314239502, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12870, + "weighted_flow_matching_loss": 0.015131704602390527, + "weighted_rkd_loss": 0.014384423196315766 + }, + { + "epoch": 3.7474541751527495, + "flow_matching_loss": 0.014211839996278286, + "grad_norm": 0.2517737150192261, + "learning_rate": 2.7830892153931975e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006939302990213036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030361325945705174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01691473862156272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8906287431716919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12880, + "weighted_flow_matching_loss": 0.014211839996278286, + "weighted_rkd_loss": 0.015064702834933996 + }, + { + "epoch": 3.750363689263893, + "flow_matching_loss": 0.014321825280785561, + "grad_norm": 0.2015225738286972, + "learning_rate": 2.7826607884944083e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006207936559803784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028739315806888045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015289804525673389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8904652774333954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12890, + "weighted_flow_matching_loss": 0.014321825280785561, + "weighted_rkd_loss": 0.01361505538225174 + }, + { + "epoch": 3.7532732033750364, + "flow_matching_loss": 0.014039111416786909, + "grad_norm": 0.21782425045967102, + "learning_rate": 2.782231971957138e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007284059189260006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003367954958230257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017936073523014785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8903016984462738, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12900, + "weighted_flow_matching_loss": 0.014039111416786909, + "weighted_rkd_loss": 0.01596852419897914 + }, + { + "epoch": 3.7561827174861797, + "flow_matching_loss": 0.016267824079841377, + "grad_norm": 0.2728240489959717, + "learning_rate": 2.78180276591165e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006820004200562835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003069023136049509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016709031630307437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8901380062103271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12910, + "weighted_flow_matching_loss": 0.016267824079841377, + "weighted_rkd_loss": 0.014873223286122084 + }, + { + "epoch": 3.7590922315973234, + "flow_matching_loss": 0.014446364156901837, + "grad_norm": 0.20748712122440338, + "learning_rate": 2.781373170488326e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00709656816907227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034678998868912458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01766103645786643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8899742066860199, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12920, + "weighted_flow_matching_loss": 0.014446364156901837, + "weighted_rkd_loss": 0.01571784121915698 + }, + { + "epoch": 3.7620017457084667, + "flow_matching_loss": 0.013221897557377816, + "grad_norm": 0.1888139247894287, + "learning_rate": 2.7809431858176655e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006690721260383725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032279381528496744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016609380673617124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8898102939128876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12930, + "weighted_flow_matching_loss": 0.013221897557377816, + "weighted_rkd_loss": 0.014779153652489186 + }, + { + "epoch": 3.76491125981961, + "flow_matching_loss": 0.01461099050939083, + "grad_norm": 0.19709165394306183, + "learning_rate": 2.7805128120302865e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007847957732155918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003562566451728344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019258482102304696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.889646303653717, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12940, + "weighted_flow_matching_loss": 0.01461099050939083, + "weighted_rkd_loss": 0.017133313789963724 + }, + { + "epoch": 3.7678207739307537, + "flow_matching_loss": 0.014492148812860251, + "grad_norm": 0.22029125690460205, + "learning_rate": 2.7800820492569254e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00653731394559145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028606516076251864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01593527952209115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8894821882247925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12950, + "weighted_flow_matching_loss": 0.014492148812860251, + "weighted_rkd_loss": 0.014174151979386807 + }, + { + "epoch": 3.770730288041897, + "flow_matching_loss": 0.015934412088245153, + "grad_norm": 0.22013917565345764, + "learning_rate": 2.779650897628436e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006898449175059796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031198534881696103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016916751861572266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8893179655075073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12960, + "weighted_flow_matching_loss": 0.015934412088245153, + "weighted_rkd_loss": 0.015044352784752846 + }, + { + "epoch": 3.77363980215304, + "flow_matching_loss": 0.013478851038962602, + "grad_norm": 0.19562599062919617, + "learning_rate": 2.7792193572757915e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006446840846911073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002902192436158657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015795874036848546, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8891536355018616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12970, + "weighted_flow_matching_loss": 0.013478851038962602, + "weighted_rkd_loss": 0.014044938888400794 + }, + { + "epoch": 3.776549316264184, + "flow_matching_loss": 0.01584439966827631, + "grad_norm": 0.2677006721496582, + "learning_rate": 2.7787874283300812e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007186067709699273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00334274263586849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0177148780785501, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8889892101287842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12980, + "weighted_flow_matching_loss": 0.01584439966827631, + "weighted_rkd_loss": 0.015748287737369537 + }, + { + "epoch": 3.7794588303753276, + "flow_matching_loss": 0.013578227441757917, + "grad_norm": 0.21214954555034637, + "learning_rate": 2.7783551109225155e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006389560643583536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029483775957487523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01572749875485897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8888246655464173, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12990, + "weighted_flow_matching_loss": 0.013578227441757917, + "weighted_rkd_loss": 0.013978908862918615 + }, + { + "epoch": 3.782368344486471, + "flow_matching_loss": 0.015617065969854594, + "grad_norm": 0.23776108026504517, + "learning_rate": 2.7779224051844195e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007456663995981216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003534750663675368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018448078725486993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8886600017547608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13000, + "weighted_flow_matching_loss": 0.015617065969854594, + "weighted_rkd_loss": 0.016393955424427988 + }, + { + "epoch": 3.785277858597614, + "flow_matching_loss": 0.01398570640012622, + "grad_norm": 0.2109193205833435, + "learning_rate": 2.777489311247239e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006954957544803619, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032301518600434064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017140066623687743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8884952604770661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13010, + "weighted_flow_matching_loss": 0.01398570640012622, + "weighted_rkd_loss": 0.015228989534080029 + }, + { + "epoch": 3.788187372708758, + "flow_matching_loss": 0.015229309629648923, + "grad_norm": 0.1978253722190857, + "learning_rate": 2.7770558292425355e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006912800995633006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032087066909298302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01703430851921439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8883303940296173, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13020, + "weighted_flow_matching_loss": 0.015229309629648923, + "weighted_rkd_loss": 0.015132121089845896 + }, + { + "epoch": 3.791096886819901, + "flow_matching_loss": 0.01534634903073311, + "grad_norm": 0.20149411261081696, + "learning_rate": 2.77662195930199e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006465501943603158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002756880037486553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01568788420408964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.888165420293808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13030, + "weighted_flow_matching_loss": 0.01534634903073311, + "weighted_rkd_loss": 0.013933491241186858 + }, + { + "epoch": 3.7940064009310444, + "flow_matching_loss": 0.012712616845965386, + "grad_norm": 0.2000812441110611, + "learning_rate": 2.7761877015574005e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006591818667948246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003026997053530067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016210634261369705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8880003392696381, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13040, + "weighted_flow_matching_loss": 0.012712616845965386, + "weighted_rkd_loss": 0.014395057410001754 + }, + { + "epoch": 3.796915915042188, + "flow_matching_loss": 0.014485939592123031, + "grad_norm": 0.1884080320596695, + "learning_rate": 2.775753056140683e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00608156188391149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026225583627820016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014785682037472724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.887835168838501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13050, + "weighted_flow_matching_loss": 0.014485939592123031, + "weighted_rkd_loss": 0.013127278350293637 + }, + { + "epoch": 3.7998254291533313, + "flow_matching_loss": 0.01366502307355404, + "grad_norm": 0.19728052616119385, + "learning_rate": 2.775318023183871e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068960913456976415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030484791845083238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016840661969035865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8876698732376098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13060, + "weighted_flow_matching_loss": 0.01366502307355404, + "weighted_rkd_loss": 0.014948949497193099 + }, + { + "epoch": 3.8027349432644746, + "flow_matching_loss": 0.015038081258535386, + "grad_norm": 0.2084517478942871, + "learning_rate": 2.774882602819117e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006978593301028013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003419357514940202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017376544140279292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8875044822692871, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13070, + "weighted_flow_matching_loss": 0.015038081258535386, + "weighted_rkd_loss": 0.015421766694635153 + }, + { + "epoch": 3.8056444573756183, + "flow_matching_loss": 0.014960604719817638, + "grad_norm": 0.23458494246006012, + "learning_rate": 2.7744467951786892e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00716891186311841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031691098818555476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01750693367794156, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8873389899730683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13080, + "weighted_flow_matching_loss": 0.014960604719817638, + "weighted_rkd_loss": 0.015534570254385472 + }, + { + "epoch": 3.8085539714867616, + "flow_matching_loss": 0.015489387698471547, + "grad_norm": 0.24376167356967926, + "learning_rate": 2.7740106003949747e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007008100533857942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031907686265185476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017206969577819108, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8871733844280243, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13090, + "weighted_flow_matching_loss": 0.015489387698471547, + "weighted_rkd_loss": 0.01526551302522421 + }, + { + "epoch": 3.8114634855979053, + "flow_matching_loss": 0.015525261778384447, + "grad_norm": 0.23225055634975433, + "learning_rate": 2.7735740186004785e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064142779912799595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029323506867513062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01576090669259429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8870076715946198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13100, + "weighted_flow_matching_loss": 0.015525261778384447, + "weighted_rkd_loss": 0.013980031106621027 + }, + { + "epoch": 3.8143729997090485, + "flow_matching_loss": 0.014669401943683625, + "grad_norm": 0.19408689439296722, + "learning_rate": 2.7731370499278222e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006929957354441285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030442998046055435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016904214676469565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8868418455123901, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13110, + "weighted_flow_matching_loss": 0.014669401943683625, + "weighted_rkd_loss": 0.014991340134292841 + }, + { + "epoch": 3.8172825138201922, + "flow_matching_loss": 0.014091173745691776, + "grad_norm": 0.17516757547855377, + "learning_rate": 2.772699694509745e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064461135771125555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003023911826312542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015916138794273137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8866759181022644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13120, + "weighted_flow_matching_loss": 0.014091173745691776, + "weighted_rkd_loss": 0.014112481474876403 + }, + { + "epoch": 3.8201920279313355, + "flow_matching_loss": 0.01622673748061061, + "grad_norm": 0.1977148950099945, + "learning_rate": 2.7722619524791046e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007034184969961643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032748515252023934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017343221139162777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.886509895324707, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13130, + "weighted_flow_matching_loss": 0.01622673748061061, + "weighted_rkd_loss": 0.015374966710805894 + }, + { + "epoch": 3.8231015420424788, + "flow_matching_loss": 0.01542305201292038, + "grad_norm": 0.20674623548984528, + "learning_rate": 2.7718238239688747e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007075869524851441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031314149033278226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017283153906464576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8863437533378601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13140, + "weighted_flow_matching_loss": 0.01542305201292038, + "weighted_rkd_loss": 0.015318816620856523 + }, + { + "epoch": 3.8260110561536225, + "flow_matching_loss": 0.012720337696373462, + "grad_norm": 0.16520844399929047, + "learning_rate": 2.7713853091121484e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006539829494431615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00287762728985399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015957286395132542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8861775100231171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13150, + "weighted_flow_matching_loss": 0.012720337696373462, + "weighted_rkd_loss": 0.014140987675637007 + }, + { + "epoch": 3.8289205702647657, + "flow_matching_loss": 0.015449755173176528, + "grad_norm": 0.21248677372932434, + "learning_rate": 2.7709464080421338e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006620878167450428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002870542462915182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016112298984080554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8860111713409424, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13160, + "weighted_flow_matching_loss": 0.015449755173176528, + "weighted_rkd_loss": 0.014275703392922879 + }, + { + "epoch": 3.831830084375909, + "flow_matching_loss": 0.01618032744154334, + "grad_norm": 0.21935752034187317, + "learning_rate": 2.770507120892158e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00649423934519291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028544823406264188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015842960961163045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8858447074890137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13170, + "weighted_flow_matching_loss": 0.01618032744154334, + "weighted_rkd_loss": 0.014034479670226575 + }, + { + "epoch": 3.8347395984870527, + "flow_matching_loss": 0.0142553704790771, + "grad_norm": 0.2280743569135666, + "learning_rate": 2.7700674477956643e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006673518754541874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029252876294776796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016272325068712234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8856781482696533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13180, + "weighted_flow_matching_loss": 0.0142553704790771, + "weighted_rkd_loss": 0.01441202275454998 + }, + { + "epoch": 3.837649112598196, + "flow_matching_loss": 0.013858848530799151, + "grad_norm": 0.21816891431808472, + "learning_rate": 2.7696273888862145e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007166642555966973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003499623970128596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017832909151911734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8855114817619324, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13190, + "weighted_flow_matching_loss": 0.013858848530799151, + "weighted_rkd_loss": 0.015791141521185638 + }, + { + "epoch": 3.8405586267093397, + "flow_matching_loss": 0.013899907004088163, + "grad_norm": 0.21501047909259796, + "learning_rate": 2.7691869442974864e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006451080413535237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028230040334165096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015725165139883757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8853447198867798, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13200, + "weighted_flow_matching_loss": 0.013899907004088163, + "weighted_rkd_loss": 0.01392215359956026 + }, + { + "epoch": 3.843468140820483, + "flow_matching_loss": 0.014538659248501062, + "grad_norm": 0.17098307609558105, + "learning_rate": 2.7687461141632754e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00673073953948915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003012138488702476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016473617311567067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8851778447628021, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13210, + "weighted_flow_matching_loss": 0.014538659248501062, + "weighted_rkd_loss": 0.014582204166799784 + }, + { + "epoch": 3.8463776549316266, + "flow_matching_loss": 0.016329449508339165, + "grad_norm": 0.19898729026317596, + "learning_rate": 2.7683048986174945e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006666438188403845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028542346321046353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016187111008912326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8850108504295349, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13220, + "weighted_flow_matching_loss": 0.016329449508339165, + "weighted_rkd_loss": 0.01432577008381486 + }, + { + "epoch": 3.84928716904277, + "flow_matching_loss": 0.014416616596281528, + "grad_norm": 0.19817788898944855, + "learning_rate": 2.7678632977941727e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006329668173566461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028749377001076938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015534274093806744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8848437666893005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13230, + "weighted_flow_matching_loss": 0.014416616596281528, + "weighted_rkd_loss": 0.01374530540779233 + }, + { + "epoch": 3.852196683153913, + "flow_matching_loss": 0.014094795193523168, + "grad_norm": 0.19497458636760712, + "learning_rate": 2.767421311827457e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066770358942449095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002935138950124383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01628921078518033, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8846765697002411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13240, + "weighted_flow_matching_loss": 0.014094795193523168, + "weighted_rkd_loss": 0.014410688634961843 + }, + { + "epoch": 3.855106197265057, + "flow_matching_loss": 0.015358765237033367, + "grad_norm": 0.17507383227348328, + "learning_rate": 2.7669789408516117e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006299257557839155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029248603852465747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015523375570774078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.88450927734375, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13250, + "weighted_flow_matching_loss": 0.015358765237033367, + "weighted_rkd_loss": 0.013730604108422995 + }, + { + "epoch": 3.8580157113762, + "flow_matching_loss": 0.015316731669008732, + "grad_norm": 0.22892923653125763, + "learning_rate": 2.766536185001016e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068967463448643684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034880232764407993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01728151608258486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8843418657779694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13260, + "weighted_flow_matching_loss": 0.015316731669008732, + "weighted_rkd_loss": 0.015282721910625695 + }, + { + "epoch": 3.8609252254873434, + "flow_matching_loss": 0.015200963523238898, + "grad_norm": 0.2297356277704239, + "learning_rate": 2.7660930444101692e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006772949034348131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003134430735372007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01668032892048359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.884174358844757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13270, + "weighted_flow_matching_loss": 0.015200963523238898, + "weighted_rkd_loss": 0.014748255908489227 + }, + { + "epoch": 3.863834739598487, + "flow_matching_loss": 0.013383951876312495, + "grad_norm": 0.18277621269226074, + "learning_rate": 2.7656495192136842e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006871703686192632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003100290964357555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01684369845315814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8840067505836486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13280, + "weighted_flow_matching_loss": 0.013383951876312495, + "weighted_rkd_loss": 0.014889962039887905 + }, + { + "epoch": 3.8667442537096304, + "flow_matching_loss": 0.015764846373349427, + "grad_norm": 0.19372449815273285, + "learning_rate": 2.7652056095462926e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006140146125108004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028390134451910854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015119305532425641, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8838390231132507, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13290, + "weighted_flow_matching_loss": 0.015764846373349427, + "weighted_rkd_loss": 0.013362996093928815 + }, + { + "epoch": 3.869653767820774, + "flow_matching_loss": 0.014944733679294586, + "grad_norm": 0.21835879981517792, + "learning_rate": 2.764761315542843e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006410261010751128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029988422291353347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015819364320486785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8836711943149567, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13300, + "weighted_flow_matching_loss": 0.014944733679294586, + "weighted_rkd_loss": 0.013979150168597698 + }, + { + "epoch": 3.8725632819319173, + "flow_matching_loss": 0.01351904971525073, + "grad_norm": 0.19487914443016052, + "learning_rate": 2.7643166373383e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063999560195952656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029203372076153757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015720249246805908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8835032701492309, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13310, + "weighted_flow_matching_loss": 0.01351904971525073, + "weighted_rkd_loss": 0.013888910692185164 + }, + { + "epoch": 3.875472796043061, + "flow_matching_loss": 0.014559641107916832, + "grad_norm": 0.25215938687324524, + "learning_rate": 2.763871575067744e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006837603449821472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032527719624340532, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01692797876894474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8833352327346802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13320, + "weighted_flow_matching_loss": 0.014559641107916832, + "weighted_rkd_loss": 0.014953048340976239 + }, + { + "epoch": 3.8783823101542043, + "flow_matching_loss": 0.015205446816980838, + "grad_norm": 0.217137411236763, + "learning_rate": 2.763426128866375e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006461528223007917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002784637385047972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015707693807780743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8831670939922333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13330, + "weighted_flow_matching_loss": 0.015205446816980838, + "weighted_rkd_loss": 0.013872534409165383 + }, + { + "epoch": 3.8812918242653476, + "flow_matching_loss": 0.014957612846046686, + "grad_norm": 0.22229549288749695, + "learning_rate": 2.7629802988695064e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006968171196058392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030200368957594035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01695637935772538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8829988479614258, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13340, + "weighted_flow_matching_loss": 0.014957612846046686, + "weighted_rkd_loss": 0.01497244881466031 + }, + { + "epoch": 3.8842013383764913, + "flow_matching_loss": 0.014506491739302874, + "grad_norm": 0.18442821502685547, + "learning_rate": 2.7625340852125703e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00659738858230412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002925349329598248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01612012628465891, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8828304946422577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13350, + "weighted_flow_matching_loss": 0.014506491739302874, + "weighted_rkd_loss": 0.014231319911777974 + }, + { + "epoch": 3.8871108524876345, + "flow_matching_loss": 0.01576806986704469, + "grad_norm": 0.20569178462028503, + "learning_rate": 2.7620874880311148e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072054615244269375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033730092458426953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017783932201564313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.882662034034729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13360, + "weighted_flow_matching_loss": 0.01576806986704469, + "weighted_rkd_loss": 0.01569710299372673 + }, + { + "epoch": 3.890020366598778, + "flow_matching_loss": 0.015437625721096993, + "grad_norm": 0.22305162250995636, + "learning_rate": 2.7616405074608033e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006793057639151811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030948739731684327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016680989135056733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8824934780597686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13370, + "weighted_flow_matching_loss": 0.015437625721096993, + "weighted_rkd_loss": 0.01472087623551488 + }, + { + "epoch": 3.8929298807099215, + "flow_matching_loss": 0.015215374995023012, + "grad_norm": 0.2353086918592453, + "learning_rate": 2.7611931436374173e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006870734365656972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032035749638453124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01694504376500845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8823248088359833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13380, + "weighted_flow_matching_loss": 0.015215374995023012, + "weighted_rkd_loss": 0.014951080083847046 + }, + { + "epoch": 3.895839394821065, + "flow_matching_loss": 0.013490007352083921, + "grad_norm": 0.2109822779893875, + "learning_rate": 2.7607453966968545e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006043565180152655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002661060728132725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01474819118157029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8821560442447662, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13390, + "weighted_flow_matching_loss": 0.013490007352083921, + "weighted_rkd_loss": 0.013010130450129508 + }, + { + "epoch": 3.898748908932208, + "flow_matching_loss": 0.01506607923656702, + "grad_norm": 0.2007301300764084, + "learning_rate": 2.760297266775128e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007047300552949309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003121362195815891, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017215963453054428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8819871604442596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13400, + "weighted_flow_matching_loss": 0.01506607923656702, + "weighted_rkd_loss": 0.01518425215035677 + }, + { + "epoch": 3.9016584230433518, + "flow_matching_loss": 0.014457394927740097, + "grad_norm": 0.22975897789001465, + "learning_rate": 2.759848754008368e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006403698353096843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002928473264910281, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01573587004095316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8818181872367858, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13410, + "weighted_flow_matching_loss": 0.014457394927740097, + "weighted_rkd_loss": 0.013876165077090264 + }, + { + "epoch": 3.9045679371544955, + "flow_matching_loss": 0.013726693857461215, + "grad_norm": 0.22302962839603424, + "learning_rate": 2.759399858532821e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006709251133725047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003129951050505042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016548453364521265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8816490948200226, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13420, + "weighted_flow_matching_loss": 0.013726693857461215, + "weighted_rkd_loss": 0.014589924551546573 + }, + { + "epoch": 3.9074774512656387, + "flow_matching_loss": 0.014107710774987936, + "grad_norm": 0.19249650835990906, + "learning_rate": 2.75895058048485e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007038816809654236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003170223068445921, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017247856874018908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8814799129962921, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13430, + "weighted_flow_matching_loss": 0.014107710774987936, + "weighted_rkd_loss": 0.01520374109968543 + }, + { + "epoch": 3.910386965376782, + "flow_matching_loss": 0.013823703676462174, + "grad_norm": 0.18353641033172607, + "learning_rate": 2.7585009200009326e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007529268926009536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033961525419726967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01845469046384096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8813106179237366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13440, + "weighted_flow_matching_loss": 0.013823703676462174, + "weighted_rkd_loss": 0.016264419350773097 + }, + { + "epoch": 3.9132964794879257, + "flow_matching_loss": 0.013473005499690771, + "grad_norm": 0.1889769732952118, + "learning_rate": 2.7580508772176648e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00671726712025702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028940644348040222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01632859855890274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.881141209602356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13450, + "weighted_flow_matching_loss": 0.013473005499690771, + "weighted_rkd_loss": 0.014387804921716452 + }, + { + "epoch": 3.916205993599069, + "flow_matching_loss": 0.014624237734824418, + "grad_norm": 0.1889302283525467, + "learning_rate": 2.7576004522717577e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007035225816071034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032071212073788045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01727757276967168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8809717118740081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13460, + "weighted_flow_matching_loss": 0.014624237734824418, + "weighted_rkd_loss": 0.015221154037863015 + }, + { + "epoch": 3.919115507710212, + "flow_matching_loss": 0.014386584423482418, + "grad_norm": 0.21055205166339874, + "learning_rate": 2.7571496453000377e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006551517732441425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029424514854326845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016045486833900214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8808021068572998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13470, + "weighted_flow_matching_loss": 0.014386584423482418, + "weighted_rkd_loss": 0.01413290286436677 + }, + { + "epoch": 3.922025021821356, + "flow_matching_loss": 0.01435056384652853, + "grad_norm": 0.20078985393047333, + "learning_rate": 2.7566984564394487e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006043716426938772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027730960049666465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014860528893768788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8806323945522309, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13480, + "weighted_flow_matching_loss": 0.01435056384652853, + "weighted_rkd_loss": 0.01308667277917266 + }, + { + "epoch": 3.924934535932499, + "flow_matching_loss": 0.014755633566528559, + "grad_norm": 0.2044057846069336, + "learning_rate": 2.7562468858270498e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006139472080394626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026916506001725795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014970594737678766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8804625570774078, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13490, + "weighted_flow_matching_loss": 0.014755633566528559, + "weighted_rkd_loss": 0.01318100718781352 + }, + { + "epoch": 3.9278440500436425, + "flow_matching_loss": 0.015117550920695066, + "grad_norm": 0.17938505113124847, + "learning_rate": 2.7557949336000165e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070462329778820275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003250694298185408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017343160323798658, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8802926540374756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13500, + "weighted_flow_matching_loss": 0.015117550920695066, + "weighted_rkd_loss": 0.015267178043723106 + }, + { + "epoch": 3.930753564154786, + "flow_matching_loss": 0.014606955181807279, + "grad_norm": 0.1736886203289032, + "learning_rate": 2.7553425998956395e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006405015476047993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027485397644340993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015558570716530084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8801226258277893, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13510, + "weighted_flow_matching_loss": 0.014606955181807279, + "weighted_rkd_loss": 0.013693441078066826 + }, + { + "epoch": 3.9336630782659294, + "flow_matching_loss": 0.013385444041341543, + "grad_norm": 0.19433341920375824, + "learning_rate": 2.7548898848513264e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006652855034917593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028643298777751625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01617004005238414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8799524903297424, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13520, + "weighted_flow_matching_loss": 0.013385444041341543, + "weighted_rkd_loss": 0.014228910394012927 + }, + { + "epoch": 3.936572592377073, + "flow_matching_loss": 0.015278023947030307, + "grad_norm": 0.18680140376091003, + "learning_rate": 2.754436788604599e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061231156345456835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028238419676199556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015070073306560516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8797822654247284, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13530, + "weighted_flow_matching_loss": 0.015278023947030307, + "weighted_rkd_loss": 0.013258372247219086 + }, + { + "epoch": 3.9394821064882164, + "flow_matching_loss": 0.012789780646562577, + "grad_norm": 0.23440870642662048, + "learning_rate": 2.7539833112930973e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00690329815261066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031079399399459363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016914536245167255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8796119332313538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13540, + "weighted_flow_matching_loss": 0.012789780646562577, + "weighted_rkd_loss": 0.01487816208973527 + }, + { + "epoch": 3.94239162059936, + "flow_matching_loss": 0.013250744249671698, + "grad_norm": 0.17254222929477692, + "learning_rate": 2.753529453054575e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006512889824807644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026288889697752894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01565466858446598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8794414937496186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13550, + "weighted_flow_matching_loss": 0.013250744249671698, + "weighted_rkd_loss": 0.013767407927662133 + }, + { + "epoch": 3.9453011347105034, + "flow_matching_loss": 0.014131175633519889, + "grad_norm": 0.19254504144191742, + "learning_rate": 2.7530752140269028e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006573924515396357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028523447923362254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016000193916261195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8792709350585938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13560, + "weighted_flow_matching_loss": 0.014131175633519889, + "weighted_rkd_loss": 0.014068475272506476 + }, + { + "epoch": 3.9482106488216466, + "flow_matching_loss": 0.014913223963230849, + "grad_norm": 0.18708598613739014, + "learning_rate": 2.7526205943480657e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006065705744549632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028568662237375975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01498827775940299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8791002929210663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13570, + "weighted_flow_matching_loss": 0.014913223963230849, + "weighted_rkd_loss": 0.013176175393164158 + }, + { + "epoch": 3.9511201629327903, + "flow_matching_loss": 0.013768612360581755, + "grad_norm": 0.1966656744480133, + "learning_rate": 2.7521655941561653e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006443332135677338, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002953560883179307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015840225107967853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8789295375347137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13580, + "weighted_flow_matching_loss": 0.013768612360581755, + "weighted_rkd_loss": 0.013922365102916955 + }, + { + "epoch": 3.9540296770439336, + "flow_matching_loss": 0.013563521206378937, + "grad_norm": 0.19661913812160492, + "learning_rate": 2.751710213589419e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006440657330676914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00296017627697438, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015841490868479014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.878758680820465, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13590, + "weighted_flow_matching_loss": 0.013563521206378937, + "weighted_rkd_loss": 0.013920814264565706 + }, + { + "epoch": 3.956939191155077, + "flow_matching_loss": 0.014572106953710317, + "grad_norm": 0.20148758590221405, + "learning_rate": 2.7512544527861597e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006414926052093506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002808380080386996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015638232138007878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8785877227783203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13600, + "weighted_flow_matching_loss": 0.014572106953710317, + "weighted_rkd_loss": 0.013739488273859023 + }, + { + "epoch": 3.9598487052662206, + "flow_matching_loss": 0.013786005973815917, + "grad_norm": 0.1492842435836792, + "learning_rate": 2.7507983118848353e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006057870760560036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002792459132615477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014908200595527887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8784166574478149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13610, + "weighted_flow_matching_loss": 0.013786005973815917, + "weighted_rkd_loss": 0.013095622975379228 + }, + { + "epoch": 3.962758219377364, + "flow_matching_loss": 0.01488974755629897, + "grad_norm": 0.22084678709506989, + "learning_rate": 2.7503417910240085e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065942369401454926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029095203615725042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01609799424186349, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8782454848289489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13620, + "weighted_flow_matching_loss": 0.01488974755629897, + "weighted_rkd_loss": 0.014137956779450178 + }, + { + "epoch": 3.9656677334885075, + "flow_matching_loss": 0.01422033766284585, + "grad_norm": 0.19032257795333862, + "learning_rate": 2.7498848903423593e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006761748576536775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030774036422371864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016600900702178477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8780742168426514, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13630, + "weighted_flow_matching_loss": 0.01422033766284585, + "weighted_rkd_loss": 0.014576809760183097 + }, + { + "epoch": 3.968577247599651, + "flow_matching_loss": 0.014097067900002003, + "grad_norm": 0.19640767574310303, + "learning_rate": 2.7494276099786817e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005758190620690584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025019732187502087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014018354378640652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8779028415679931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13640, + "weighted_flow_matching_loss": 0.014097067900002003, + "weighted_rkd_loss": 0.012306749261915684 + }, + { + "epoch": 3.9714867617107945, + "flow_matching_loss": 0.012807605043053626, + "grad_norm": 0.1873341053724289, + "learning_rate": 2.7489699500718854e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006626095250248909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031873584259301422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01643954887986183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8777313530445099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13650, + "weighted_flow_matching_loss": 0.012807605043053626, + "weighted_rkd_loss": 0.014429405704140663 + }, + { + "epoch": 3.9743962758219378, + "flow_matching_loss": 0.013158835005015135, + "grad_norm": 0.24408546090126038, + "learning_rate": 2.7485119107609953e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00606025573797524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002762975252699107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014883486740291118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8775597810745239, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13660, + "weighted_flow_matching_loss": 0.013158835005015135, + "weighted_rkd_loss": 0.013061160035431386 + }, + { + "epoch": 3.977305789933081, + "flow_matching_loss": 0.013772565871477127, + "grad_norm": 0.2067662924528122, + "learning_rate": 2.748053492185152e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006256632460281253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029764431994408367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015489708352833986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8773880898952484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13670, + "weighted_flow_matching_loss": 0.013772565871477127, + "weighted_rkd_loss": 0.013590471353381872 + }, + { + "epoch": 3.9802153040442247, + "flow_matching_loss": 0.013368800468742847, + "grad_norm": 0.21540483832359314, + "learning_rate": 2.7475946944836103e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006379981245845556, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028860393445938825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015646001696586607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8772162973880768, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13680, + "weighted_flow_matching_loss": 0.013368800468742847, + "weighted_rkd_loss": 0.013724826276302338 + }, + { + "epoch": 3.983124818155368, + "flow_matching_loss": 0.016763212252408267, + "grad_norm": 0.2108507752418518, + "learning_rate": 2.7471355177957414e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006849412387236953, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00324720551725477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016946030501276254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.877044403553009, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13690, + "weighted_flow_matching_loss": 0.016763212252408267, + "weighted_rkd_loss": 0.014862434100359678 + }, + { + "epoch": 3.9860343322665113, + "flow_matching_loss": 0.014265501685440541, + "grad_norm": 0.1861128807067871, + "learning_rate": 2.746675962261031e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006397421984001994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030328912194818257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015827735140919685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8768724024295806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13700, + "weighted_flow_matching_loss": 0.014265501685440541, + "weighted_rkd_loss": 0.01387900933623314 + }, + { + "epoch": 3.988943846377655, + "flow_matching_loss": 0.015076951123774052, + "grad_norm": 0.22005093097686768, + "learning_rate": 2.74621602801908e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006660273484885693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031910892575979233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016511636227369307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8767002999782563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13710, + "weighted_flow_matching_loss": 0.015076951123774052, + "weighted_rkd_loss": 0.014475722704082728 + }, + { + "epoch": 3.9918533604887982, + "flow_matching_loss": 0.013187254127115012, + "grad_norm": 0.2408423274755478, + "learning_rate": 2.745755715209604e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00675733839161694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003071614191867411, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01658629085868597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8765281140804291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13720, + "weighted_flow_matching_loss": 0.013187254127115012, + "weighted_rkd_loss": 0.014538379851728678 + }, + { + "epoch": 3.994762874599942, + "flow_matching_loss": 0.013804137334227562, + "grad_norm": 0.18015050888061523, + "learning_rate": 2.7452950239724337e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006545648956671357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003015235299244523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01610653307288885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.876355791091919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13730, + "weighted_flow_matching_loss": 0.013804137334227562, + "weighted_rkd_loss": 0.014115034788846969 + }, + { + "epoch": 3.997672388711085, + "flow_matching_loss": 0.014040630217641592, + "grad_norm": 0.19445493817329407, + "learning_rate": 2.7448339544475156e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005721224611625076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027007776545360683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014143226854503154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8761833906173706, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13740, + "weighted_flow_matching_loss": 0.014040630217641592, + "weighted_rkd_loss": 0.01239203317090869 + }, + { + "epoch": 4.000581902822229, + "flow_matching_loss": 0.014190183766186238, + "grad_norm": 0.17830590903759003, + "learning_rate": 2.7443725067749098e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006103586871176958, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023884911788627505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014595665037631989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8760108649730682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13750, + "weighted_flow_matching_loss": 0.014190183766186238, + "weighted_rkd_loss": 0.012785960920155048 + }, + { + "epoch": 4.003491416933372, + "flow_matching_loss": 0.01322759259492159, + "grad_norm": 0.18192456662654877, + "learning_rate": 2.7439106810947923e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006787444651126862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002966955001465976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016541844327002765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8758382499217987, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13760, + "weighted_flow_matching_loss": 0.01322759259492159, + "weighted_rkd_loss": 0.014487916603684425 + }, + { + "epoch": 4.0064009310445154, + "flow_matching_loss": 0.01407631728798151, + "grad_norm": 0.20991043746471405, + "learning_rate": 2.7434484775474532e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006433366099372506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027864798437803985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015653212182223798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8756655275821685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13770, + "weighted_flow_matching_loss": 0.01407631728798151, + "weighted_rkd_loss": 0.013706927746534347 + }, + { + "epoch": 4.009310445155659, + "flow_matching_loss": 0.014647805225104093, + "grad_norm": 0.196796715259552, + "learning_rate": 2.7429858962732984e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006188379600644112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027724422281607984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015149201452732085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8754926919937134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13780, + "weighted_flow_matching_loss": 0.014647805225104093, + "weighted_rkd_loss": 0.01326297940686345 + }, + { + "epoch": 4.012219959266803, + "flow_matching_loss": 0.014455722365528344, + "grad_norm": 0.1755574345588684, + "learning_rate": 2.7425229374128475e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006609672820195556, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031558996997773647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016375245340168477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8753197848796844, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13790, + "weighted_flow_matching_loss": 0.014455722365528344, + "weighted_rkd_loss": 0.014333559945225716 + }, + { + "epoch": 4.015129473377946, + "flow_matching_loss": 0.01442234991118312, + "grad_norm": 0.20214776694774628, + "learning_rate": 2.742059601106735e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006892803963273764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003217347082681954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017002954985946415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8751467525959015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13800, + "weighted_flow_matching_loss": 0.01442234991118312, + "weighted_rkd_loss": 0.014879995584487915 + }, + { + "epoch": 4.018038987489089, + "flow_matching_loss": 0.014878661278635264, + "grad_norm": 0.2017940878868103, + "learning_rate": 2.74159588749571e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006576033774763346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028516627615317704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01600373052060604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8749736249446869, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13810, + "weighted_flow_matching_loss": 0.014878661278635264, + "weighted_rkd_loss": 0.014002892468124629 + }, + { + "epoch": 4.020948501600233, + "flow_matching_loss": 0.013837584946304559, + "grad_norm": 0.18149197101593018, + "learning_rate": 2.7411317967206373e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064742909744381905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030233528930693867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015971934888511896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8748003959655761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13820, + "weighted_flow_matching_loss": 0.013837584946304559, + "weighted_rkd_loss": 0.013972231652587652 + }, + { + "epoch": 4.023858015711376, + "flow_matching_loss": 0.01563668381422758, + "grad_norm": 0.18534086644649506, + "learning_rate": 2.7406673289224945e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006475591799244285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003252145671285689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01620332933962345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8746270596981048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13830, + "weighted_flow_matching_loss": 0.01563668381422758, + "weighted_rkd_loss": 0.01417182944715023 + }, + { + "epoch": 4.02676752982252, + "flow_matching_loss": 0.014486365113407373, + "grad_norm": 0.20507757365703583, + "learning_rate": 2.740202484242376e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00652193040587008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026318078860640528, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01567566879093647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8744536221027375, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13840, + "weighted_flow_matching_loss": 0.014486365113407373, + "weighted_rkd_loss": 0.013707688450813294 + }, + { + "epoch": 4.029677043933663, + "flow_matching_loss": 0.014393532369285821, + "grad_norm": 0.22719956934452057, + "learning_rate": 2.7397372628214873e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067369877360761166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003366678790189326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01684065433219075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8742800772190094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13850, + "weighted_flow_matching_loss": 0.014393532369285821, + "weighted_rkd_loss": 0.014723420049995184 + }, + { + "epoch": 4.032586558044806, + "flow_matching_loss": 0.015289124939590693, + "grad_norm": 0.20502741634845734, + "learning_rate": 2.739271664801152e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006987806409597397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034004827379249037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017376095708459615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8741064310073853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13860, + "weighted_flow_matching_loss": 0.015289124939590693, + "weighted_rkd_loss": 0.015188564080744982 + }, + { + "epoch": 4.03549607215595, + "flow_matching_loss": 0.014654884394258261, + "grad_norm": 0.2075023204088211, + "learning_rate": 2.738805690322806e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006561977462843061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002903436473570764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016027391143143176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8739326775074006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13870, + "weighted_flow_matching_loss": 0.014654884394258261, + "weighted_rkd_loss": 0.014006833918392659 + }, + { + "epoch": 4.038405586267094, + "flow_matching_loss": 0.01612782748416066, + "grad_norm": 0.19035285711288452, + "learning_rate": 2.7383393395280004e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006542742997407913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002870105882175267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01595559185370803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8737588286399841, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13880, + "weighted_flow_matching_loss": 0.01612782748416066, + "weighted_rkd_loss": 0.01394134871661663 + }, + { + "epoch": 4.041315100378237, + "flow_matching_loss": 0.01289679342880845, + "grad_norm": 0.21004687249660492, + "learning_rate": 2.7378726125584e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064973623491823675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002909117075614631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015903841704130173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8735848665237427, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13890, + "weighted_flow_matching_loss": 0.01289679342880845, + "weighted_rkd_loss": 0.013893381413072347 + }, + { + "epoch": 4.04422461448938, + "flow_matching_loss": 0.012590421456843614, + "grad_norm": 0.20076332986354828, + "learning_rate": 2.7374055095557832e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006203001132234931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026728131459094586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015078815445303918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8734108209609985, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13900, + "weighted_flow_matching_loss": 0.012590421456843614, + "weighted_rkd_loss": 0.013169972784817218 + }, + { + "epoch": 4.047134128600524, + "flow_matching_loss": 0.014814625214785337, + "grad_norm": 0.21120218932628632, + "learning_rate": 2.7369380306620446e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006852925475686789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030002017272636296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01670605270192027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8732366681098938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13910, + "weighted_flow_matching_loss": 0.014814625214785337, + "weighted_rkd_loss": 0.014588337298482657 + }, + { + "epoch": 4.0500436427116675, + "flow_matching_loss": 0.01620271522551775, + "grad_norm": 0.204387366771698, + "learning_rate": 2.7364701760191923e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00643144971691072, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003090260992757976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01595316044986248, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8730624139308929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13920, + "weighted_flow_matching_loss": 0.01620271522551775, + "weighted_rkd_loss": 0.01392820430919528 + }, + { + "epoch": 4.05295315682281, + "flow_matching_loss": 0.016682132519781588, + "grad_norm": 0.2598475217819214, + "learning_rate": 2.7360019457693468e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006877093575894833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032765476033091547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017030734568834305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8728880405426025, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13930, + "weighted_flow_matching_loss": 0.016682132519781588, + "weighted_rkd_loss": 0.014865959249436855 + }, + { + "epoch": 4.055862670933954, + "flow_matching_loss": 0.01377133084461093, + "grad_norm": 0.17248131334781647, + "learning_rate": 2.7355333400547457e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006715591251850128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002969183446839452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01640036581084132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.872713589668274, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13940, + "weighted_flow_matching_loss": 0.01377133084461093, + "weighted_rkd_loss": 0.014312782883644104 + }, + { + "epoch": 4.058772185045098, + "flow_matching_loss": 0.01488434299826622, + "grad_norm": 0.16819536685943604, + "learning_rate": 2.7350643590177378e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006358477612957358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002968115755356848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01568507105112076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8725390136241913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13950, + "weighted_flow_matching_loss": 0.01488434299826622, + "weighted_rkd_loss": 0.013685806281864643 + }, + { + "epoch": 4.0616816991562406, + "flow_matching_loss": 0.013673352170735597, + "grad_norm": 0.14826489984989166, + "learning_rate": 2.7345950028007878e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00656620068475604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029867565724998714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016119157895445824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8723643481731415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13960, + "weighted_flow_matching_loss": 0.013673352170735597, + "weighted_rkd_loss": 0.01406176295131445 + }, + { + "epoch": 4.064591213267384, + "flow_matching_loss": 0.016606531012803315, + "grad_norm": 0.21500952541828156, + "learning_rate": 2.7341252715464736e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006084023136645555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002542414981871843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014710461162030697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8721895754337311, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13970, + "weighted_flow_matching_loss": 0.016606531012803315, + "weighted_rkd_loss": 0.012830276694148778 + }, + { + "epoch": 4.067500727378528, + "flow_matching_loss": 0.014064575824886561, + "grad_norm": 0.15598958730697632, + "learning_rate": 2.7336551653974868e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005674938438460231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025887469295412303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013938624039292335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8720147013664246, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13980, + "weighted_flow_matching_loss": 0.014064575824886561, + "weighted_rkd_loss": 0.012154731526970863 + }, + { + "epoch": 4.070410241489671, + "flow_matching_loss": 0.013776667881757021, + "grad_norm": 0.19906212389469147, + "learning_rate": 2.733184684496634e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00720864087343216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033941653091460466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017811447102576494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8718397319316864, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13990, + "weighted_flow_matching_loss": 0.013776667881757021, + "weighted_rkd_loss": 0.015528758987784386 + }, + { + "epoch": 4.0733197556008145, + "flow_matching_loss": 0.015874146576970817, + "grad_norm": 0.25527334213256836, + "learning_rate": 2.7327138289868337e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006357868760824203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003005790524184704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01572152813896537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8716646611690522, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14000, + "weighted_flow_matching_loss": 0.015874146576970817, + "weighted_rkd_loss": 0.013703857455402613 + }, + { + "epoch": 4.076229269711958, + "flow_matching_loss": 0.014510011672973633, + "grad_norm": 0.19698689877986908, + "learning_rate": 2.7322425990111208e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00636241021566093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029068590374663472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01563167991116643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8714894652366638, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14010, + "weighted_flow_matching_loss": 0.014510011672973633, + "weighted_rkd_loss": 0.013622861448675394 + }, + { + "epoch": 4.079138783823102, + "flow_matching_loss": 0.015369166061282158, + "grad_norm": 0.1844555288553238, + "learning_rate": 2.7317709947126416e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006759304041042924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027828664286062123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016301474813371897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8713141977787018, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14020, + "weighted_flow_matching_loss": 0.015369166061282158, + "weighted_rkd_loss": 0.014203714672476053 + }, + { + "epoch": 4.082048297934245, + "flow_matching_loss": 0.014689155854284763, + "grad_norm": 0.2395295649766922, + "learning_rate": 2.731299016234657e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006682152347639203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002681942516937852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016046246979385614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8711388170719147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14030, + "weighted_flow_matching_loss": 0.014689155854284763, + "weighted_rkd_loss": 0.01397843100130558 + }, + { + "epoch": 4.084957812045388, + "flow_matching_loss": 0.015341269224882126, + "grad_norm": 0.18379557132720947, + "learning_rate": 2.730826663720542e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066874018404632805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003013543947599828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01638834774494171, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8709633350372314, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14040, + "weighted_flow_matching_loss": 0.015341269224882126, + "weighted_rkd_loss": 0.014273626171052456 + }, + { + "epoch": 4.087867326156532, + "flow_matching_loss": 0.014760674070566892, + "grad_norm": 0.20756573975086212, + "learning_rate": 2.730353937313785e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005891113262623549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024888986139558257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014271125104278326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8707877457141876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14050, + "weighted_flow_matching_loss": 0.014760674070566892, + "weighted_rkd_loss": 0.012427092902362347 + }, + { + "epoch": 4.090776840267675, + "flow_matching_loss": 0.014682666212320328, + "grad_norm": 0.1808401495218277, + "learning_rate": 2.729880837157987e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006313812965527177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002792167221195996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0154197933152318, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8706120610237121, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14060, + "weighted_flow_matching_loss": 0.014682666212320328, + "weighted_rkd_loss": 0.01342465654015541 + }, + { + "epoch": 4.093686354378819, + "flow_matching_loss": 0.012410435173660517, + "grad_norm": 0.18042190372943878, + "learning_rate": 2.729407363396864e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00571457501500845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002529875293839723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013959025405347348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.870436280965805, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14070, + "weighted_flow_matching_loss": 0.012410435173660517, + "weighted_rkd_loss": 0.012150434032082557 + }, + { + "epoch": 4.096595868489962, + "flow_matching_loss": 0.013393501564860345, + "grad_norm": 0.20076040923595428, + "learning_rate": 2.7289335161742444e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007316320482641458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003340128529816866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017972769495099784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.870260375738144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14080, + "weighted_flow_matching_loss": 0.013393501564860345, + "weighted_rkd_loss": 0.015640855208039282 + }, + { + "epoch": 4.099505382601105, + "flow_matching_loss": 0.014574253279715777, + "grad_norm": 0.21818923950195312, + "learning_rate": 2.7284592956340707e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006694318866357208, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003143759537488222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016532397270202635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8700843870639801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14090, + "weighted_flow_matching_loss": 0.014574253279715777, + "weighted_rkd_loss": 0.014384470041841268 + }, + { + "epoch": 4.102414896712249, + "flow_matching_loss": 0.016279044654220344, + "grad_norm": 0.23718774318695068, + "learning_rate": 2.727984701920399e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005889365309849381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026089033111929894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014387633837759494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8699082970619202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14100, + "weighted_flow_matching_loss": 0.016279044654220344, + "weighted_rkd_loss": 0.012515911646187306 + }, + { + "epoch": 4.105324410823393, + "flow_matching_loss": 0.013645821157842875, + "grad_norm": 0.17776857316493988, + "learning_rate": 2.7275097351773977e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006465679360553622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028569098562002183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015788268484175205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8697321057319641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14110, + "weighted_flow_matching_loss": 0.013645821157842875, + "weighted_rkd_loss": 0.013731556851416827 + }, + { + "epoch": 4.108233924934536, + "flow_matching_loss": 0.013215458486229181, + "grad_norm": 0.19889914989471436, + "learning_rate": 2.727034395549349e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006150894099846482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002792426268570125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015094214491546154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8695558071136474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14120, + "weighted_flow_matching_loss": 0.013215458486229181, + "weighted_rkd_loss": 0.013125348184257746 + }, + { + "epoch": 4.111143439045679, + "flow_matching_loss": 0.014889538008719682, + "grad_norm": 0.15852545201778412, + "learning_rate": 2.726558683180649e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072451465297490355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031085676746442914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0175988607108593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8693794012069702, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14130, + "weighted_flow_matching_loss": 0.014889538008719682, + "weighted_rkd_loss": 0.015300089493393898 + }, + { + "epoch": 4.114052953156823, + "flow_matching_loss": 0.013665339723229409, + "grad_norm": 0.15096789598464966, + "learning_rate": 2.7260825982158067e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005814858758822083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002501266496255994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014130983967334032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8692029178142547, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14140, + "weighted_flow_matching_loss": 0.013665339723229409, + "weighted_rkd_loss": 0.012282716576009989 + }, + { + "epoch": 4.1169624672679666, + "flow_matching_loss": 0.014648384228348733, + "grad_norm": 0.1705772429704666, + "learning_rate": 2.725606140799444e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006734036514535547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029350814409554006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016403154563158752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8690263092517853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14150, + "weighted_flow_matching_loss": 0.014648384228348733, + "weighted_rkd_loss": 0.01425487855449319 + }, + { + "epoch": 4.119871981379109, + "flow_matching_loss": 0.014541223365813493, + "grad_norm": 0.21527647972106934, + "learning_rate": 2.7251293110762957e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007097651297226548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031246050726622342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017319907899945974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8688496053218842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14160, + "weighted_flow_matching_loss": 0.014541223365813493, + "weighted_rkd_loss": 0.015048349741846323 + }, + { + "epoch": 4.122781495490253, + "flow_matching_loss": 0.013966955151408911, + "grad_norm": 0.20204991102218628, + "learning_rate": 2.7246521091912106e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069769600871950384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003310277545824647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017264197673648597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8686728060245514, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14170, + "weighted_flow_matching_loss": 0.013966955151408911, + "weighted_rkd_loss": 0.014996964018791914 + }, + { + "epoch": 4.125691009601397, + "flow_matching_loss": 0.01311950795352459, + "grad_norm": 0.1589774340391159, + "learning_rate": 2.7241745352891497e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00643345732241869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002869200357235968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015736114978790284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8684958994388581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14180, + "weighted_flow_matching_loss": 0.01311950795352459, + "weighted_rkd_loss": 0.01366685824468732 + }, + { + "epoch": 4.12860052371254, + "flow_matching_loss": 0.013787433598190546, + "grad_norm": 0.19819720089435577, + "learning_rate": 2.7236965895151874e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006894617807120085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003136621997691691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016925857681781054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8683188855648041, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14190, + "weighted_flow_matching_loss": 0.013787433598190546, + "weighted_rkd_loss": 0.01469706855714321 + }, + { + "epoch": 4.131510037823683, + "flow_matching_loss": 0.014145647455006839, + "grad_norm": 0.16733962297439575, + "learning_rate": 2.7232182720145105e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006410671025514602, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002939606411382556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01576094850897789, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8681417942047119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14200, + "weighted_flow_matching_loss": 0.014145647455006839, + "weighted_rkd_loss": 0.013682803139090538 + }, + { + "epoch": 4.134419551934827, + "flow_matching_loss": 0.014699246454983949, + "grad_norm": 0.16379225254058838, + "learning_rate": 2.7227395829324208e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006675381865352392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002981799829285592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016332563664764165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8679645955562592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14210, + "weighted_flow_matching_loss": 0.014699246454983949, + "weighted_rkd_loss": 0.01417624605819583 + }, + { + "epoch": 4.137329066045971, + "flow_matching_loss": 0.014057976938784123, + "grad_norm": 0.21756577491760254, + "learning_rate": 2.7222605224143298e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005648671044036746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002337674144655466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013635016232728957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8677872896194458, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14220, + "weighted_flow_matching_loss": 0.014057976938784123, + "weighted_rkd_loss": 0.011832238174974918 + }, + { + "epoch": 4.1402385801571135, + "flow_matching_loss": 0.012780234962701798, + "grad_norm": 0.16025251150131226, + "learning_rate": 2.721781090605764e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006359263462945819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028404273907653987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015558954328298569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8676098704338073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14230, + "weighted_flow_matching_loss": 0.012780234962701798, + "weighted_rkd_loss": 0.013499111775308848 + }, + { + "epoch": 4.143148094268257, + "flow_matching_loss": 0.013573873229324818, + "grad_norm": 0.2104230523109436, + "learning_rate": 2.721301287652362e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060151231475174425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002671593346167356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014701839722692966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8674323678016662, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14240, + "weighted_flow_matching_loss": 0.013573873229324818, + "weighted_rkd_loss": 0.012752892263233662 + }, + { + "epoch": 4.146057608379401, + "flow_matching_loss": 0.01484726769849658, + "grad_norm": 0.19518080353736877, + "learning_rate": 2.7208211136998758e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006385029200464487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031717446632683276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015941802971065046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8672547578811646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14250, + "weighted_flow_matching_loss": 0.01484726769849658, + "weighted_rkd_loss": 0.013825625833123922 + }, + { + "epoch": 4.148967122490544, + "flow_matching_loss": 0.014480877481400967, + "grad_norm": 0.20738956332206726, + "learning_rate": 2.7203405688941693e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006020198646001518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029403183260001244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014980715606361628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8670770525932312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14260, + "weighted_flow_matching_loss": 0.014480877481400967, + "weighted_rkd_loss": 0.012989369034767152 + }, + { + "epoch": 4.1518766366016875, + "flow_matching_loss": 0.01397413806989789, + "grad_norm": 0.21754100918769836, + "learning_rate": 2.719859653381219e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006370639288797974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002812106430064887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015553384926170111, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8668992340564727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14270, + "weighted_flow_matching_loss": 0.01397413806989789, + "weighted_rkd_loss": 0.013483214750885963 + }, + { + "epoch": 4.154786150712831, + "flow_matching_loss": 0.014124175626784563, + "grad_norm": 0.1938628852367401, + "learning_rate": 2.719378367307114e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006507106870412826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028700596187263725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015884273406118154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8667213201522828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14280, + "weighted_flow_matching_loss": 0.014124175626784563, + "weighted_rkd_loss": 0.013767246436327697 + }, + { + "epoch": 4.157695664823974, + "flow_matching_loss": 0.014400944206863641, + "grad_norm": 0.21616919338703156, + "learning_rate": 2.7188967108180574e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007066295342519879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003132678032852709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01726526888087392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.86654332280159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14290, + "weighted_flow_matching_loss": 0.014400944206863641, + "weighted_rkd_loss": 0.014961045701056719 + }, + { + "epoch": 4.160605178935118, + "flow_matching_loss": 0.014990229159593582, + "grad_norm": 0.17017939686775208, + "learning_rate": 2.7184146840603634e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006099223345518112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028065063757821917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015004952996969223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8663652062416076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14300, + "weighted_flow_matching_loss": 0.014990229159593582, + "weighted_rkd_loss": 0.012999779917299748 + }, + { + "epoch": 4.163514693046261, + "flow_matching_loss": 0.014161887019872666, + "grad_norm": 0.2057117372751236, + "learning_rate": 2.7179322871804584e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006466813432052731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028382940450683235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015771920885890724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8661869943141938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14310, + "weighted_flow_matching_loss": 0.014161887019872666, + "weighted_rkd_loss": 0.013661406654864549 + }, + { + "epoch": 4.166424207157405, + "flow_matching_loss": 0.013788511231541633, + "grad_norm": 0.21064844727516174, + "learning_rate": 2.717449520324882e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006574261886999011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029385143890976904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016087037976831198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8660086810588836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14320, + "weighted_flow_matching_loss": 0.013788511231541633, + "weighted_rkd_loss": 0.013931440282613039 + }, + { + "epoch": 4.169333721268548, + "flow_matching_loss": 0.013641471043229103, + "grad_norm": 0.19690223038196564, + "learning_rate": 2.7169663836402864e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006262489920482039, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002717138477601111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01524211848154664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8658302664756775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14330, + "weighted_flow_matching_loss": 0.013641471043229103, + "weighted_rkd_loss": 0.013197108823806047 + }, + { + "epoch": 4.172243235379692, + "flow_matching_loss": 0.012709395308047534, + "grad_norm": 0.18109357357025146, + "learning_rate": 2.7164828772734352e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005550240445882082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024109854362905025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013511466328054667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8656517446041108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14340, + "weighted_flow_matching_loss": 0.012709395308047534, + "weighted_rkd_loss": 0.01169627271592617 + }, + { + "epoch": 4.175152749490835, + "flow_matching_loss": 0.01594475731253624, + "grad_norm": 0.17792168259620667, + "learning_rate": 2.7159990013712057e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058465668931603435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026394327986054124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014332566503435373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8654731392860413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14350, + "weighted_flow_matching_loss": 0.01594475731253624, + "weighted_rkd_loss": 0.012404537200927735 + }, + { + "epoch": 4.178062263601978, + "flow_matching_loss": 0.015079795476049184, + "grad_norm": 0.22435703873634338, + "learning_rate": 2.7155147560805863e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005987716373056173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00252910841954872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014504541270434856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8652944266796112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14360, + "weighted_flow_matching_loss": 0.015079795476049184, + "weighted_rkd_loss": 0.012550696730613708 + }, + { + "epoch": 4.180971777713122, + "flow_matching_loss": 0.014704492036253214, + "grad_norm": 0.19332773983478546, + "learning_rate": 2.7150301415486777e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005804116511717439, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00259386288234964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014202095940709114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.865115612745285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14370, + "weighted_flow_matching_loss": 0.014704492036253214, + "weighted_rkd_loss": 0.012286446709185838 + }, + { + "epoch": 4.183881291824266, + "flow_matching_loss": 0.015113365557044744, + "grad_norm": 0.19112952053546906, + "learning_rate": 2.7145451579226932e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006601259624585509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002984584216028452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016187103651463985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8649366915225982, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14380, + "weighted_flow_matching_loss": 0.015113365557044744, + "weighted_rkd_loss": 0.014000900741666555 + }, + { + "epoch": 4.186790805935408, + "flow_matching_loss": 0.01442566690966487, + "grad_norm": 0.18647456169128418, + "learning_rate": 2.7140598053499577e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006436797836795449, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002884323615580797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015757919009774923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8647576749324799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14390, + "weighted_flow_matching_loss": 0.01442566690966487, + "weighted_rkd_loss": 0.013626777846366168 + }, + { + "epoch": 4.189700320046552, + "flow_matching_loss": 0.015357061196118594, + "grad_norm": 0.20622052252292633, + "learning_rate": 2.7135740839779087e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006507937470450998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002874519699253142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01589039508253336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8645785629749299, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14400, + "weighted_flow_matching_loss": 0.015357061196118594, + "weighted_rkd_loss": 0.013738459441810846 + }, + { + "epoch": 4.192609834157696, + "flow_matching_loss": 0.015174979250878096, + "grad_norm": 0.20023363828659058, + "learning_rate": 2.7130879939540957e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056344364304095505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024752627359703183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013744135573506355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8643993556499481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14410, + "weighted_flow_matching_loss": 0.015174979250878096, + "weighted_rkd_loss": 0.011880424618721009 + }, + { + "epoch": 4.195519348268839, + "flow_matching_loss": 0.015463524777442218, + "grad_norm": 0.1969889998435974, + "learning_rate": 2.7126015354261798e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006365339271724224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028197002131491898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01555037871003151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8642200410366059, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14420, + "weighted_flow_matching_loss": 0.015463524777442218, + "weighted_rkd_loss": 0.013438980001956225 + }, + { + "epoch": 4.198428862379982, + "flow_matching_loss": 0.014104383531957865, + "grad_norm": 0.22042497992515564, + "learning_rate": 2.712114708541934e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006528548523783683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028907442232593895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015947841107845306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8640406250953674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14430, + "weighted_flow_matching_loss": 0.014104383531957865, + "weighted_rkd_loss": 0.013779522199183703 + }, + { + "epoch": 4.201338376491126, + "flow_matching_loss": 0.014915749710053206, + "grad_norm": 0.17402784526348114, + "learning_rate": 2.7116275134492445e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006275945901870727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028253834694623946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015377275180071592, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8638611137866974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14440, + "weighted_flow_matching_loss": 0.014915749710053206, + "weighted_rkd_loss": 0.013283873535692691 + }, + { + "epoch": 4.20424789060227, + "flow_matching_loss": 0.016377331409603357, + "grad_norm": 0.16023069620132446, + "learning_rate": 2.7111399502961073e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006268896674737334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026886679930612444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015226461365818978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8636814892292023, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14450, + "weighted_flow_matching_loss": 0.016377331409603357, + "weighted_rkd_loss": 0.013150869961827993 + }, + { + "epoch": 4.207157404713413, + "flow_matching_loss": 0.013588189985603094, + "grad_norm": 0.18715202808380127, + "learning_rate": 2.710652019230631e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00581685146316886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024885366903617976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014122239872813224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8635017812252045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14460, + "weighted_flow_matching_loss": 0.013588189985603094, + "weighted_rkd_loss": 0.012194526474922895 + }, + { + "epoch": 4.210066918824556, + "flow_matching_loss": 0.0141908529214561, + "grad_norm": 0.1528552770614624, + "learning_rate": 2.710163720401037e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060453638434410095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026103143813088537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014701042138040066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8633219718933105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14470, + "weighted_flow_matching_loss": 0.0141908529214561, + "weighted_rkd_loss": 0.012691780552268028 + }, + { + "epoch": 4.2129764329357, + "flow_matching_loss": 0.013415229599922895, + "grad_norm": 0.14764170348644257, + "learning_rate": 2.709675053955658e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005929237464442849, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025460361619479956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014404511079192162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8631420612335206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14480, + "weighted_flow_matching_loss": 0.013415229599922895, + "weighted_rkd_loss": 0.012433183286339045 + }, + { + "epoch": 4.215885947046843, + "flow_matching_loss": 0.014698317740112543, + "grad_norm": 0.1881900280714035, + "learning_rate": 2.7091860200429362e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005618794355541468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024913552682846785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013728944025933743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8629620432853699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14490, + "weighted_flow_matching_loss": 0.014698317740112543, + "weighted_rkd_loss": 0.011847517918795347 + }, + { + "epoch": 4.2187954611579865, + "flow_matching_loss": 0.015270031429827214, + "grad_norm": 0.18333306908607483, + "learning_rate": 2.7086966188114285e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006078970246016979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002896717842668295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015054658427834511, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8627819299697876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14500, + "weighted_flow_matching_loss": 0.015270031429827214, + "weighted_rkd_loss": 0.012988941557705403 + }, + { + "epoch": 4.22170497526913, + "flow_matching_loss": 0.013874129951000213, + "grad_norm": 0.1819087415933609, + "learning_rate": 2.708206850409802e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006331996992230415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027739066630601884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015437900274991988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8626017153263092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14510, + "weighted_flow_matching_loss": 0.013874129951000213, + "weighted_rkd_loss": 0.013316784892231226 + }, + { + "epoch": 4.224614489380274, + "flow_matching_loss": 0.013660066854208707, + "grad_norm": 0.21739733219146729, + "learning_rate": 2.707716714986835e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006837288243696093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032809693133458496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016955545730888845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8624214053153991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14520, + "weighted_flow_matching_loss": 0.013660066854208707, + "weighted_rkd_loss": 0.014622837770730257 + }, + { + "epoch": 4.227524003491417, + "flow_matching_loss": 0.016033711191266776, + "grad_norm": 0.18350186944007874, + "learning_rate": 2.707226212691418e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007035035640001297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003019133280031383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01708920458331704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8622409999370575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14530, + "weighted_flow_matching_loss": 0.016033711191266776, + "weighted_rkd_loss": 0.014734970405697823 + }, + { + "epoch": 4.2304335176025605, + "flow_matching_loss": 0.01602762071415782, + "grad_norm": 0.19202207028865814, + "learning_rate": 2.7067353436725525e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005734307644888759, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00280946537386626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0142780807800591, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8620604932308197, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14540, + "weighted_flow_matching_loss": 0.01602762071415782, + "weighted_rkd_loss": 0.012308561243116856 + }, + { + "epoch": 4.233343031713704, + "flow_matching_loss": 0.014348071534186602, + "grad_norm": 0.16898326575756073, + "learning_rate": 2.706244108079352e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005872398498468101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002694338117726147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014439135044813155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8618798851966858, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14550, + "weighted_flow_matching_loss": 0.014348071534186602, + "weighted_rkd_loss": 0.012444750918075443 + }, + { + "epoch": 4.236252545824847, + "flow_matching_loss": 0.014592116139829159, + "grad_norm": 0.20889373123645782, + "learning_rate": 2.7057525060610396e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064080316107720135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029943140340037643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015810377150774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8616991758346557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14560, + "weighted_flow_matching_loss": 0.014592116139829159, + "weighted_rkd_loss": 0.013623730186372996 + }, + { + "epoch": 4.239162059935991, + "flow_matching_loss": 0.012859457451850177, + "grad_norm": 0.16760528087615967, + "learning_rate": 2.7052605377669527e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00659947763197124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028699158574454486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01606887113302946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8615183651447296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14570, + "weighted_flow_matching_loss": 0.012859457451850177, + "weighted_rkd_loss": 0.013843712210655213 + }, + { + "epoch": 4.242071574047134, + "flow_matching_loss": 0.013933352567255497, + "grad_norm": 0.18250921368598938, + "learning_rate": 2.704768203346537e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006449032109230757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002966128138359636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015864192601293327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8613374590873718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14580, + "weighted_flow_matching_loss": 0.013933352567255497, + "weighted_rkd_loss": 0.013664522301405668 + }, + { + "epoch": 4.244981088158277, + "flow_matching_loss": 0.015014952514320612, + "grad_norm": 0.16632625460624695, + "learning_rate": 2.7042755029493513e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00591844622977078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027693423442542555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0146062346175313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8611564576625824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14590, + "weighted_flow_matching_loss": 0.015014952514320612, + "weighted_rkd_loss": 0.012578221969306469 + }, + { + "epoch": 4.247890602269421, + "flow_matching_loss": 0.015038322750478983, + "grad_norm": 0.1747918725013733, + "learning_rate": 2.703782436725065e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005610995553433895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002552515163552016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013774506282061338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8609753549098969, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14600, + "weighted_flow_matching_loss": 0.015038322750478983, + "weighted_rkd_loss": 0.01185954650864005 + }, + { + "epoch": 4.250800116380565, + "flow_matching_loss": 0.014393153041601181, + "grad_norm": 0.18614277243614197, + "learning_rate": 2.7032890048234585e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006384727219119668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029194582253694533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01568891266360879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8607941508293152, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14610, + "weighted_flow_matching_loss": 0.014393153041601181, + "weighted_rkd_loss": 0.013505003042519092 + }, + { + "epoch": 4.2537096304917075, + "flow_matching_loss": 0.014289462380111218, + "grad_norm": 0.2051873356103897, + "learning_rate": 2.702795207394423e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058135062921792265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002610631287097931, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014237643871456384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8606128454208374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14620, + "weighted_flow_matching_loss": 0.014289462380111218, + "weighted_rkd_loss": 0.012253154721111058 + }, + { + "epoch": 4.256619144602851, + "flow_matching_loss": 0.013350775092840194, + "grad_norm": 0.15227936208248138, + "learning_rate": 2.702301044587962e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005729789193719625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025872963480651377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01404687473550439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8604314446449279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14630, + "weighted_flow_matching_loss": 0.013350775092840194, + "weighted_rkd_loss": 0.012086330261081458 + }, + { + "epoch": 4.259528658713995, + "flow_matching_loss": 0.014579354040324687, + "grad_norm": 0.17869800329208374, + "learning_rate": 2.7018065165541885e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005993198323994875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024918375071138143, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014478234201669693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8602499425411224, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14640, + "weighted_flow_matching_loss": 0.014579354040324687, + "weighted_rkd_loss": 0.01245482238009572 + }, + { + "epoch": 4.262438172825139, + "flow_matching_loss": 0.013913271203637122, + "grad_norm": 0.1798238754272461, + "learning_rate": 2.701311623443327e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005854495521634817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025996669894084333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014308658055961132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8600683510303497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14650, + "weighted_flow_matching_loss": 0.013913271203637122, + "weighted_rkd_loss": 0.012306425720453262 + }, + { + "epoch": 4.265347686936281, + "flow_matching_loss": 0.013545748591423035, + "grad_norm": 0.22157812118530273, + "learning_rate": 2.700816365405713e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00687769902870059, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003243510890752077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016998908761888743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8598866522312164, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14660, + "weighted_flow_matching_loss": 0.013545748591423035, + "weighted_rkd_loss": 0.014617113582789897 + }, + { + "epoch": 4.268257201047425, + "flow_matching_loss": 0.014316289871931075, + "grad_norm": 0.21330778300762177, + "learning_rate": 2.7003207425917926e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005801290133967995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025279478868469596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014130527898669244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8597048580646515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14670, + "weighted_flow_matching_loss": 0.014316289871931075, + "weighted_rkd_loss": 0.012148083187639713 + }, + { + "epoch": 4.271166715158569, + "flow_matching_loss": 0.015296630654484033, + "grad_norm": 0.20833030343055725, + "learning_rate": 2.6998247551521238e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062213831581175326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028480379609391094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015290804021060467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.859522956609726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14680, + "weighted_flow_matching_loss": 0.015296630654484033, + "weighted_rkd_loss": 0.0131427351385355 + }, + { + "epoch": 4.274076229269712, + "flow_matching_loss": 0.01489145988598466, + "grad_norm": 0.17067116498947144, + "learning_rate": 2.6993284032373732e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005843938840553164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027334277052432298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014421305432915688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8593409657478333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14690, + "weighted_flow_matching_loss": 0.01489145988598466, + "weighted_rkd_loss": 0.01239282190799713 + }, + { + "epoch": 4.276985743380855, + "flow_matching_loss": 0.015745624899864197, + "grad_norm": 0.1860392540693283, + "learning_rate": 2.6988316869983208e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066381738055497404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003068816172890365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016345163993537425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8591588735580444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14700, + "weighted_flow_matching_loss": 0.015745624899864197, + "weighted_rkd_loss": 0.014043155312538146 + }, + { + "epoch": 4.279895257491999, + "flow_matching_loss": 0.01408538008108735, + "grad_norm": 0.26287850737571716, + "learning_rate": 2.698334606585855e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064700087532401085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027985960128717124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015738613437861204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8589766800403595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14710, + "weighted_flow_matching_loss": 0.01408538008108735, + "weighted_rkd_loss": 0.013519050646573305 + }, + { + "epoch": 4.282804771603142, + "flow_matching_loss": 0.01380145624279976, + "grad_norm": 0.18398542702198029, + "learning_rate": 2.6978371621509757e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00639978377148509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028394623077474535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015639029908925295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8587943911552429, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14720, + "weighted_flow_matching_loss": 0.01380145624279976, + "weighted_rkd_loss": 0.013430769089609385 + }, + { + "epoch": 4.285714285714286, + "flow_matching_loss": 0.015333693381398917, + "grad_norm": 0.18077009916305542, + "learning_rate": 2.697339353844794e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006519623193889856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028869424480944873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01592618878930807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8586120009422302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14730, + "weighted_flow_matching_loss": 0.015333693381398917, + "weighted_rkd_loss": 0.01367451436817646 + }, + { + "epoch": 4.288623799825429, + "flow_matching_loss": 0.013994960859417916, + "grad_norm": 0.18894924223423004, + "learning_rate": 2.6968411818185303e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005996434157714248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028004535008221863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014793321955949068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8584295213222504, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14740, + "weighted_flow_matching_loss": 0.013994960859417916, + "weighted_rkd_loss": 0.01269897324964404 + }, + { + "epoch": 4.291533313936573, + "flow_matching_loss": 0.015057269856333733, + "grad_norm": 0.17951838672161102, + "learning_rate": 2.6963426462235167e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006122949346899986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002755660132970661, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01500155869871378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8582469403743744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14750, + "weighted_flow_matching_loss": 0.015057269856333733, + "weighted_rkd_loss": 0.01287502096965909 + }, + { + "epoch": 4.294442828047716, + "flow_matching_loss": 0.015874965488910674, + "grad_norm": 0.20097286999225616, + "learning_rate": 2.6958437472111947e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065729821566492316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002846743422560394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01599270775914192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8580642580986023, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14760, + "weighted_flow_matching_loss": 0.015874965488910674, + "weighted_rkd_loss": 0.01372279180213809 + }, + { + "epoch": 4.2973523421588595, + "flow_matching_loss": 0.013105198089033366, + "grad_norm": 0.1844671666622162, + "learning_rate": 2.695344484933116e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005949553335085511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002689095714595169, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014588202256709338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8578814744949341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14770, + "weighted_flow_matching_loss": 0.013105198089033366, + "weighted_rkd_loss": 0.012514942605048419 + }, + { + "epoch": 4.300261856270003, + "flow_matching_loss": 0.014542948082089424, + "grad_norm": 0.20652040839195251, + "learning_rate": 2.6948448595409443e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058614130597561594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002635635668411851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0143584618344903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8576985955238342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14780, + "weighted_flow_matching_loss": 0.014542948082089424, + "weighted_rkd_loss": 0.012315196869894861 + }, + { + "epoch": 4.303171370381146, + "flow_matching_loss": 0.014705585315823555, + "grad_norm": 0.16695569455623627, + "learning_rate": 2.694344871186453e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005406799400225282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002445614291355014, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013259213231503964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8575156211853028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14790, + "weighted_flow_matching_loss": 0.014705585315823555, + "weighted_rkd_loss": 0.011369945295155049 + }, + { + "epoch": 4.30608088449229, + "flow_matching_loss": 0.013813696336001158, + "grad_norm": 0.1612231433391571, + "learning_rate": 2.6938445200215237e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005889204423874616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024656454916112125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014244054444134235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8573325514793396, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14800, + "weighted_flow_matching_loss": 0.013813696336001158, + "weighted_rkd_loss": 0.012211890146136283 + }, + { + "epoch": 4.3089903986034335, + "flow_matching_loss": 0.014416618458926677, + "grad_norm": 0.18660375475883484, + "learning_rate": 2.6933438061981513e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068498064763844015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034866095753386616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017186222411692143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8571493744850158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14810, + "weighted_flow_matching_loss": 0.014416618458926677, + "weighted_rkd_loss": 0.01473104851320386 + }, + { + "epoch": 4.311899912714576, + "flow_matching_loss": 0.01494414210319519, + "grad_norm": 0.2551155686378479, + "learning_rate": 2.6928427298684395e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006575500639155507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030340081779286266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016185009572654962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8569661021232605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14820, + "weighted_flow_matching_loss": 0.01494414210319519, + "weighted_rkd_loss": 0.013869969919323922 + }, + { + "epoch": 4.31480942682572, + "flow_matching_loss": 0.013061403855681419, + "grad_norm": 0.18747888505458832, + "learning_rate": 2.6923412911846008e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00652301162481308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028536450583487747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015899668261408807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8567827403545379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14830, + "weighted_flow_matching_loss": 0.013061403855681419, + "weighted_rkd_loss": 0.013622596114873885 + }, + { + "epoch": 4.317718940936864, + "flow_matching_loss": 0.01362918270751834, + "grad_norm": 0.19316573441028595, + "learning_rate": 2.6918394902989604e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006626256695017219, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029838323360309007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016236345935612916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8565992653369904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14840, + "weighted_flow_matching_loss": 0.01362918270751834, + "weighted_rkd_loss": 0.013908033631742001 + }, + { + "epoch": 4.320628455048007, + "flow_matching_loss": 0.015257804654538631, + "grad_norm": 0.18660712242126465, + "learning_rate": 2.6913373273639508e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006260372838005424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002959650824777782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015480396524071693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8564157068729401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14850, + "weighted_flow_matching_loss": 0.015257804654538631, + "weighted_rkd_loss": 0.01325768493115902 + }, + { + "epoch": 4.32353796915915, + "flow_matching_loss": 0.01478106789290905, + "grad_norm": 0.16836905479431152, + "learning_rate": 2.690834802532117e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005370313161984086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024367621983401476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013177388440817594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8562320530414581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14860, + "weighted_flow_matching_loss": 0.01478106789290905, + "weighted_rkd_loss": 0.011282899975776672 + }, + { + "epoch": 4.326447483270294, + "flow_matching_loss": 0.015403585880994797, + "grad_norm": 0.22145432233810425, + "learning_rate": 2.6903319159561125e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006118197692558169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027258681831881402, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014962263684719802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8560482919216156, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14870, + "weighted_flow_matching_loss": 0.015403585880994797, + "weighted_rkd_loss": 0.012808425072580575 + }, + { + "epoch": 4.329356997381438, + "flow_matching_loss": 0.0139077203348279, + "grad_norm": 0.1716635823249817, + "learning_rate": 2.6898286677887005e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066201505716890095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002925575734116137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016165876761078835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8558644354343414, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14880, + "weighted_flow_matching_loss": 0.0139077203348279, + "weighted_rkd_loss": 0.013835838902741671 + }, + { + "epoch": 4.3322665114925805, + "flow_matching_loss": 0.014209140837192536, + "grad_norm": 0.18202704191207886, + "learning_rate": 2.689325058182755e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006509664701297879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002875534538179636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015894864033907653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8556804835796357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14890, + "weighted_flow_matching_loss": 0.014209140837192536, + "weighted_rkd_loss": 0.01360090933740139 + }, + { + "epoch": 4.335176025603724, + "flow_matching_loss": 0.014891103748232127, + "grad_norm": 0.17287299036979675, + "learning_rate": 2.6888210872912594e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058893266599625346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002513041556812823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014291694760322571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8554964244365693, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14900, + "weighted_flow_matching_loss": 0.014891103748232127, + "weighted_rkd_loss": 0.01222655326128006 + }, + { + "epoch": 4.338085539714868, + "flow_matching_loss": 0.014455372840166092, + "grad_norm": 0.20795203745365143, + "learning_rate": 2.688316755267306e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064192367251962425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027595874620601536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015598060935735703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8553122580051422, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14910, + "weighted_flow_matching_loss": 0.014455372840166092, + "weighted_rkd_loss": 0.013341237604618073 + }, + { + "epoch": 4.340995053826011, + "flow_matching_loss": 0.014393936190754176, + "grad_norm": 0.19068807363510132, + "learning_rate": 2.6878120622640988e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005769908986985683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002458146109711379, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01399796400219202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.855128037929535, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14920, + "weighted_flow_matching_loss": 0.014393936190754176, + "weighted_rkd_loss": 0.01197008229792118 + }, + { + "epoch": 4.343904567937154, + "flow_matching_loss": 0.01451462609693408, + "grad_norm": 0.1954079419374466, + "learning_rate": 2.6873070084349486e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005793029582127929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002480745199136436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014066804386675357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8549436926841736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14930, + "weighted_flow_matching_loss": 0.01451462609693408, + "weighted_rkd_loss": 0.012026283890008926 + }, + { + "epoch": 4.346814082048298, + "flow_matching_loss": 0.014600615575909614, + "grad_norm": 0.19665847718715668, + "learning_rate": 2.6868015939332788e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006694695027545094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003160405531525612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016549795493483545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8547592461109161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14940, + "weighted_flow_matching_loss": 0.014600615575909614, + "weighted_rkd_loss": 0.014146092720329762 + }, + { + "epoch": 4.349723596159441, + "flow_matching_loss": 0.014342368207871914, + "grad_norm": 0.2190856635570526, + "learning_rate": 2.68629581891262e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006296683195978403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002774587646126747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01536795413121581, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8545747101306915, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14950, + "weighted_flow_matching_loss": 0.014342368207871914, + "weighted_rkd_loss": 0.013132955878973007 + }, + { + "epoch": 4.352633110270585, + "flow_matching_loss": 0.013534333184361457, + "grad_norm": 0.16574150323867798, + "learning_rate": 2.685789683526614e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006299055041745305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027096391073428094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015307749155908823, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8543900668621063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14960, + "weighted_flow_matching_loss": 0.013534333184361457, + "weighted_rkd_loss": 0.013078754208981991 + }, + { + "epoch": 4.355542624381728, + "flow_matching_loss": 0.013801356218755246, + "grad_norm": 0.19491782784461975, + "learning_rate": 2.685283187929011e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006636502454057336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003220959100872278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01649396400898695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8542053580284119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14970, + "weighted_flow_matching_loss": 0.013801356218755246, + "weighted_rkd_loss": 0.014089303556829691 + }, + { + "epoch": 4.358452138492872, + "flow_matching_loss": 0.016346553433686496, + "grad_norm": 0.17611141502857208, + "learning_rate": 2.6847763322736713e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005848923069424927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029036857071332634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014601531997323036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8540205121040344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14980, + "weighted_flow_matching_loss": 0.016346553433686496, + "weighted_rkd_loss": 0.012470062542706727 + }, + { + "epoch": 4.361361652604015, + "flow_matching_loss": 0.014201536495238542, + "grad_norm": 0.18703630566596985, + "learning_rate": 2.684269116714564e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006273632729426026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028690831968560815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01541634863242507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8538355827331543, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14990, + "weighted_flow_matching_loss": 0.014201536495238542, + "weighted_rkd_loss": 0.013163040857762098 + }, + { + "epoch": 4.364271166715159, + "flow_matching_loss": 0.013392443582415581, + "grad_norm": 0.17715072631835938, + "learning_rate": 2.6837615414057677e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005880184611305595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00255034351721406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01431071273982525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8536505699157715, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15000, + "weighted_flow_matching_loss": 0.013392443582415581, + "weighted_rkd_loss": 0.012216304522007703 + }, + { + "epoch": 4.367180680826302, + "flow_matching_loss": 0.014365181885659694, + "grad_norm": 0.22052066028118134, + "learning_rate": 2.6832536065014708e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006196171045303345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002556117728818208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014948459714651108, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8534654378890991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15010, + "weighted_flow_matching_loss": 0.014365181885659694, + "weighted_rkd_loss": 0.012757994513958693 + }, + { + "epoch": 4.370090194937445, + "flow_matching_loss": 0.01670350069180131, + "grad_norm": 0.1679365485906601, + "learning_rate": 2.68274531215597e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005878540244884789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002490578219294548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014247658662497997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.853280222415924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15020, + "weighted_flow_matching_loss": 0.01670350069180131, + "weighted_rkd_loss": 0.012157336343079805 + }, + { + "epoch": 4.372999709048589, + "flow_matching_loss": 0.014984485134482384, + "grad_norm": 0.1741015613079071, + "learning_rate": 2.6822366585236716e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006137979403138161, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002711209701374173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014987168554216623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8530949056148529, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15030, + "weighted_flow_matching_loss": 0.014984485134482384, + "weighted_rkd_loss": 0.012785513047128915 + }, + { + "epoch": 4.3759092231597325, + "flow_matching_loss": 0.015230135526508093, + "grad_norm": 0.17153915762901306, + "learning_rate": 2.6817276457590924e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067987809889018536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002852291357703507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016449853498488664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.852909505367279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15040, + "weighted_flow_matching_loss": 0.015230135526508093, + "weighted_rkd_loss": 0.014030216168612242 + }, + { + "epoch": 4.378818737270876, + "flow_matching_loss": 0.014377017319202424, + "grad_norm": 0.14736495912075043, + "learning_rate": 2.6812182740168555e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005863504251465201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026109553524293005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014337963797152043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8527239918708801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15050, + "weighted_flow_matching_loss": 0.014377017319202424, + "weighted_rkd_loss": 0.012226348463445901 + }, + { + "epoch": 4.381728251382019, + "flow_matching_loss": 0.013072763569653034, + "grad_norm": 0.17009343206882477, + "learning_rate": 2.6807085434516953e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006632123375311494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028939555399119855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016158202290534975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8525383770465851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15060, + "weighted_flow_matching_loss": 0.013072763569653034, + "weighted_rkd_loss": 0.013775510713458061 + }, + { + "epoch": 4.384637765493163, + "flow_matching_loss": 0.014148245565593243, + "grad_norm": 0.20369431376457214, + "learning_rate": 2.6801984542184544e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007770217163488269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034162512747570872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018956685438752174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8523526847362518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15070, + "weighted_flow_matching_loss": 0.014148245565593243, + "weighted_rkd_loss": 0.016157805826514958 + }, + { + "epoch": 4.3875472796043065, + "flow_matching_loss": 0.013959443289786577, + "grad_norm": 0.16814424097537994, + "learning_rate": 2.6796880064720845e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005587498610839248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026084665092639624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013783463556319476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.852166885137558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15080, + "weighted_flow_matching_loss": 0.013959443289786577, + "weighted_rkd_loss": 0.011745765339583158 + }, + { + "epoch": 4.390456793715449, + "flow_matching_loss": 0.015568314399570227, + "grad_norm": 0.17238353192806244, + "learning_rate": 2.6791772003676462e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006114413077011704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025444986647926273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014773324877023698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.851980984210968, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15090, + "weighted_flow_matching_loss": 0.015568314399570227, + "weighted_rkd_loss": 0.012586659844964742 + }, + { + "epoch": 4.393366307826593, + "flow_matching_loss": 0.013806181866675616, + "grad_norm": 0.2051565945148468, + "learning_rate": 2.6786660360603087e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006902071088552475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002795084146782756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016599226277321577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8517949998378753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15100, + "weighted_flow_matching_loss": 0.013806181866675616, + "weighted_rkd_loss": 0.014139093086123466 + }, + { + "epoch": 4.396275821937737, + "flow_matching_loss": 0.014029177743941546, + "grad_norm": 0.17854583263397217, + "learning_rate": 2.6781545137053503e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005898101534694433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026961584109812973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014492361526936292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8516089141368866, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15110, + "weighted_flow_matching_loss": 0.014029177743941546, + "weighted_rkd_loss": 0.012341842148452997 + }, + { + "epoch": 4.3991853360488795, + "flow_matching_loss": 0.013550951797515153, + "grad_norm": 0.17876675724983215, + "learning_rate": 2.677642633458158e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00523020289838314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002265981584787369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012726387288421393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8514227271080017, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15120, + "weighted_flow_matching_loss": 0.013550951797515153, + "weighted_rkd_loss": 0.010835540946573019 + }, + { + "epoch": 4.402094850160023, + "flow_matching_loss": 0.014344675466418266, + "grad_norm": 0.1679052710533142, + "learning_rate": 2.6771303954742274e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005622811755165458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025887824594974518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013834405783563852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8512364506721497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15130, + "weighted_flow_matching_loss": 0.014344675466418266, + "weighted_rkd_loss": 0.011776412278413773 + }, + { + "epoch": 4.405004364271167, + "flow_matching_loss": 0.015558200143277644, + "grad_norm": 0.1523008495569229, + "learning_rate": 2.6766177999091633e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006405427865684032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003095950814895332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015906806290149688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8510500729084015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15140, + "weighted_flow_matching_loss": 0.015558200143277644, + "weighted_rkd_loss": 0.013537520356476308 + }, + { + "epoch": 4.40791387838231, + "flow_matching_loss": 0.013273522071540356, + "grad_norm": 0.17246781289577484, + "learning_rate": 2.676104846918678e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005843489384278655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026305287145078184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014317507483065128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8508635997772217, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15150, + "weighted_flow_matching_loss": 0.013273522071540356, + "weighted_rkd_loss": 0.012182292621582746 + }, + { + "epoch": 4.4108233924934535, + "flow_matching_loss": 0.013326785992830992, + "grad_norm": 0.16901475191116333, + "learning_rate": 2.675591536658594e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005869081662967801, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026882786070927976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014426441956311464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8506770312786103, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15160, + "weighted_flow_matching_loss": 0.013326785992830992, + "weighted_rkd_loss": 0.01227230904623866 + }, + { + "epoch": 4.413732906604597, + "flow_matching_loss": 0.015113721322268247, + "grad_norm": 0.18003828823566437, + "learning_rate": 2.6750778692848405e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005609372444450855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025041479850187896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013722892850637436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8504903614521027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15170, + "weighted_flow_matching_loss": 0.015113721322268247, + "weighted_rkd_loss": 0.011671223025768996 + }, + { + "epoch": 4.416642420715741, + "flow_matching_loss": 0.014985421486198903, + "grad_norm": 0.16931158304214478, + "learning_rate": 2.6745638449534562e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005894496804103255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026100034825503825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014398996997624636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8503036022186279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15180, + "weighted_flow_matching_loss": 0.014985421486198903, + "weighted_rkd_loss": 0.012243571784347295 + }, + { + "epoch": 4.419551934826884, + "flow_matching_loss": 0.012456835713237524, + "grad_norm": 0.20355944335460663, + "learning_rate": 2.6740494638205888e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005757623678073287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025624627945944668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014077710174024104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8501167476177216, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15190, + "weighted_flow_matching_loss": 0.012456835713237524, + "weighted_rkd_loss": 0.011967652197927236 + }, + { + "epoch": 4.422461448938027, + "flow_matching_loss": 0.013095467910170555, + "grad_norm": 0.197760209441185, + "learning_rate": 2.6735347260424928e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005924325995147228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025830950937233864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014431747142225504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.849929791688919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15200, + "weighted_flow_matching_loss": 0.013095467910170555, + "weighted_rkd_loss": 0.012265935353934766 + }, + { + "epoch": 4.425370963049171, + "flow_matching_loss": 0.013285278063267469, + "grad_norm": 0.2356666475534439, + "learning_rate": 2.6730196317755326e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006378764845430851, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002656467375345528, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015413997042924166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8497427463531494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15210, + "weighted_flow_matching_loss": 0.013285278063267469, + "weighted_rkd_loss": 0.013097866717725993 + }, + { + "epoch": 4.428280477160314, + "flow_matching_loss": 0.011742026265710592, + "grad_norm": 0.18434688448905945, + "learning_rate": 2.6725041811761805e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063385302666574715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026020836317911743, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015279144328087569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8495556056499481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15220, + "weighted_flow_matching_loss": 0.011742026265710592, + "weighted_rkd_loss": 0.012980547081679106 + }, + { + "epoch": 4.431189991271458, + "flow_matching_loss": 0.013323961291462183, + "grad_norm": 0.15006296336650848, + "learning_rate": 2.671988374401016e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007254387624561787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031101683853194118, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01761894365772605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8493683576583863, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15230, + "weighted_flow_matching_loss": 0.013323961291462183, + "weighted_rkd_loss": 0.014964980073273182 + }, + { + "epoch": 4.434099505382601, + "flow_matching_loss": 0.014593095425516367, + "grad_norm": 0.210915207862854, + "learning_rate": 2.671472211606728e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006149936048313975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027624481124803426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015062320046126842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8491810202598572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15240, + "weighted_flow_matching_loss": 0.014593095425516367, + "weighted_rkd_loss": 0.012790643516927958 + }, + { + "epoch": 4.437009019493744, + "flow_matching_loss": 0.014103217143565416, + "grad_norm": 0.19711683690547943, + "learning_rate": 2.6709556929501128e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005999933648854494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026266565779224037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014626523759216071, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8489935874938965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15250, + "weighted_flow_matching_loss": 0.014103217143565416, + "weighted_rkd_loss": 0.012417799700051546 + }, + { + "epoch": 4.439918533604888, + "flow_matching_loss": 0.013259402755647898, + "grad_norm": 0.18152792751789093, + "learning_rate": 2.670438818588076e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005899306642822922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027544083073735236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014553021639585495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8488060593605041, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15260, + "weighted_flow_matching_loss": 0.013259402755647898, + "weighted_rkd_loss": 0.012352696526795626 + }, + { + "epoch": 4.442828047716032, + "flow_matching_loss": 0.014792061690241099, + "grad_norm": 0.16191522777080536, + "learning_rate": 2.6699215886776287e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006278419541195035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029073609155602755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015464199893176555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8486184418201447, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15270, + "weighted_flow_matching_loss": 0.014792061690241099, + "weighted_rkd_loss": 0.013123161159455776 + }, + { + "epoch": 4.445737561827175, + "flow_matching_loss": 0.013296201638877393, + "grad_norm": 0.1904374361038208, + "learning_rate": 2.6694040033758933e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006676619127392769, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031891566701233386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016542394738644363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.848430722951889, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15280, + "weighted_flow_matching_loss": 0.013296201638877393, + "weighted_rkd_loss": 0.014035151991993188 + }, + { + "epoch": 4.448647075938318, + "flow_matching_loss": 0.012476899567991495, + "grad_norm": 0.17014466226100922, + "learning_rate": 2.6688860628400982e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006165778264403343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026850212598219514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015016577765345573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8482428967952729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15290, + "weighted_flow_matching_loss": 0.012476899567991495, + "weighted_rkd_loss": 0.012737773079425096 + }, + { + "epoch": 4.451556590049462, + "flow_matching_loss": 0.013159089162945748, + "grad_norm": 0.19411200284957886, + "learning_rate": 2.6683677672275797e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006100138463079929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002607139083556831, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014807416032999754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8480549931526185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15300, + "weighted_flow_matching_loss": 0.013159089162945748, + "weighted_rkd_loss": 0.01255743345245719 + }, + { + "epoch": 4.4544661041606055, + "flow_matching_loss": 0.014730932842940091, + "grad_norm": 0.18537090718746185, + "learning_rate": 2.6678491166957825e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006025448394939304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027828798396512867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014833776745945215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8478669822216034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15310, + "weighted_flow_matching_loss": 0.014730932842940091, + "weighted_rkd_loss": 0.012577023170888425 + }, + { + "epoch": 4.457375618271748, + "flow_matching_loss": 0.013795346673578023, + "grad_norm": 0.17624230682849884, + "learning_rate": 2.6673301114022592e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066810586024075745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027338839136064053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016096000932157038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8476788878440857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15320, + "weighted_flow_matching_loss": 0.013795346673578023, + "weighted_rkd_loss": 0.01364424340426922 + }, + { + "epoch": 4.460285132382892, + "flow_matching_loss": 0.014450593292713166, + "grad_norm": 0.17235083878040314, + "learning_rate": 2.666810751504669e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006887612026184797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031773563008755446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01695258030667901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8474906861782074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15330, + "weighted_flow_matching_loss": 0.014450593292713166, + "weighted_rkd_loss": 0.014367081876844167 + }, + { + "epoch": 4.463194646494036, + "flow_matching_loss": 0.015164302382618188, + "grad_norm": 0.18000546097755432, + "learning_rate": 2.6662910371607807e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00562286083586514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025672418414615095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013812963478267192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8473023772239685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15340, + "weighted_flow_matching_loss": 0.015164302382618188, + "weighted_rkd_loss": 0.01170383533462882 + }, + { + "epoch": 4.466104160605179, + "flow_matching_loss": 0.013168525788933039, + "grad_norm": 0.18371589481830597, + "learning_rate": 2.6657709685284696e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005874660005792976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002550174237694591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014299494307488203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8471140086650848, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15350, + "weighted_flow_matching_loss": 0.013168525788933039, + "weighted_rkd_loss": 0.012113246507942677 + }, + { + "epoch": 4.469013674716322, + "flow_matching_loss": 0.014289846923202277, + "grad_norm": 0.1830817610025406, + "learning_rate": 2.6652505457657183e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006194710871204734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027810461819171906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015170467738062143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8469255268573761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15360, + "weighted_flow_matching_loss": 0.014289846923202277, + "weighted_rkd_loss": 0.012848189286887645 + }, + { + "epoch": 4.471923188827466, + "flow_matching_loss": 0.013832877203822137, + "grad_norm": 0.21721191704273224, + "learning_rate": 2.664729769030618e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006724837329238653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027158565702848137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016165531426668166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8467369556427002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15370, + "weighted_flow_matching_loss": 0.013832877203822137, + "weighted_rkd_loss": 0.013687896821647883 + }, + { + "epoch": 4.47483270293861, + "flow_matching_loss": 0.014486865513026715, + "grad_norm": 0.2620255649089813, + "learning_rate": 2.6642086384813667e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005708440858870745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002767906873486936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01418478861451149, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8465482711791992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15380, + "weighted_flow_matching_loss": 0.014486865513026715, + "weighted_rkd_loss": 0.012008131202310323 + }, + { + "epoch": 4.4777422170497525, + "flow_matching_loss": 0.012513806764036416, + "grad_norm": 0.18095092475414276, + "learning_rate": 2.6636871542762703e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006023195199668407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002799427672289312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014845818001776933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8463595151901245, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15390, + "weighted_flow_matching_loss": 0.012513806764036416, + "weighted_rkd_loss": 0.012564865034073591 + }, + { + "epoch": 4.480651731160896, + "flow_matching_loss": 0.013330232445150613, + "grad_norm": 0.21518664062023163, + "learning_rate": 2.6631653165737418e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006148959510028362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002718252153135836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015016170963644982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8461706340312958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15400, + "weighted_flow_matching_loss": 0.013330232445150613, + "weighted_rkd_loss": 0.012706194538623095 + }, + { + "epoch": 4.48356124527204, + "flow_matching_loss": 0.01385624911636114, + "grad_norm": 0.19068624079227448, + "learning_rate": 2.662643125532302e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006856329645961523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003125279280357063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01683793878182769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8459816873073578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15410, + "weighted_flow_matching_loss": 0.01385624911636114, + "weighted_rkd_loss": 0.01424457598477602 + }, + { + "epoch": 4.486470759383183, + "flow_matching_loss": 0.015197847597301006, + "grad_norm": 0.18563833832740784, + "learning_rate": 2.6621205813105777e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006533640157431364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031393423792906105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016206622775644063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8457926332950592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15420, + "weighted_flow_matching_loss": 0.015197847597301006, + "weighted_rkd_loss": 0.013707443978637457 + }, + { + "epoch": 4.4893802734943264, + "flow_matching_loss": 0.013886940386146306, + "grad_norm": 0.23060688376426697, + "learning_rate": 2.6615976840673056e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006642467807978392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028736904729157685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016158626042306425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8456034898757935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15430, + "weighted_flow_matching_loss": 0.013886940386146306, + "weighted_rkd_loss": 0.013663728442043065 + }, + { + "epoch": 4.49228978760547, + "flow_matching_loss": 0.013389635179191828, + "grad_norm": 0.24230293929576874, + "learning_rate": 2.6610744339613265e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00554693778976798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026545636472292246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013748439121991396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8454142451286316, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15440, + "weighted_flow_matching_loss": 0.013389635179191828, + "weighted_rkd_loss": 0.011623120959848165 + }, + { + "epoch": 4.495199301716613, + "flow_matching_loss": 0.014538021944463252, + "grad_norm": 0.18970879912376404, + "learning_rate": 2.6605508311515916e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006944564916193485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003191597806289792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017080727592110635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.845224916934967, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15450, + "weighted_flow_matching_loss": 0.014538021944463252, + "weighted_rkd_loss": 0.014437029883265496 + }, + { + "epoch": 4.498108815827757, + "flow_matching_loss": 0.013409700710326433, + "grad_norm": 0.21118827164173126, + "learning_rate": 2.6600268757971566e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006076834304258228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026208288734778764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014774497505277396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8450354814529419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15460, + "weighted_flow_matching_loss": 0.013409700710326433, + "weighted_rkd_loss": 0.01248489273712039 + }, + { + "epoch": 4.5010183299389, + "flow_matching_loss": 0.014188131969422101, + "grad_norm": 0.18335263431072235, + "learning_rate": 2.659502568057185e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005872003734111786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002593428874388337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014337436389178037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8448459506034851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15470, + "weighted_flow_matching_loss": 0.014188131969422101, + "weighted_rkd_loss": 0.012112908996641636 + }, + { + "epoch": 4.503927844050043, + "flow_matching_loss": 0.01436835089698434, + "grad_norm": 0.19661815464496613, + "learning_rate": 2.658977908090949e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00634999992325902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002966342773288488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015666342712938784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8446563422679901, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15480, + "weighted_flow_matching_loss": 0.01436835089698434, + "weighted_rkd_loss": 0.013232770748436451 + }, + { + "epoch": 4.506837358161187, + "flow_matching_loss": 0.014273880887776614, + "grad_norm": 0.16582919657230377, + "learning_rate": 2.6584528960578245e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071925527416169645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030772560043260456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017462361417710782, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8444666206836701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15490, + "weighted_flow_matching_loss": 0.014273880887776614, + "weighted_rkd_loss": 0.01474643051624298 + }, + { + "epoch": 4.509746872272331, + "flow_matching_loss": 0.014898780081421138, + "grad_norm": 0.19189776480197906, + "learning_rate": 2.657927532117298e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006278236722573638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003096914105117321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01565338745713234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8442768156528473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15500, + "weighted_flow_matching_loss": 0.014898780081421138, + "weighted_rkd_loss": 0.013215810060501099 + }, + { + "epoch": 4.512656386383474, + "flow_matching_loss": 0.014124532882124185, + "grad_norm": 0.18545493483543396, + "learning_rate": 2.6574018164289605e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006106571760028601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00304483484942466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015257978346198797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.844086903333664, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15510, + "weighted_flow_matching_loss": 0.014124532882124185, + "weighted_rkd_loss": 0.01287909559905529 + }, + { + "epoch": 4.515565900494617, + "flow_matching_loss": 0.0141638801433146, + "grad_norm": 0.21306441724300385, + "learning_rate": 2.6568757491525103e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005954145127907396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026986110024154186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014606901351362467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8438969135284424, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15520, + "weighted_flow_matching_loss": 0.0141638801433146, + "weighted_rkd_loss": 0.012326694652438163 + }, + { + "epoch": 4.518475414605761, + "flow_matching_loss": 0.014583136420696974, + "grad_norm": 0.15664447844028473, + "learning_rate": 2.6563493304477534e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066694874782115224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030022710212506353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016341246012598277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8437068164348602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15530, + "weighted_flow_matching_loss": 0.014583136420696974, + "weighted_rkd_loss": 0.013787381257861853 + }, + { + "epoch": 4.521384928716905, + "flow_matching_loss": 0.013631100952625274, + "grad_norm": 0.17963489890098572, + "learning_rate": 2.655822560474601e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005995938600972295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028501452645286916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014842022396624088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8435166358947754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15540, + "weighted_flow_matching_loss": 0.013631100952625274, + "weighted_rkd_loss": 0.012519595958292485 + }, + { + "epoch": 4.524294442828047, + "flow_matching_loss": 0.013605820760130883, + "grad_norm": 0.2161926031112671, + "learning_rate": 2.655295439393073e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005796746769919992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002652452327311039, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01424594596028328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.84332634806633, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15550, + "weighted_flow_matching_loss": 0.013605820760130883, + "weighted_rkd_loss": 0.01201401287689805 + }, + { + "epoch": 4.527203956939191, + "flow_matching_loss": 0.014602745696902276, + "grad_norm": 0.1871359497308731, + "learning_rate": 2.654767967363294e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00610241019167006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002701798325870186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01490661846473813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8431359827518463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15560, + "weighted_flow_matching_loss": 0.014602745696902276, + "weighted_rkd_loss": 0.012568308878690005 + }, + { + "epoch": 4.530113471050335, + "flow_matching_loss": 0.014694784302264452, + "grad_norm": 0.16948291659355164, + "learning_rate": 2.6542401445454967e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006338676810264588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002738507406320423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015415860991925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8429455041885376, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15570, + "weighted_flow_matching_loss": 0.014694784302264452, + "weighted_rkd_loss": 0.012994767259806394 + }, + { + "epoch": 4.5330229851614785, + "flow_matching_loss": 0.013569669611752034, + "grad_norm": 0.2037096619606018, + "learning_rate": 2.6537119711000187e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005800524028018117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024999713874422013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01410101940855384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8427549481391907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15580, + "weighted_flow_matching_loss": 0.013569669611752034, + "weighted_rkd_loss": 0.011883719451725483 + }, + { + "epoch": 4.535932499272621, + "flow_matching_loss": 0.011783641390502453, + "grad_norm": 0.17010727524757385, + "learning_rate": 2.6531834471873068e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005967267137020826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025624231318943204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014496957231312991, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8425642848014832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15590, + "weighted_flow_matching_loss": 0.011783641390502453, + "weighted_rkd_loss": 0.012214682064950466 + }, + { + "epoch": 4.538842013383765, + "flow_matching_loss": 0.0141597930341959, + "grad_norm": 0.15988761186599731, + "learning_rate": 2.6526545729679115e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005585389258340001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024903625948354603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01366114104166627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.842373538017273, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15600, + "weighted_flow_matching_loss": 0.0141597930341959, + "weighted_rkd_loss": 0.011507782246917486 + }, + { + "epoch": 4.541751527494909, + "flow_matching_loss": 0.014035113900899888, + "grad_norm": 0.18697088956832886, + "learning_rate": 2.6521253486024908e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006089773657731712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002439454756677151, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014619001932442189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8421826958656311, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15610, + "weighted_flow_matching_loss": 0.014035113900899888, + "weighted_rkd_loss": 0.012311760475859046 + }, + { + "epoch": 4.5446610416060516, + "flow_matching_loss": 0.013231019210070371, + "grad_norm": 0.16893646121025085, + "learning_rate": 2.651595774251809e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005845558410510421, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025285401032306256, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014219656959176064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8419917583465576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15620, + "weighted_flow_matching_loss": 0.013231019210070371, + "weighted_rkd_loss": 0.01197283398360014 + }, + { + "epoch": 4.547570555717195, + "flow_matching_loss": 0.014099250547587872, + "grad_norm": 0.20165085792541504, + "learning_rate": 2.6510658500767378e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006481517804786563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002985994890332222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01594903049990535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8418007314205169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15630, + "weighted_flow_matching_loss": 0.014099250547587872, + "weighted_rkd_loss": 0.013426014687865973 + }, + { + "epoch": 4.550480069828339, + "flow_matching_loss": 0.015772462636232377, + "grad_norm": 0.16641958057880402, + "learning_rate": 2.650535576238253e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005664298962801695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024650245904922485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013793622609227896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8416095972061157, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15640, + "weighted_flow_matching_loss": 0.015772462636232377, + "weighted_rkd_loss": 0.011608858965337277 + }, + { + "epoch": 4.553389583939482, + "flow_matching_loss": 0.013468792568892241, + "grad_norm": 0.203900545835495, + "learning_rate": 2.650004952897438e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059559091459959745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027705412125214935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014682359620928764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8414183855056763, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15650, + "weighted_flow_matching_loss": 0.013468792568892241, + "weighted_rkd_loss": 0.012354004848748445 + }, + { + "epoch": 4.5562990980506255, + "flow_matching_loss": 0.01412406489253044, + "grad_norm": 0.15272296965122223, + "learning_rate": 2.649473980215483e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006048292992636562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029285624623298645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015025148447602988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8412270784378052, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15660, + "weighted_flow_matching_loss": 0.01412406489253044, + "weighted_rkd_loss": 0.012639582809060811 + }, + { + "epoch": 4.559208612161769, + "flow_matching_loss": 0.01408252865076065, + "grad_norm": 0.16268840432167053, + "learning_rate": 2.648942658353683e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006320182699710131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027782946708612144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01541865998879075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.841035658121109, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15670, + "weighted_flow_matching_loss": 0.01408252865076065, + "weighted_rkd_loss": 0.012967728357762099 + }, + { + "epoch": 4.562118126272912, + "flow_matching_loss": 0.01377541609108448, + "grad_norm": 0.22014141082763672, + "learning_rate": 2.648410987473439e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005950823333114386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002726540365256369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014628187008202075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8408441603183746, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15680, + "weighted_flow_matching_loss": 0.01377541609108448, + "weighted_rkd_loss": 0.0122999656945467 + }, + { + "epoch": 4.565027640384056, + "flow_matching_loss": 0.013473009131848812, + "grad_norm": 0.19940124452114105, + "learning_rate": 2.6478789677362596e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005762131046503782, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002785567019600421, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014309829100966454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8406525671482086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15690, + "weighted_flow_matching_loss": 0.013473009131848812, + "weighted_rkd_loss": 0.012029568664729595 + }, + { + "epoch": 4.567937154495199, + "flow_matching_loss": 0.013096723333001136, + "grad_norm": 0.22145196795463562, + "learning_rate": 2.6473465993037572e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006239445833489299, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002740469784475863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01521936133503914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8404608845710755, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15700, + "weighted_flow_matching_loss": 0.013096723333001136, + "weighted_rkd_loss": 0.012791277188807727 + }, + { + "epoch": 4.570846668606343, + "flow_matching_loss": 0.014419854059815407, + "grad_norm": 0.13759759068489075, + "learning_rate": 2.6468138823376522e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006266026478260756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027072938391938806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015239346772432327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8402691006660461, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15710, + "weighted_flow_matching_loss": 0.014419854059815407, + "weighted_rkd_loss": 0.01280509913340211 + }, + { + "epoch": 4.573756182717486, + "flow_matching_loss": 0.012924960348755121, + "grad_norm": 0.15962986648082733, + "learning_rate": 2.6462808169997695e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006406439235433936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002825397625565529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015638276096433402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8400772213935852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15720, + "weighted_flow_matching_loss": 0.012924960348755121, + "weighted_rkd_loss": 0.01313743032515049 + }, + { + "epoch": 4.57666569682863, + "flow_matching_loss": 0.014723936654627323, + "grad_norm": 0.20529624819755554, + "learning_rate": 2.6457474034520402e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006559762498363852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003162335744127631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016281860694289208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8398852646350861, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15730, + "weighted_flow_matching_loss": 0.014723936654627323, + "weighted_rkd_loss": 0.013674911111593246 + }, + { + "epoch": 4.579575210939773, + "flow_matching_loss": 0.012985633499920369, + "grad_norm": 0.15507152676582336, + "learning_rate": 2.6452136418565016e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052732746116817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002454350411426276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01300089992582798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8396931946277618, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15740, + "weighted_flow_matching_loss": 0.012985633499920369, + "weighted_rkd_loss": 0.010916796745732427 + }, + { + "epoch": 4.582484725050916, + "flow_matching_loss": 0.013583007641136647, + "grad_norm": 0.19958697259426117, + "learning_rate": 2.6446795323752962e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005902461614459753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027545914519578217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014559514727443457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8395010471343994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15750, + "weighted_flow_matching_loss": 0.013583007641136647, + "weighted_rkd_loss": 0.012222732789814473 + }, + { + "epoch": 4.58539423916206, + "flow_matching_loss": 0.012653921823948621, + "grad_norm": 0.1650596559047699, + "learning_rate": 2.6441450751706725e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061574229039251804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002567056845873594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014881902746856213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8393088042736053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15760, + "weighted_flow_matching_loss": 0.012653921823948621, + "weighted_rkd_loss": 0.012490585818886757 + }, + { + "epoch": 4.588303753273204, + "flow_matching_loss": 0.012293610628694296, + "grad_norm": 0.17818275094032288, + "learning_rate": 2.6436102704049843e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005957987112924457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027512572472915054, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014667231403291226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8391164600849151, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15770, + "weighted_flow_matching_loss": 0.012293610628694296, + "weighted_rkd_loss": 0.012307511828839778 + }, + { + "epoch": 4.591213267384346, + "flow_matching_loss": 0.014583496376872063, + "grad_norm": 0.2110544741153717, + "learning_rate": 2.6430751182406913e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005771487532183528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002543870243243873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014086845237761736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8389240205287933, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15780, + "weighted_flow_matching_loss": 0.014583496376872063, + "weighted_rkd_loss": 0.01181785324588418 + }, + { + "epoch": 4.59412278149549, + "flow_matching_loss": 0.014753040578216314, + "grad_norm": 0.1748286485671997, + "learning_rate": 2.6425396188403585e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005891253566369414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025134813506156204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014295988529920579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8387314975261688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15790, + "weighted_flow_matching_loss": 0.014753040578216314, + "weighted_rkd_loss": 0.01199051896110177 + }, + { + "epoch": 4.597032295606634, + "flow_matching_loss": 0.012936047464609145, + "grad_norm": 0.19466239213943481, + "learning_rate": 2.6420037723666565e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00586985694244504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025326446862891316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014272358734160662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8385388791561127, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15800, + "weighted_flow_matching_loss": 0.012936047464609145, + "weighted_rkd_loss": 0.011967870965600013 + }, + { + "epoch": 4.599941809717777, + "flow_matching_loss": 0.014463020488619804, + "grad_norm": 0.15273690223693848, + "learning_rate": 2.641467578982361e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005945534398779273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002504719328135252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014395788125693799, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8383461713790894, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15810, + "weighted_flow_matching_loss": 0.014463020488619804, + "weighted_rkd_loss": 0.012068627867847681 + }, + { + "epoch": 4.60285132382892, + "flow_matching_loss": 0.013641273882240058, + "grad_norm": 0.20587176084518433, + "learning_rate": 2.6409310388503542e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006211938848719001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002883135434240103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015307013224810363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8381533741950988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15820, + "weighted_flow_matching_loss": 0.013641273882240058, + "weighted_rkd_loss": 0.012829664535820485 + }, + { + "epoch": 4.605760837940064, + "flow_matching_loss": 0.012867502681910992, + "grad_norm": 0.15790943801403046, + "learning_rate": 2.640394152133622e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005484696524217725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024455094942823054, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01341490251943469, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8379604697227478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15830, + "weighted_flow_matching_loss": 0.012867502681910992, + "weighted_rkd_loss": 0.01124119507148862 + }, + { + "epoch": 4.608670352051208, + "flow_matching_loss": 0.015022817067801952, + "grad_norm": 0.17575842142105103, + "learning_rate": 2.6398569189952573e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006611452857032418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002950842445716262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016173748206347226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8377674877643585, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15840, + "weighted_flow_matching_loss": 0.015022817067801952, + "weighted_rkd_loss": 0.013549810741096735 + }, + { + "epoch": 4.611579866162351, + "flow_matching_loss": 0.012778459489345551, + "grad_norm": 0.20658496022224426, + "learning_rate": 2.6393193395984563e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005972333764657378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002786554768681526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01473122239112854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8375744223594666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15850, + "weighted_flow_matching_loss": 0.012778459489345551, + "weighted_rkd_loss": 0.012338474299758673 + }, + { + "epoch": 4.614489380273494, + "flow_matching_loss": 0.015235639549791813, + "grad_norm": 0.22216662764549255, + "learning_rate": 2.6387814141065216e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006531904311850667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00278128816280514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01584509704262018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.837381225824356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15860, + "weighted_flow_matching_loss": 0.015235639549791813, + "weighted_rkd_loss": 0.013268368225544691 + }, + { + "epoch": 4.617398894384638, + "flow_matching_loss": 0.012853435706347227, + "grad_norm": 0.1883953958749771, + "learning_rate": 2.6382431426828615e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005715378164313734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002531572652515024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01396232875995338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8371879696846009, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15870, + "weighted_flow_matching_loss": 0.012853435706347227, + "weighted_rkd_loss": 0.011689040064811706 + }, + { + "epoch": 4.620308408495781, + "flow_matching_loss": 0.01576787745580077, + "grad_norm": 0.18134909868240356, + "learning_rate": 2.6377045254909885e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005811042990535498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025792264845222236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01420131241902709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.836994606256485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15880, + "weighted_flow_matching_loss": 0.01576787745580077, + "weighted_rkd_loss": 0.011886407528072596 + }, + { + "epoch": 4.6232179226069245, + "flow_matching_loss": 0.014131619315594434, + "grad_norm": 0.18990790843963623, + "learning_rate": 2.6371655626945196e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005938191479071975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002640558569692075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014516941364854574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8368011474609375, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15890, + "weighted_flow_matching_loss": 0.014131619315594434, + "weighted_rkd_loss": 0.012147754430770874 + }, + { + "epoch": 4.626127436718068, + "flow_matching_loss": 0.014782614726573228, + "grad_norm": 0.1801673322916031, + "learning_rate": 2.6366262544571778e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005405798181891441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002394323970656842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013205920159816743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8366076111793518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15900, + "weighted_flow_matching_loss": 0.014782614726573228, + "weighted_rkd_loss": 0.01104817222803831 + }, + { + "epoch": 4.629036950829212, + "flow_matching_loss": 0.014692725148051977, + "grad_norm": 0.17657062411308289, + "learning_rate": 2.636086600942791e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006359660159796476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028124349308200182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015531755238771438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.83641397356987, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15910, + "weighted_flow_matching_loss": 0.014692725148051977, + "weighted_rkd_loss": 0.012991070188581944 + }, + { + "epoch": 4.631946464940355, + "flow_matching_loss": 0.014054228086024523, + "grad_norm": 0.15901976823806763, + "learning_rate": 2.6355466023152913e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005719025153666735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024980058311484753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013936056196689606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.836220246553421, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15920, + "weighted_flow_matching_loss": 0.014054228086024523, + "weighted_rkd_loss": 0.011653622332960367 + }, + { + "epoch": 4.6348559790514985, + "flow_matching_loss": 0.013000993337482215, + "grad_norm": 0.19547881186008453, + "learning_rate": 2.6350062587387164e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005849474482238293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027579738991335033, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014456922747194767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.836026418209076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15930, + "weighted_flow_matching_loss": 0.013000993337482215, + "weighted_rkd_loss": 0.012086271587759257 + }, + { + "epoch": 4.637765493162642, + "flow_matching_loss": 0.01537158265709877, + "grad_norm": 0.19228428602218628, + "learning_rate": 2.634465570377208e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006336598307825625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028490695869550107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01552226608619094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8358325183391571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15940, + "weighted_flow_matching_loss": 0.01537158265709877, + "weighted_rkd_loss": 0.012973831035196781 + }, + { + "epoch": 4.640675007273785, + "flow_matching_loss": 0.014469419792294502, + "grad_norm": 0.14944778382778168, + "learning_rate": 2.633924537395013e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005380977084860206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023164851823821663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013078439421951771, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8356385111808777, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15950, + "weighted_flow_matching_loss": 0.014469419792294502, + "weighted_rkd_loss": 0.010928868129849435 + }, + { + "epoch": 4.643584521384929, + "flow_matching_loss": 0.015377741958945989, + "grad_norm": 0.17592087388038635, + "learning_rate": 2.6333831599564835e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064655558206141, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029097018530592324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015840813610702754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8354444146156311, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15960, + "weighted_flow_matching_loss": 0.015377741958945989, + "weighted_rkd_loss": 0.013234141189604998 + }, + { + "epoch": 4.646494035496072, + "flow_matching_loss": 0.013428445346653462, + "grad_norm": 0.2281910926103592, + "learning_rate": 2.632841438226075e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061239517293870446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026632431196048855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01491114655509591, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8352502286434174, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15970, + "weighted_flow_matching_loss": 0.013428445346653462, + "weighted_rkd_loss": 0.012454538326710462 + }, + { + "epoch": 4.649403549607215, + "flow_matching_loss": 0.016314864344894885, + "grad_norm": 0.16023486852645874, + "learning_rate": 2.6322993723683486e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005950833158567548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002484397671651095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014386063907295465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.835055947303772, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15980, + "weighted_flow_matching_loss": 0.016314864344894885, + "weighted_rkd_loss": 0.012013150658458472 + }, + { + "epoch": 4.652313063718359, + "flow_matching_loss": 0.014813877921551467, + "grad_norm": 0.18971648812294006, + "learning_rate": 2.6317569625479695e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006150829838588834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028680569492280485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015169716533273458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8348615646362305, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15990, + "weighted_flow_matching_loss": 0.014813877921551467, + "weighted_rkd_loss": 0.012664689030498267 + }, + { + "epoch": 4.655222577829503, + "flow_matching_loss": 0.014009984582662583, + "grad_norm": 0.15699517726898193, + "learning_rate": 2.6312142089297076e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005783619033172727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024991016369313003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014066339656710624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8346671044826508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16000, + "weighted_flow_matching_loss": 0.014009984582662583, + "weighted_rkd_loss": 0.011740663275122642 + }, + { + "epoch": 4.6581320919406455, + "flow_matching_loss": 0.013324275985360146, + "grad_norm": 0.16838860511779785, + "learning_rate": 2.6306711116784366e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005946016404777765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002584996132645756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01447702907025814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8344725489616394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16010, + "weighted_flow_matching_loss": 0.013324275985360146, + "weighted_rkd_loss": 0.012080768309533595 + }, + { + "epoch": 4.661041606051789, + "flow_matching_loss": 0.014564551319926978, + "grad_norm": 0.20055770874023438, + "learning_rate": 2.6301276709591358e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006164946733042598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002687390521168709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015017283894121648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8342779040336609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16020, + "weighted_flow_matching_loss": 0.014564551319926978, + "weighted_rkd_loss": 0.01252858443185687 + }, + { + "epoch": 4.663951120162933, + "flow_matching_loss": 0.014939631707966328, + "grad_norm": 0.23048673570156097, + "learning_rate": 2.6295838869368878e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006050656037405133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002681616402696818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01478292839601636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8340831637382508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16030, + "weighted_flow_matching_loss": 0.014939631707966328, + "weighted_rkd_loss": 0.012330119824036955 + }, + { + "epoch": 4.666860634274077, + "flow_matching_loss": 0.013195583131164312, + "grad_norm": 0.15441741049289703, + "learning_rate": 2.6290397597768803e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056949292309582235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002446175995282829, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013836034573614597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8338883221149445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16040, + "weighted_flow_matching_loss": 0.013195583131164312, + "weighted_rkd_loss": 0.011537679471075535 + }, + { + "epoch": 4.669770148385219, + "flow_matching_loss": 0.01280459938570857, + "grad_norm": 0.17041082680225372, + "learning_rate": 2.6284952896444037e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006204855721443891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002839637058787048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015249348618090153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8336934089660645, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16050, + "weighted_flow_matching_loss": 0.01280459938570857, + "weighted_rkd_loss": 0.012713267561048269 + }, + { + "epoch": 4.672679662496363, + "flow_matching_loss": 0.013262454513460397, + "grad_norm": 0.17242109775543213, + "learning_rate": 2.6279504767048547e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005541125219315291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024252601666375994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013507510628551244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8334983944892883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16060, + "weighted_flow_matching_loss": 0.013262454513460397, + "weighted_rkd_loss": 0.01125848488882184 + }, + { + "epoch": 4.675589176607507, + "flow_matching_loss": 0.01362818256020546, + "grad_norm": 0.17519061267375946, + "learning_rate": 2.6274053211237323e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005398407485336066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024021184421144427, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013198933470994234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8333032906055451, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16070, + "weighted_flow_matching_loss": 0.01362818256020546, + "weighted_rkd_loss": 0.010998697485774755 + }, + { + "epoch": 4.67849869071865, + "flow_matching_loss": 0.014725138526409864, + "grad_norm": 0.18949198722839355, + "learning_rate": 2.626859823066641e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006138119287788868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025739460601471365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01485018478706479, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8331080913543701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16080, + "weighted_flow_matching_loss": 0.014725138526409864, + "weighted_rkd_loss": 0.012371713668107987 + }, + { + "epoch": 4.681408204829793, + "flow_matching_loss": 0.013617242220789194, + "grad_norm": 0.17776180803775787, + "learning_rate": 2.6263139826992886e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00503517643082887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022121183690615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012282471358776092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8329128086566925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16090, + "weighted_flow_matching_loss": 0.013617242220789194, + "weighted_rkd_loss": 0.010230128234252334 + }, + { + "epoch": 4.684317718940937, + "flow_matching_loss": 0.014264800865203143, + "grad_norm": 0.17901434004306793, + "learning_rate": 2.625767800187487e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006165350321680307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026663531898520887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01499705370515585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8327174246311188, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16100, + "weighted_flow_matching_loss": 0.014264800865203143, + "weighted_rkd_loss": 0.012488321587443351 + }, + { + "epoch": 4.687227233052081, + "flow_matching_loss": 0.013142896257340908, + "grad_norm": 0.16482098400592804, + "learning_rate": 2.6252212756971514e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005917057162150741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026044683647342024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014438582584261894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8325219690799713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16110, + "weighted_flow_matching_loss": 0.013142896257340908, + "weighted_rkd_loss": 0.012020495720207691 + }, + { + "epoch": 4.690136747163224, + "flow_matching_loss": 0.015253443270921707, + "grad_norm": 0.16538803279399872, + "learning_rate": 2.6246744093943022e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00560368001461029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025277042062953115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013735064212232828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8323263883590698, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16120, + "weighted_flow_matching_loss": 0.015253443270921707, + "weighted_rkd_loss": 0.011431969329714775 + }, + { + "epoch": 4.693046261274367, + "flow_matching_loss": 0.012616526056081056, + "grad_norm": 0.1632794737815857, + "learning_rate": 2.6241272014450625e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004959126422181726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021643172134645284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012082570046186448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.832130742073059, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16130, + "weighted_flow_matching_loss": 0.012616526056081056, + "weighted_rkd_loss": 0.01005421681329608 + }, + { + "epoch": 4.69595577538551, + "flow_matching_loss": 0.01449700053781271, + "grad_norm": 0.1554059535264969, + "learning_rate": 2.6235796520156603e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005302377557381987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022806339082308115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012885389104485511, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8319349884986877, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16140, + "weighted_flow_matching_loss": 0.01449700053781271, + "weighted_rkd_loss": 0.010719822999089957 + }, + { + "epoch": 4.698865289496654, + "flow_matching_loss": 0.013757082540541888, + "grad_norm": 0.17196965217590332, + "learning_rate": 2.623031761272426e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005356938578188419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002378586784470826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013092463929206133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8317391633987427, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16150, + "weighted_flow_matching_loss": 0.013757082540541888, + "weighted_rkd_loss": 0.010889538936316967 + }, + { + "epoch": 4.7017748036077975, + "flow_matching_loss": 0.014206983800977468, + "grad_norm": 0.15682874619960785, + "learning_rate": 2.6224835293817948e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00552238398231566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002368195739109069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013412963785231113, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8315432369709015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16160, + "weighted_flow_matching_loss": 0.014206983800977468, + "weighted_rkd_loss": 0.011153441574424505 + }, + { + "epoch": 4.704684317718941, + "flow_matching_loss": 0.01351099107414484, + "grad_norm": 0.18269379436969757, + "learning_rate": 2.6219349565103044e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060089577455073595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002744909282773733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014762824680656195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8313472211360932, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16170, + "weighted_flow_matching_loss": 0.01351099107414484, + "weighted_rkd_loss": 0.012273015175014734 + }, + { + "epoch": 4.707593831830084, + "flow_matching_loss": 0.015187408775091171, + "grad_norm": 0.1452736109495163, + "learning_rate": 2.6213860428245974e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005983282951638103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002639044588431716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014605610631406307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8311511099338531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16180, + "weighted_flow_matching_loss": 0.015187408775091171, + "weighted_rkd_loss": 0.012139534205198288 + }, + { + "epoch": 4.710503345941228, + "flow_matching_loss": 0.014703015517443418, + "grad_norm": 0.21091622114181519, + "learning_rate": 2.620836788491419e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006009894097223878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024267435190267863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014446531608700752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8309549152851105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16190, + "weighted_flow_matching_loss": 0.014703015517443418, + "weighted_rkd_loss": 0.012004402745515109 + }, + { + "epoch": 4.7134128600523715, + "flow_matching_loss": 0.013274364452809096, + "grad_norm": 0.17037995159626007, + "learning_rate": 2.6202871936776183e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00548083889298141, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023868737858720123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013348551467061043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8307586252689362, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16200, + "weighted_flow_matching_loss": 0.013274364452809096, + "weighted_rkd_loss": 0.011089470330625772 + }, + { + "epoch": 4.716322374163514, + "flow_matching_loss": 0.01338179036974907, + "grad_norm": 0.1909295618534088, + "learning_rate": 2.6197372585501477e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006241819029673934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027868758887052538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015270514041185379, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8305622458457946, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16210, + "weighted_flow_matching_loss": 0.01338179036974907, + "weighted_rkd_loss": 0.012683193385601043 + }, + { + "epoch": 4.719231888274658, + "flow_matching_loss": 0.015009225253015757, + "grad_norm": 0.18554192781448364, + "learning_rate": 2.6191869832760626e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006095391185954213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002738433622289449, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014929215703159571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.830365777015686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16220, + "weighted_flow_matching_loss": 0.015009225253015757, + "weighted_rkd_loss": 0.01239673988893628 + }, + { + "epoch": 4.722141402385802, + "flow_matching_loss": 0.014835519716143608, + "grad_norm": 0.22354841232299805, + "learning_rate": 2.618636368022523e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006144417496398091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002803217084147036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015092052146792411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8301692128181457, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16230, + "weighted_flow_matching_loss": 0.014835519716143608, + "weighted_rkd_loss": 0.012528934795409442 + }, + { + "epoch": 4.725050916496945, + "flow_matching_loss": 0.014226538967341185, + "grad_norm": 0.15462976694107056, + "learning_rate": 2.6180854129567902e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005970953963696957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002528911572881043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014470819570124149, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8299725651741028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16240, + "weighted_flow_matching_loss": 0.014226538967341185, + "weighted_rkd_loss": 0.012010366469621659 + }, + { + "epoch": 4.727960430608088, + "flow_matching_loss": 0.014983478002250195, + "grad_norm": 0.18299773335456848, + "learning_rate": 2.6175341182462303e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005589629709720612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023532445658929644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013532503973692656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8297758221626281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16250, + "weighted_flow_matching_loss": 0.014983478002250195, + "weighted_rkd_loss": 0.011228939983993769 + }, + { + "epoch": 4.730869944719232, + "flow_matching_loss": 0.014008455444127321, + "grad_norm": 0.17039890587329865, + "learning_rate": 2.6169824840583124e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005614438466727734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024810463655740023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01370992325246334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.829578983783722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16260, + "weighted_flow_matching_loss": 0.014008455444127321, + "weighted_rkd_loss": 0.011373511794954539 + }, + { + "epoch": 4.733779458830376, + "flow_matching_loss": 0.014159954618662596, + "grad_norm": 0.21262924373149872, + "learning_rate": 2.6164305105606076e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006147924670949578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002731071365997195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015026920568197965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.829382061958313, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16270, + "weighted_flow_matching_loss": 0.014159954618662596, + "weighted_rkd_loss": 0.01246300907805562 + }, + { + "epoch": 4.7366889729415185, + "flow_matching_loss": 0.014821443799883127, + "grad_norm": 0.14267534017562866, + "learning_rate": 2.6158781979207918e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048777676653116945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002075317606795579, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011830852925777435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8291850507259368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16280, + "weighted_flow_matching_loss": 0.014821443799883127, + "weighted_rkd_loss": 0.009809968108311296 + }, + { + "epoch": 4.739598487052662, + "flow_matching_loss": 0.012788129597902298, + "grad_norm": 0.24583600461483002, + "learning_rate": 2.615325546306642e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005998892756178975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002634419803507626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01463220538571477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8289879441261292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16290, + "weighted_flow_matching_loss": 0.012788129597902298, + "weighted_rkd_loss": 0.012129822000861167 + }, + { + "epoch": 4.742508001163806, + "flow_matching_loss": 0.013655418530106544, + "grad_norm": 0.18447653949260712, + "learning_rate": 2.61477255588604e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006183497607707977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027396560995839535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015106651186943054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8287907540798187, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16300, + "weighted_flow_matching_loss": 0.013655418530106544, + "weighted_rkd_loss": 0.012520265206694603 + }, + { + "epoch": 4.745417515274949, + "flow_matching_loss": 0.013739829137921333, + "grad_norm": 0.20623764395713806, + "learning_rate": 2.614219226826969e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00559450346045196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026207156712189316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013809722755104304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8285934627056122, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16310, + "weighted_flow_matching_loss": 0.013739829137921333, + "weighted_rkd_loss": 0.011442634835839271 + }, + { + "epoch": 4.748327029386092, + "flow_matching_loss": 0.015312467142939568, + "grad_norm": 0.14153112471103668, + "learning_rate": 2.6136655592975166e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005860195541754365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026052128640003504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014325603935867548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8283960938453674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16320, + "weighted_flow_matching_loss": 0.015312467142939568, + "weighted_rkd_loss": 0.011867299675941467 + }, + { + "epoch": 4.751236543497236, + "flow_matching_loss": 0.013403274305164813, + "grad_norm": 0.1815088540315628, + "learning_rate": 2.613111553465872e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005422028433531523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025564025505445897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013400459568947554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8281986296176911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16330, + "weighted_flow_matching_loss": 0.013403274305164813, + "weighted_rkd_loss": 0.011098231934010983 + }, + { + "epoch": 4.754146057608379, + "flow_matching_loss": 0.015062342025339603, + "grad_norm": 0.2305532991886139, + "learning_rate": 2.6125572095003276e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006001046672463417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026571180555038155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014659211225807666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.828001081943512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16340, + "weighted_flow_matching_loss": 0.015062342025339603, + "weighted_rkd_loss": 0.012137802504003047 + }, + { + "epoch": 4.757055571719523, + "flow_matching_loss": 0.01570644611492753, + "grad_norm": 0.1609696000814438, + "learning_rate": 2.6120025275692777e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005623753950931132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026236111647449435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01387111870571971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8278034329414368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16350, + "weighted_flow_matching_loss": 0.01570644611492753, + "weighted_rkd_loss": 0.011482569482177496 + }, + { + "epoch": 4.759965085830666, + "flow_matching_loss": 0.015017289575189352, + "grad_norm": 0.20417580008506775, + "learning_rate": 2.6114475078412212e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005392587743699551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002394876361358911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013180051930248737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8276057004928589, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16360, + "weighted_flow_matching_loss": 0.015017289575189352, + "weighted_rkd_loss": 0.010907831229269504 + }, + { + "epoch": 4.76287459994181, + "flow_matching_loss": 0.01465085968375206, + "grad_norm": 0.21114039421081543, + "learning_rate": 2.6108921504847575e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060388858430087565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028699856135062872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014947757590562105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8274078905582428, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16370, + "weighted_flow_matching_loss": 0.01465085968375206, + "weighted_rkd_loss": 0.012367895105853677 + }, + { + "epoch": 4.765784114052953, + "flow_matching_loss": 0.0129075575619936, + "grad_norm": 0.14915265142917633, + "learning_rate": 2.6103364556685902e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006270201969891786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002757443266455084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015297847241163254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8272099673748017, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16380, + "weighted_flow_matching_loss": 0.0129075575619936, + "weighted_rkd_loss": 0.012654507346451283 + }, + { + "epoch": 4.768693628164097, + "flow_matching_loss": 0.014543082565069199, + "grad_norm": 0.18226084113121033, + "learning_rate": 2.6097804235615242e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006671647215262056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030432471772655844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016386541724205016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8270119607448578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16390, + "weighted_flow_matching_loss": 0.014543082565069199, + "weighted_rkd_loss": 0.013551868963986635 + }, + { + "epoch": 4.77160314227524, + "flow_matching_loss": 0.013621060363948344, + "grad_norm": 0.1659657061100006, + "learning_rate": 2.6092240543324676e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005207253410480917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022877910174429416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01270229769870639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8268138706684113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16400, + "weighted_flow_matching_loss": 0.013621060363948344, + "weighted_rkd_loss": 0.010502440249547362 + }, + { + "epoch": 4.774512656386383, + "flow_matching_loss": 0.013577501196414233, + "grad_norm": 0.17020615935325623, + "learning_rate": 2.6086673481504303e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005876035895198584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024243526277132332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014176424499601126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8266156852245331, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16410, + "weighted_flow_matching_loss": 0.013577501196414233, + "weighted_rkd_loss": 0.011718415841460227 + }, + { + "epoch": 4.777422170497527, + "flow_matching_loss": 0.013572080247104167, + "grad_norm": 0.15875522792339325, + "learning_rate": 2.608110305184526e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00630012433975935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002690801804419607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015291050169616937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8264174163341522, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16420, + "weighted_flow_matching_loss": 0.013572080247104167, + "weighted_rkd_loss": 0.01263678865507245 + }, + { + "epoch": 4.7803316846086705, + "flow_matching_loss": 0.013180877920240164, + "grad_norm": 0.1435595154762268, + "learning_rate": 2.6075529256039687e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005484089208766818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024525434826500716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013420721981674433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8262190520763397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16430, + "weighted_flow_matching_loss": 0.013180877920240164, + "weighted_rkd_loss": 0.011088503617793321 + }, + { + "epoch": 4.783241198719814, + "flow_matching_loss": 0.012472815997898578, + "grad_norm": 0.199574276804924, + "learning_rate": 2.6069952095780756e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00595484902150929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002661707974039018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014571405947208405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.826020622253418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16440, + "weighted_flow_matching_loss": 0.012472815997898578, + "weighted_rkd_loss": 0.012036300357431173 + }, + { + "epoch": 4.786150712830957, + "flow_matching_loss": 0.014308930654078723, + "grad_norm": 0.20421850681304932, + "learning_rate": 2.6064371572762667e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006194528564810753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025835029082372785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014972560200840234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8258220613002777, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16450, + "weighted_flow_matching_loss": 0.014308930654078723, + "weighted_rkd_loss": 0.012364702764898539 + }, + { + "epoch": 4.789060226942101, + "flow_matching_loss": 0.012238510884344578, + "grad_norm": 0.23112830519676208, + "learning_rate": 2.605878768868063e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005854322109371424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002517184743192047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014225829020142555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8256234347820282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16460, + "weighted_flow_matching_loss": 0.012238510884344578, + "weighted_rkd_loss": 0.01174508137628436 + }, + { + "epoch": 4.7919697410532445, + "flow_matching_loss": 0.01331252995878458, + "grad_norm": 0.17497149109840393, + "learning_rate": 2.605320044523088e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005716157844290137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024880984332412482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013920414354652166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8254247188568116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16470, + "weighted_flow_matching_loss": 0.01331252995878458, + "weighted_rkd_loss": 0.011490372382104397 + }, + { + "epoch": 4.794879255164387, + "flow_matching_loss": 0.013713316805660725, + "grad_norm": 0.12044532597064972, + "learning_rate": 2.6047609844110683e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00618141230661422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029230354237370195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015285859908908605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8252259135246277, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16480, + "weighted_flow_matching_loss": 0.013713316805660725, + "weighted_rkd_loss": 0.012614336656406521 + }, + { + "epoch": 4.797788769275531, + "flow_matching_loss": 0.012551251612603665, + "grad_norm": 0.1770520955324173, + "learning_rate": 2.604201588701831e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006587501289322973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003007542551495135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016182545013725756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8250270128250122, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16490, + "weighted_flow_matching_loss": 0.012551251612603665, + "weighted_rkd_loss": 0.013351133745163678 + }, + { + "epoch": 4.800698283386675, + "flow_matching_loss": 0.013116326462477446, + "grad_norm": 0.17123834788799286, + "learning_rate": 2.6036418575653057e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005569371860474348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024606568971648813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013599400687962771, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8248280227184296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16500, + "weighted_flow_matching_loss": 0.013116326462477446, + "weighted_rkd_loss": 0.01121718054637313 + }, + { + "epoch": 4.8036077974978175, + "flow_matching_loss": 0.014606920443475247, + "grad_norm": 0.15687398612499237, + "learning_rate": 2.603081791171524e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006441042898222804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002878166316077113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01576025215908885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8246289491653442, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16510, + "weighted_flow_matching_loss": 0.014606920443475247, + "weighted_rkd_loss": 0.012996369693428277 + }, + { + "epoch": 4.806517311608961, + "flow_matching_loss": 0.012990734167397022, + "grad_norm": 0.27194252610206604, + "learning_rate": 2.602521389690619e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00602944721467793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002785868721548468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014844763185828924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8244297862052917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16520, + "weighted_flow_matching_loss": 0.012990734167397022, + "weighted_rkd_loss": 0.012238414864987135 + }, + { + "epoch": 4.809426825720105, + "flow_matching_loss": 0.012333588022738696, + "grad_norm": 0.18140769004821777, + "learning_rate": 2.601960653292827e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006521746376529336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027325530303642155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015776045620441437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.824230533838272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16530, + "weighted_flow_matching_loss": 0.012333588022738696, + "weighted_rkd_loss": 0.013003141526132822 + }, + { + "epoch": 4.812336339831248, + "flow_matching_loss": 0.0141806167550385, + "grad_norm": 0.17779235541820526, + "learning_rate": 2.6013995821484833e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062845366075634955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002882907795719802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015451981127262116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8240311980247498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16540, + "weighted_flow_matching_loss": 0.0141806167550385, + "weighted_rkd_loss": 0.012732858676463366 + }, + { + "epoch": 4.8152458539423915, + "flow_matching_loss": 0.01277127480134368, + "grad_norm": 0.14211644232273102, + "learning_rate": 2.600838176428028e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00533814039081335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002313739014789462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012990019656717778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8238317549228669, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16550, + "weighted_flow_matching_loss": 0.01277127480134368, + "weighted_rkd_loss": 0.010701683908700943 + }, + { + "epoch": 4.818155368053535, + "flow_matching_loss": 0.013504771050065756, + "grad_norm": 0.20358672738075256, + "learning_rate": 2.6002764363020004e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005744267767295241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002513107820414007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014001643285155296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8236322402954102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16560, + "weighted_flow_matching_loss": 0.013504771050065756, + "weighted_rkd_loss": 0.011532171536237002 + }, + { + "epoch": 4.821064882164679, + "flow_matching_loss": 0.014985853526741266, + "grad_norm": 0.15776915848255157, + "learning_rate": 2.5997143619410427e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005687340069562197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002609538065735251, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01398421823978424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8234326422214509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16570, + "weighted_flow_matching_loss": 0.014985853526741266, + "weighted_rkd_loss": 0.011515012755990029 + }, + { + "epoch": 4.823974396275822, + "flow_matching_loss": 0.013629595469683408, + "grad_norm": 0.16151146590709686, + "learning_rate": 2.5991519535158976e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005695509305223822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002644380321726203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01403539888560772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8232329368591309, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16580, + "weighted_flow_matching_loss": 0.013629595469683408, + "weighted_rkd_loss": 0.011554386094212532 + }, + { + "epoch": 4.826883910386965, + "flow_matching_loss": 0.015275141689926386, + "grad_norm": 0.19116918742656708, + "learning_rate": 2.598589211197411e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054994890931993725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023564284201711416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0133554064668715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8230331540107727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16590, + "weighted_flow_matching_loss": 0.015275141689926386, + "weighted_rkd_loss": 0.010991923417896033 + }, + { + "epoch": 4.829793424498109, + "flow_matching_loss": 0.014480243902653456, + "grad_norm": 0.18299217522144318, + "learning_rate": 2.598026135156528e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005728432442992926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002330499654635787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013787364587187767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8228332817554473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16600, + "weighted_flow_matching_loss": 0.014480243902653456, + "weighted_rkd_loss": 0.01134468149393797 + }, + { + "epoch": 4.832702938609252, + "flow_matching_loss": 0.012623845133930445, + "grad_norm": 0.1603710651397705, + "learning_rate": 2.5974627255642974e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005623532040044665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027545738383196296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014001637883484363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8226333200931549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16610, + "weighted_flow_matching_loss": 0.012623845133930445, + "weighted_rkd_loss": 0.011518256552517415 + }, + { + "epoch": 4.835612452720396, + "flow_matching_loss": 0.013980197254568338, + "grad_norm": 0.13880865275859833, + "learning_rate": 2.5968989825918674e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054797830525785685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024004110484384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013359977211803199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8224332571029663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16620, + "weighted_flow_matching_loss": 0.013980197254568338, + "weighted_rkd_loss": 0.010987772885710002 + }, + { + "epoch": 4.838521966831539, + "flow_matching_loss": 0.0144144956022501, + "grad_norm": 0.20169612765312195, + "learning_rate": 2.5963349064104887e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005636407295241952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025101258070208132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01378294052556157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8222331225872039, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16630, + "weighted_flow_matching_loss": 0.0144144956022501, + "weighted_rkd_loss": 0.011332730809226633 + }, + { + "epoch": 4.841431480942683, + "flow_matching_loss": 0.01488074129447341, + "grad_norm": 0.16359968483448029, + "learning_rate": 2.5957704971915116e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005957160750404001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026338861556723715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014548207633197308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8220328688621521, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16640, + "weighted_flow_matching_loss": 0.01488074129447341, + "weighted_rkd_loss": 0.011959121376276017 + }, + { + "epoch": 4.844340995053826, + "flow_matching_loss": 0.013028935063630342, + "grad_norm": 0.1822356879711151, + "learning_rate": 2.5952057551063905e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006408586865290999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002803202369250357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015620376262813807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8218325793743133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16650, + "weighted_flow_matching_loss": 0.013028935063630342, + "weighted_rkd_loss": 0.012837291415780782 + }, + { + "epoch": 4.84725050916497, + "flow_matching_loss": 0.012697302550077439, + "grad_norm": 0.15705081820487976, + "learning_rate": 2.594640680326678e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052715042838826776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022259625024162234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012768970988690852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8216321706771851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16660, + "weighted_flow_matching_loss": 0.012697302550077439, + "weighted_rkd_loss": 0.010491390898823737 + }, + { + "epoch": 4.850160023276112, + "flow_matching_loss": 0.013281539268791675, + "grad_norm": 0.15684793889522552, + "learning_rate": 2.594075273024029e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005930207390338183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002674358745571226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014534773398190736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8214316844940186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16670, + "weighted_flow_matching_loss": 0.013281539268791675, + "weighted_rkd_loss": 0.011939321178942919 + }, + { + "epoch": 4.853069537387256, + "flow_matching_loss": 0.013185745384544135, + "grad_norm": 0.17000173032283783, + "learning_rate": 2.5935095333701994e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005787638621404767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002811253094114363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014386530499905348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8212311029434204, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16680, + "weighted_flow_matching_loss": 0.013185745384544135, + "weighted_rkd_loss": 0.011814686562865972 + }, + { + "epoch": 4.8559790514984, + "flow_matching_loss": 0.015571037586778403, + "grad_norm": 0.18102724850177765, + "learning_rate": 2.5929434615370462e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005960807390511036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002760011563077569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014681626483798027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8210304260253907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16690, + "weighted_flow_matching_loss": 0.015571037586778403, + "weighted_rkd_loss": 0.012054022029042244 + }, + { + "epoch": 4.8588885656095435, + "flow_matching_loss": 0.01415300564840436, + "grad_norm": 0.1677231639623642, + "learning_rate": 2.592377057696527e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005262792995199561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022424545139074327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012768040504306554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8208296775817872, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16700, + "weighted_flow_matching_loss": 0.01415300564840436, + "weighted_rkd_loss": 0.010480352910235524 + }, + { + "epoch": 4.861798079720686, + "flow_matching_loss": 0.013641295582056045, + "grad_norm": 0.19340120255947113, + "learning_rate": 2.5918103220206997e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005767750786617398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002497541520278901, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014033043012022971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8206288397312165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16710, + "weighted_flow_matching_loss": 0.013641295582056045, + "weighted_rkd_loss": 0.011515900027006865 + }, + { + "epoch": 4.86470759383183, + "flow_matching_loss": 0.014214114006608724, + "grad_norm": 0.17918661236763, + "learning_rate": 2.5912432546817247e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005623471550643444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002367410308215767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01361435353755951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8204279005527496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16720, + "weighted_flow_matching_loss": 0.014214114006608724, + "weighted_rkd_loss": 0.01116966139525175 + }, + { + "epoch": 4.867617107942974, + "flow_matching_loss": 0.013987488951534033, + "grad_norm": 0.1585010439157486, + "learning_rate": 2.5906758558518613e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006009454373270273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026679055066779257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014686814229935407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8202268779277802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16730, + "weighted_flow_matching_loss": 0.013987488951534033, + "weighted_rkd_loss": 0.012046440783888102 + }, + { + "epoch": 4.870526622054117, + "flow_matching_loss": 0.014584804698824883, + "grad_norm": 0.12777116894721985, + "learning_rate": 2.5901081257034706e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005011545773595571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023085589869879185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012331650592386723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.820025771856308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16740, + "weighted_flow_matching_loss": 0.014584804698824883, + "weighted_rkd_loss": 0.010112340282648801 + }, + { + "epoch": 4.87343613616526, + "flow_matching_loss": 0.013506491668522358, + "grad_norm": 0.162565678358078, + "learning_rate": 2.5895400644090138e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00548965223133564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023903886671178044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013369693048298359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8198245763778687, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16750, + "weighted_flow_matching_loss": 0.013506491668522358, + "weighted_rkd_loss": 0.010960825905203819 + }, + { + "epoch": 4.876345650276404, + "flow_matching_loss": 0.013491815142333507, + "grad_norm": 0.16251954436302185, + "learning_rate": 2.5889716721410535e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005364547343924641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023545420495793223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013083636667579413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8196232914924622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16760, + "weighted_flow_matching_loss": 0.013491815142333507, + "weighted_rkd_loss": 0.01072361832484603 + }, + { + "epoch": 4.879255164387548, + "flow_matching_loss": 0.01231650160625577, + "grad_norm": 0.19477707147598267, + "learning_rate": 2.5884029490722515e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005717092705890536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024310473818331957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013865232840180397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.819421923160553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16770, + "weighted_flow_matching_loss": 0.01231650160625577, + "weighted_rkd_loss": 0.011361445020884275 + }, + { + "epoch": 4.8821646784986905, + "flow_matching_loss": 0.01483418783172965, + "grad_norm": 0.16006942093372345, + "learning_rate": 2.587833895375371e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005019567115232349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002202979405410588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012242113519459964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8192204654216766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16780, + "weighted_flow_matching_loss": 0.01483418783172965, + "weighted_rkd_loss": 0.010029032360762358 + }, + { + "epoch": 4.885074192609834, + "flow_matching_loss": 0.01585231935605407, + "grad_norm": 0.18610738217830658, + "learning_rate": 2.587264511223276e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005871423287317157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002734436676837504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01447728332132101, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8190189123153686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16790, + "weighted_flow_matching_loss": 0.01585231935605407, + "weighted_rkd_loss": 0.011857165023684502 + }, + { + "epoch": 4.887983706720978, + "flow_matching_loss": 0.01365164453163743, + "grad_norm": 0.20388418436050415, + "learning_rate": 2.5866947967889295e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00584425306878984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002650534617714584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014339040778577328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8188172817230225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16800, + "weighted_flow_matching_loss": 0.01365164453163743, + "weighted_rkd_loss": 0.011740940250456333 + }, + { + "epoch": 4.890893220832121, + "flow_matching_loss": 0.012940295599400997, + "grad_norm": 0.14881308376789093, + "learning_rate": 2.586124752245397e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006241959519684315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002734915807377547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015218835044652223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8186155557632446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16810, + "weighted_flow_matching_loss": 0.012940295599400997, + "weighted_rkd_loss": 0.012458310928195715 + }, + { + "epoch": 4.8938027349432645, + "flow_matching_loss": 0.015187345072627067, + "grad_norm": 0.19034458696842194, + "learning_rate": 2.585554377765842e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058350440114736555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025381394079886377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01420822711661458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8184137463569641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16820, + "weighted_flow_matching_loss": 0.015187345072627067, + "weighted_rkd_loss": 0.011628157272934914 + }, + { + "epoch": 4.896712249054408, + "flow_matching_loss": 0.013210983853787184, + "grad_norm": 0.20592860877513885, + "learning_rate": 2.584983673523529e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005727935582399368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002678416413255036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01413428746163845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8182118475437165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16830, + "weighted_flow_matching_loss": 0.013210983853787184, + "weighted_rkd_loss": 0.011564859468489886 + }, + { + "epoch": 4.899621763165551, + "flow_matching_loss": 0.014209663216024637, + "grad_norm": 0.210158571600914, + "learning_rate": 2.5844126396918236e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005953340558335185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002769477048423141, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014676158223301173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.818009877204895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16840, + "weighted_flow_matching_loss": 0.014209663216024637, + "weighted_rkd_loss": 0.012005266081541777 + }, + { + "epoch": 4.902531277276695, + "flow_matching_loss": 0.013382307160645724, + "grad_norm": 0.17965207993984222, + "learning_rate": 2.5838412764441905e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006193952495232224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025284336530603468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014916338492184877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8178078055381774, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16850, + "weighted_flow_matching_loss": 0.013382307160645724, + "weighted_rkd_loss": 0.012198692094534636 + }, + { + "epoch": 4.905440791387838, + "flow_matching_loss": 0.01522020362317562, + "grad_norm": 0.16690826416015625, + "learning_rate": 2.583269583954195e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062249608803540465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002701466262806207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015151388105005026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8176056444644928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16860, + "weighted_flow_matching_loss": 0.01522020362317562, + "weighted_rkd_loss": 0.012387872859835624 + }, + { + "epoch": 4.908350305498981, + "flow_matching_loss": 0.013734316173940896, + "grad_norm": 0.15109208226203918, + "learning_rate": 2.5826975623955017e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005676359357312322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023722984013147654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013725017104297877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8174033880233764, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16870, + "weighted_flow_matching_loss": 0.013734316173940896, + "weighted_rkd_loss": 0.011218887753784657 + }, + { + "epoch": 4.911259819610125, + "flow_matching_loss": 0.01285766325891018, + "grad_norm": 0.17800085246562958, + "learning_rate": 2.5821252119418762e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006296876817941666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002899393020197749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015493146702647209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8172010719776154, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16880, + "weighted_flow_matching_loss": 0.01285766325891018, + "weighted_rkd_loss": 0.01266098041087389 + }, + { + "epoch": 4.914169333721269, + "flow_matching_loss": 0.013249451015144586, + "grad_norm": 0.20814836025238037, + "learning_rate": 2.581552532767183e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005678725941106677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025226361583918335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013880088087171316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8169986486434937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16890, + "weighted_flow_matching_loss": 0.013249451015144586, + "weighted_rkd_loss": 0.011340044997632503 + }, + { + "epoch": 4.917078847832412, + "flow_matching_loss": 0.013507269229739905, + "grad_norm": 0.21485178172588348, + "learning_rate": 2.580979525045387e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005414162902161479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002538949274457991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013367275055497885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8167961418628693, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16900, + "weighted_flow_matching_loss": 0.013507269229739905, + "weighted_rkd_loss": 0.010918303858488798 + }, + { + "epoch": 4.919988361943555, + "flow_matching_loss": 0.014162975084036588, + "grad_norm": 0.13491639494895935, + "learning_rate": 2.580406188950553e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005515294475480914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025392687064595522, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013569857832044363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8165935277938843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16910, + "weighted_flow_matching_loss": 0.014162975084036588, + "weighted_rkd_loss": 0.011081182211637498 + }, + { + "epoch": 4.922897876054699, + "flow_matching_loss": 0.014736669417470694, + "grad_norm": 0.14609812200069427, + "learning_rate": 2.5798325246568453e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005440736142918468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023259637411683796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013207436073571444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8163908660411835, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16920, + "weighted_flow_matching_loss": 0.014736669417470694, + "weighted_rkd_loss": 0.01078243162482977 + }, + { + "epoch": 4.925807390165843, + "flow_matching_loss": 0.012261358881369233, + "grad_norm": 0.170848086476326, + "learning_rate": 2.579258532338528e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056853446178138254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002334409800823778, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013705099001526833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8161881029605865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16930, + "weighted_flow_matching_loss": 0.012261358881369233, + "weighted_rkd_loss": 0.011185920145362616 + }, + { + "epoch": 4.928716904276985, + "flow_matching_loss": 0.014550551027059554, + "grad_norm": 0.20061932504177094, + "learning_rate": 2.5786842121699644e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006102774199098349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024737462867051364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014679294731467962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8159852504730225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16940, + "weighted_flow_matching_loss": 0.014550551027059554, + "weighted_rkd_loss": 0.011978060472756624 + }, + { + "epoch": 4.931626418388129, + "flow_matching_loss": 0.014476583059877157, + "grad_norm": 0.18677806854248047, + "learning_rate": 2.5781095643256186e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006103095063008368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002871808037161827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015077998116612434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8157823085784912, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16950, + "weighted_flow_matching_loss": 0.014476583059877157, + "weighted_rkd_loss": 0.012300348561257123 + }, + { + "epoch": 4.934535932499273, + "flow_matching_loss": 0.012387855863198637, + "grad_norm": 0.1879866123199463, + "learning_rate": 2.577534588980053e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006034613307565451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026605836348608134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014729810412973166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8155792891979218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16960, + "weighted_flow_matching_loss": 0.012387855863198637, + "weighted_rkd_loss": 0.012013363186269999 + }, + { + "epoch": 4.9374454466104165, + "flow_matching_loss": 0.013521172106266022, + "grad_norm": 0.1905381828546524, + "learning_rate": 2.5769592863079298e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006302089523524046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002992418734356761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015596597641706466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8153761684894562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16970, + "weighted_flow_matching_loss": 0.013521172106266022, + "weighted_rkd_loss": 0.01271714698523283 + }, + { + "epoch": 4.940354960721559, + "flow_matching_loss": 0.0140532816760242, + "grad_norm": 0.1591569036245346, + "learning_rate": 2.576383656484011e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006364337354898453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002952960238326341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015681634843349456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8151729881763459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16980, + "weighted_flow_matching_loss": 0.0140532816760242, + "weighted_rkd_loss": 0.012783258687704802 + }, + { + "epoch": 4.943264474832703, + "flow_matching_loss": 0.015753765124827623, + "grad_norm": 0.1588994860649109, + "learning_rate": 2.5758076996831572e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005699990829452872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002528158889617771, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013928140699863433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8149696886539459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16990, + "weighted_flow_matching_loss": 0.015753765124827623, + "weighted_rkd_loss": 0.011351120006293059 + }, + { + "epoch": 4.946173988943847, + "flow_matching_loss": 0.013307852111756802, + "grad_norm": 0.17138884961605072, + "learning_rate": 2.5752314160803295e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005569915333762765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025385372224263846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013678368087857962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8147663176059723, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17000, + "weighted_flow_matching_loss": 0.013307852111756802, + "weighted_rkd_loss": 0.011144762299954892 + }, + { + "epoch": 4.94908350305499, + "flow_matching_loss": 0.014044974464923143, + "grad_norm": 0.20833536982536316, + "learning_rate": 2.5746548058505873e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005634853197261691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025250687613151968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013794775120913982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.814562863111496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17010, + "weighted_flow_matching_loss": 0.014044974464923143, + "weighted_rkd_loss": 0.011236584978178144 + }, + { + "epoch": 4.951993017166133, + "flow_matching_loss": 0.012770852074027061, + "grad_norm": 0.14793661236763, + "learning_rate": 2.57407786916909e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00552183291874826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002561845874879509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013605511933565139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8143593192100524, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17020, + "weighted_flow_matching_loss": 0.012770852074027061, + "weighted_rkd_loss": 0.011079756077378988 + }, + { + "epoch": 4.954902531277277, + "flow_matching_loss": 0.012529023736715317, + "grad_norm": 0.186561718583107, + "learning_rate": 2.573500606211095e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005812353733927011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002570956340059638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014195663947612047, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8141556859016419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17030, + "weighted_flow_matching_loss": 0.012529023736715317, + "weighted_rkd_loss": 0.011557369539514185 + }, + { + "epoch": 4.95781204538842, + "flow_matching_loss": 0.015343482419848443, + "grad_norm": 0.14801397919654846, + "learning_rate": 2.57292301715196e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053511792561039325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024310584412887692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013133416883647442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8139519691467285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17040, + "weighted_flow_matching_loss": 0.015343482419848443, + "weighted_rkd_loss": 0.010689993388950825 + }, + { + "epoch": 4.9607215594995635, + "flow_matching_loss": 0.014044040720909834, + "grad_norm": 0.19256184995174408, + "learning_rate": 2.572345102167142e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006625287607312203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028371913358569145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01608776655048132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8137481808662415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17050, + "weighted_flow_matching_loss": 0.014044040720909834, + "weighted_rkd_loss": 0.01309136524796486 + }, + { + "epoch": 4.963631073610707, + "flow_matching_loss": 0.013115029316395521, + "grad_norm": 0.1826266497373581, + "learning_rate": 2.571766861432195e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005871189618483185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027424272615462542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014484806451946497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8135442793369293, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17060, + "weighted_flow_matching_loss": 0.013115029316395521, + "weighted_rkd_loss": 0.011784050893038512 + }, + { + "epoch": 4.96654058772185, + "flow_matching_loss": 0.014869096782058477, + "grad_norm": 0.16209523379802704, + "learning_rate": 2.5711882951227738e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005364385806024075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023088664398528635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013037637900561094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8133403062820435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17070, + "weighted_flow_matching_loss": 0.014869096782058477, + "weighted_rkd_loss": 0.010604032315313816 + }, + { + "epoch": 4.969450101832994, + "flow_matching_loss": 0.012193317525088787, + "grad_norm": 0.16730816662311554, + "learning_rate": 2.570609403414632e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004899888741783797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021319397841580213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011931717302650214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.813136237859726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17080, + "weighted_flow_matching_loss": 0.012193317525088787, + "weighted_rkd_loss": 0.009702071966603398 + }, + { + "epoch": 4.9723596159441374, + "flow_matching_loss": 0.013721208460628987, + "grad_norm": 0.22372780740261078, + "learning_rate": 2.5700301864836212e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00714207049459219, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003073100873734802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017357241734862326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8129320979118347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17090, + "weighted_flow_matching_loss": 0.013721208460628987, + "weighted_rkd_loss": 0.014110112935304642 + }, + { + "epoch": 4.975269130055281, + "flow_matching_loss": 0.013466698303818702, + "grad_norm": 0.1732117235660553, + "learning_rate": 2.569450644505692e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006133759557269513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027526196790859103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015020138863474131, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8127278625965119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17100, + "weighted_flow_matching_loss": 0.013466698303818702, + "weighted_rkd_loss": 0.012207265431061386 + }, + { + "epoch": 4.978178644166424, + "flow_matching_loss": 0.013696006406098605, + "grad_norm": 0.19033922255039215, + "learning_rate": 2.5688707776568943e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005878596752882004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025749179418198763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014332111552357673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8125235497951507, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17110, + "weighted_flow_matching_loss": 0.013696006406098605, + "weighted_rkd_loss": 0.011645109113305806 + }, + { + "epoch": 4.981088158277568, + "flow_matching_loss": 0.013558633252978325, + "grad_norm": 0.19671708345413208, + "learning_rate": 2.5682905861133764e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005251540476456285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023254982428625227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012828579172492027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8123191297054291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17120, + "weighted_flow_matching_loss": 0.013558633252978325, + "weighted_rkd_loss": 0.010420865006744861 + }, + { + "epoch": 4.983997672388711, + "flow_matching_loss": 0.015755977388471364, + "grad_norm": 0.18656566739082336, + "learning_rate": 2.567710070051385e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062401705421507355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002645351295359433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015125692449510097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8121146559715271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17130, + "weighted_flow_matching_loss": 0.015755977388471364, + "weighted_rkd_loss": 0.012283838726580143 + }, + { + "epoch": 4.986907186499854, + "flow_matching_loss": 0.01571832187473774, + "grad_norm": 0.16635727882385254, + "learning_rate": 2.5671292296472652e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005876697762869298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002682733675464988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014436129201203584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.811910080909729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17140, + "weighted_flow_matching_loss": 0.01571832187473774, + "weighted_rkd_loss": 0.011720925569534302 + }, + { + "epoch": 4.989816700610998, + "flow_matching_loss": 0.015330823231488466, + "grad_norm": 0.1958642601966858, + "learning_rate": 2.5665480650774616e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005689006298780441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002670525712892413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014048538263887167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8117054283618927, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17150, + "weighted_flow_matching_loss": 0.015330823231488466, + "weighted_rkd_loss": 0.01140326289460063 + }, + { + "epoch": 4.992726214722142, + "flow_matching_loss": 0.01406589411199093, + "grad_norm": 0.15855880081653595, + "learning_rate": 2.565966576518517e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005930118728429079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027679397258907556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0146281773224473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8115006804466247, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17160, + "weighted_flow_matching_loss": 0.01406589411199093, + "weighted_rkd_loss": 0.01187075162306428 + }, + { + "epoch": 4.995635728833285, + "flow_matching_loss": 0.014925509504973889, + "grad_norm": 0.15307992696762085, + "learning_rate": 2.5653847641470706e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006070049712434411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002645307581406087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014785406831651925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8112958490848541, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17170, + "weighted_flow_matching_loss": 0.014925509504973889, + "weighted_rkd_loss": 0.011995434015989303 + }, + { + "epoch": 4.998545242944428, + "flow_matching_loss": 0.013173841498792171, + "grad_norm": 0.15053008496761322, + "learning_rate": 2.564802628139863e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005409735161811114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002289421879686415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013108892273157835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8110909283161163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17180, + "weighted_flow_matching_loss": 0.013173841498792171, + "weighted_rkd_loss": 0.010632564127445222 + }, + { + "epoch": 5.001454757055572, + "flow_matching_loss": 0.01385658485814929, + "grad_norm": 0.16521640121936798, + "learning_rate": 2.5642201686737318e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005700064054690301, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026179646025411784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014018092397600412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8108859062194824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17190, + "weighted_flow_matching_loss": 0.01385658485814929, + "weighted_rkd_loss": 0.011367135634645819 + }, + { + "epoch": 5.004364271166716, + "flow_matching_loss": 0.01341583402827382, + "grad_norm": 0.17966404557228088, + "learning_rate": 2.563637385925612e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005861971247941256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002659005578607321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01438294816762209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8106808423995971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17200, + "weighted_flow_matching_loss": 0.01341583402827382, + "weighted_rkd_loss": 0.011659989133477211 + }, + { + "epoch": 5.007273785277858, + "flow_matching_loss": 0.01408664220944047, + "grad_norm": 0.1813056617975235, + "learning_rate": 2.5630542800725388e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0050373054575175045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022356147528626025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012310225516557693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8104756712913513, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17210, + "weighted_flow_matching_loss": 0.01408664220944047, + "weighted_rkd_loss": 0.009977084212005138 + }, + { + "epoch": 5.010183299389002, + "flow_matching_loss": 0.014933736715465783, + "grad_norm": 0.14576798677444458, + "learning_rate": 2.562470851291643e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005902389669790864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025421838974580167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01434696326032281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8102704167366028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17220, + "weighted_flow_matching_loss": 0.014933736715465783, + "weighted_rkd_loss": 0.011624892707914114 + }, + { + "epoch": 5.013092813500146, + "flow_matching_loss": 0.01328926724381745, + "grad_norm": 0.16785679757595062, + "learning_rate": 2.561887099760156e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005281440168619156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002309200179297477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012872080411761999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.810065072774887, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17230, + "weighted_flow_matching_loss": 0.01328926724381745, + "weighted_rkd_loss": 0.010427219234406948 + }, + { + "epoch": 5.016002327611289, + "flow_matching_loss": 0.016677141189575195, + "grad_norm": 0.15517929196357727, + "learning_rate": 2.5613030256554054e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005329414224252105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002381879254244268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0130407078191638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8098596453666687, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17240, + "weighted_flow_matching_loss": 0.016677141189575195, + "weighted_rkd_loss": 0.010561098903417587 + }, + { + "epoch": 5.018911841722432, + "flow_matching_loss": 0.014291747007519007, + "grad_norm": 0.18529784679412842, + "learning_rate": 2.5607186291548174e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005094658047892153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021414405200630426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012330756708979606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8096541285514831, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17250, + "weighted_flow_matching_loss": 0.014291747007519007, + "weighted_rkd_loss": 0.00998369618318975 + }, + { + "epoch": 5.021821355833576, + "flow_matching_loss": 0.014489809330552817, + "grad_norm": 0.17405369877815247, + "learning_rate": 2.5601339104359167e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005560968513600528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025167349143885077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013638671673834323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8094485104084015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17260, + "weighted_flow_matching_loss": 0.014489809330552817, + "weighted_rkd_loss": 0.011039823899045587 + }, + { + "epoch": 5.024730869944719, + "flow_matching_loss": 0.015121553745120763, + "grad_norm": 0.21143130958080292, + "learning_rate": 2.559548869676325e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005512393126264214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002466949715744704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013491735979914665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8092428505420685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17270, + "weighted_flow_matching_loss": 0.015121553745120763, + "weighted_rkd_loss": 0.010918007418513297 + }, + { + "epoch": 5.0276403840558626, + "flow_matching_loss": 0.012668988946825267, + "grad_norm": 0.19402112066745758, + "learning_rate": 2.558963507053763e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048708278918638825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002102006075438112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011843661963939666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8090370833873749, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17280, + "weighted_flow_matching_loss": 0.012668988946825267, + "weighted_rkd_loss": 0.009581955941393971 + }, + { + "epoch": 5.030549898167006, + "flow_matching_loss": 0.014453008491545915, + "grad_norm": 0.18215689063072205, + "learning_rate": 2.5583778227460476e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005783120985142887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024468684452585874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014013110473752022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8088312327861786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17290, + "weighted_flow_matching_loss": 0.014453008491545915, + "weighted_rkd_loss": 0.011334201786667109 + }, + { + "epoch": 5.03345941227815, + "flow_matching_loss": 0.014020064659416676, + "grad_norm": 0.14089582860469818, + "learning_rate": 2.5577918169310943e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066125336801633235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002907497121486813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016132564563304187, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8086252987384797, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17300, + "weighted_flow_matching_loss": 0.014020064659416676, + "weighted_rkd_loss": 0.013045253697782755 + }, + { + "epoch": 5.036368926389293, + "flow_matching_loss": 0.012996522709727287, + "grad_norm": 0.18469999730587006, + "learning_rate": 2.5572054897869167e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005303282476961613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002402966911904514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01300953170284629, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8084192752838135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17310, + "weighted_flow_matching_loss": 0.012996522709727287, + "weighted_rkd_loss": 0.010517132375389338 + }, + { + "epoch": 5.0392784405004365, + "flow_matching_loss": 0.012882381118834019, + "grad_norm": 0.1477389633655548, + "learning_rate": 2.556618841491625e-05, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004742504865862429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002010417520068586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011495427368208766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8082131624221802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17320, + "weighted_flow_matching_loss": 0.012882381118834019, + "weighted_rkd_loss": 0.009290755027905107 + }, + { + "epoch": 5.04218795461158, + "flow_matching_loss": 0.012726509245112538, + "grad_norm": 0.1699957698583603, + "learning_rate": 2.5560318722234276e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005309490743093193, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002400545054115355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013019526656717063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.808007001876831, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17330, + "weighted_flow_matching_loss": 0.012726509245112538, + "weighted_rkd_loss": 0.010519876424223185 + }, + { + "epoch": 5.045097468722723, + "flow_matching_loss": 0.014814989920705558, + "grad_norm": 0.21935343742370605, + "learning_rate": 2.5554445821606306e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006142189539968967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027889888151548803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015073367767035961, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8078006982803345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17340, + "weighted_flow_matching_loss": 0.014814989920705558, + "weighted_rkd_loss": 0.0121763177216053 + }, + { + "epoch": 5.048006982833867, + "flow_matching_loss": 0.015147662162780762, + "grad_norm": 0.19169870018959045, + "learning_rate": 2.554856971481637e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005263770371675492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024219217128120364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012949462421238423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8075943470001221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17350, + "weighted_flow_matching_loss": 0.015147662162780762, + "weighted_rkd_loss": 0.010457900539040565 + }, + { + "epoch": 5.05091649694501, + "flow_matching_loss": 0.013282474875450135, + "grad_norm": 0.17382512986660004, + "learning_rate": 2.554269040364947e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005541544500738382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002408313122577965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013491402380168438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8073879063129425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17360, + "weighted_flow_matching_loss": 0.013282474875450135, + "weighted_rkd_loss": 0.010892807506024837 + }, + { + "epoch": 5.053826011056153, + "flow_matching_loss": 0.015195154119282961, + "grad_norm": 0.1632007509469986, + "learning_rate": 2.553680788989159e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006002400862053036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026880238438025117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014692825451493263, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8071813762187958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17370, + "weighted_flow_matching_loss": 0.015195154119282961, + "weighted_rkd_loss": 0.011859794612973928 + }, + { + "epoch": 5.056735525167297, + "flow_matching_loss": 0.01325204586610198, + "grad_norm": 0.19613364338874817, + "learning_rate": 2.5530922175329677e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055446352809667586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024339719791896643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013523242622613906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8069747626781464, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17380, + "weighted_flow_matching_loss": 0.01325204586610198, + "weighted_rkd_loss": 0.010912938974797726 + }, + { + "epoch": 5.059645039278441, + "flow_matching_loss": 0.014293221849948167, + "grad_norm": 0.2106294184923172, + "learning_rate": 2.5525033261751667e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006527315266430378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003304253343958408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016358884051442146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8067680656909942, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17390, + "weighted_flow_matching_loss": 0.014293221849948167, + "weighted_rkd_loss": 0.013197991531342268 + }, + { + "epoch": 5.0625545533895835, + "flow_matching_loss": 0.013405084144324064, + "grad_norm": 0.15923061966896057, + "learning_rate": 2.5519141150946445e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006071265833452344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002745043789036572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014887575525790453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8065612852573395, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17400, + "weighted_flow_matching_loss": 0.013405084144324064, + "weighted_rkd_loss": 0.01200782977975905 + }, + { + "epoch": 5.065464067500727, + "flow_matching_loss": 0.01323374817147851, + "grad_norm": 0.1762947142124176, + "learning_rate": 2.551324584470389e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005760167352855205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025072140735574065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014027548674494027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.806354421377182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17410, + "weighted_flow_matching_loss": 0.01323374817147851, + "weighted_rkd_loss": 0.011311176139861346 + }, + { + "epoch": 5.068373581611871, + "flow_matching_loss": 0.01294002840295434, + "grad_norm": 0.15025554597377777, + "learning_rate": 2.5507347344814825e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005977193079888821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026104256743565203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014564811997115613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8061474800109864, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17420, + "weighted_flow_matching_loss": 0.01294002840295434, + "weighted_rkd_loss": 0.011741332709789276 + }, + { + "epoch": 5.071283095723015, + "flow_matching_loss": 0.014441039320081473, + "grad_norm": 0.14370374381542206, + "learning_rate": 2.5501445653071072e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005285450676456094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002236053440719843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012806954700499773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.805940443277359, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17430, + "weighted_flow_matching_loss": 0.014441039320081473, + "weighted_rkd_loss": 0.010321637894958258 + }, + { + "epoch": 5.074192609834157, + "flow_matching_loss": 0.014794039819389582, + "grad_norm": 0.1690169721841812, + "learning_rate": 2.5495540771265406e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005590812209993601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023179286974482237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0134995530359447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8057333290576935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17440, + "weighted_flow_matching_loss": 0.014794039819389582, + "weighted_rkd_loss": 0.01087704161182046 + }, + { + "epoch": 5.077102123945301, + "flow_matching_loss": 0.014517757669091224, + "grad_norm": 0.1686968207359314, + "learning_rate": 2.5489632701191565e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005577923404052854, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025861985748633744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013742045406252146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8055261254310608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17450, + "weighted_flow_matching_loss": 0.014517757669091224, + "weighted_rkd_loss": 0.011069573648273944 + }, + { + "epoch": 5.080011638056445, + "flow_matching_loss": 0.013448304031044245, + "grad_norm": 0.17905114591121674, + "learning_rate": 2.5483721444644276e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005739647429436445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002601403882727027, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014080699067562819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8053188502788544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17460, + "weighted_flow_matching_loss": 0.013448304031044245, + "weighted_rkd_loss": 0.011339427344501019 + }, + { + "epoch": 5.082921152167588, + "flow_matching_loss": 0.01421155510470271, + "grad_norm": 0.1665896326303482, + "learning_rate": 2.5477807003419222e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056274755159392955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028162808739580214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014071232080459595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8051114678382874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17470, + "weighted_flow_matching_loss": 0.01421155510470271, + "weighted_rkd_loss": 0.011328885052353143 + }, + { + "epoch": 5.085830666278731, + "flow_matching_loss": 0.014418425690382718, + "grad_norm": 0.18017178773880005, + "learning_rate": 2.547188937931305e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005258016986772418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023321836488321424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012848217599093914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8049040317535401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17480, + "weighted_flow_matching_loss": 0.014418425690382718, + "weighted_rkd_loss": 0.010341528709977866 + }, + { + "epoch": 5.088740180389875, + "flow_matching_loss": 0.01361497761681676, + "grad_norm": 0.1543046236038208, + "learning_rate": 2.5465968574123372e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006044027442112565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025355402496643363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014623595122247934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8046965003013611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17490, + "weighted_flow_matching_loss": 0.01361497761681676, + "weighted_rkd_loss": 0.011767564155161381 + }, + { + "epoch": 5.091649694501018, + "flow_matching_loss": 0.014600502513349056, + "grad_norm": 0.15524916350841522, + "learning_rate": 2.5460044589648787e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005969863990321756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002549255336634815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014488983433693647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8044888794422149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17500, + "weighted_flow_matching_loss": 0.014600502513349056, + "weighted_rkd_loss": 0.01165636656805873 + }, + { + "epoch": 5.094559208612162, + "flow_matching_loss": 0.012696125078946352, + "grad_norm": 0.15015637874603271, + "learning_rate": 2.5454117427688833e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055845414288342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002543343952856958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013712426740676164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8042811751365662, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17510, + "weighted_flow_matching_loss": 0.012696125078946352, + "weighted_rkd_loss": 0.011028653755784035 + }, + { + "epoch": 5.097468722723305, + "flow_matching_loss": 0.013529642578214407, + "grad_norm": 0.15278246998786926, + "learning_rate": 2.5448187090044033e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059250487945973875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026743011199869215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014524398650974036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8040733933448792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17520, + "weighted_flow_matching_loss": 0.013529642578214407, + "weighted_rkd_loss": 0.011678723897784949 + }, + { + "epoch": 5.100378236834449, + "flow_matching_loss": 0.013619101606309414, + "grad_norm": 0.17511266469955444, + "learning_rate": 2.544225357851586e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057838988490402695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025426231557503344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014110420737415552, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8038655281066894, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17530, + "weighted_flow_matching_loss": 0.013619101606309414, + "weighted_rkd_loss": 0.011342868395149707 + }, + { + "epoch": 5.103287750945592, + "flow_matching_loss": 0.013254643138498067, + "grad_norm": 0.15696460008621216, + "learning_rate": 2.543631689490676e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007120663812384009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003671843255870044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017913170903921128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8036575615406036, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17540, + "weighted_flow_matching_loss": 0.013254643138498067, + "weighted_rkd_loss": 0.014395987801253795 + }, + { + "epoch": 5.1061972650567355, + "flow_matching_loss": 0.013530711922794581, + "grad_norm": 0.1903221607208252, + "learning_rate": 2.543037704102015e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006025354098528624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027720659738406537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014822774194180965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8034495353698731, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17550, + "weighted_flow_matching_loss": 0.013530711922794581, + "weighted_rkd_loss": 0.011909274850040674 + }, + { + "epoch": 5.109106779167879, + "flow_matching_loss": 0.012964142113924026, + "grad_norm": 0.17006129026412964, + "learning_rate": 2.5424434018660386e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068030782975256445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030324414605274797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016638597939163446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8032414138317108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17560, + "weighted_flow_matching_loss": 0.012964142113924026, + "weighted_rkd_loss": 0.013364843465387822 + }, + { + "epoch": 5.112016293279022, + "flow_matching_loss": 0.013990082498639821, + "grad_norm": 0.19342203438282013, + "learning_rate": 2.541848782963281e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006255102902650833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002677485055755824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015187690872699023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8030332267284394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17570, + "weighted_flow_matching_loss": 0.013990082498639821, + "weighted_rkd_loss": 0.012196110095828772 + }, + { + "epoch": 5.114925807390166, + "flow_matching_loss": 0.013577992748469114, + "grad_norm": 0.15493734180927277, + "learning_rate": 2.5412538475743714e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005750003270804882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025139273377135394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014013933949172497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8028249382972718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17580, + "weighted_flow_matching_loss": 0.013577992748469114, + "weighted_rkd_loss": 0.011250670719891786 + }, + { + "epoch": 5.1178353215013095, + "flow_matching_loss": 0.014738747384399176, + "grad_norm": 0.13233640789985657, + "learning_rate": 2.540658595880036e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005891270330175757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026549470145255326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01443748790770769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8026165664196014, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17590, + "weighted_flow_matching_loss": 0.014738747384399176, + "weighted_rkd_loss": 0.011587775126099587 + }, + { + "epoch": 5.120744835612452, + "flow_matching_loss": 0.012812416907399893, + "grad_norm": 0.20542773604393005, + "learning_rate": 2.540063028061096e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005768867302685976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027131262118928134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01425086073577404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8024081230163574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17600, + "weighted_flow_matching_loss": 0.012812416907399893, + "weighted_rkd_loss": 0.011435056384652854 + }, + { + "epoch": 5.123654349723596, + "flow_matching_loss": 0.0144527074880898, + "grad_norm": 0.1764289140701294, + "learning_rate": 2.5394671442984692e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052421984262764456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023840042180381716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012868401035666466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8021995902061463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17610, + "weighted_flow_matching_loss": 0.0144527074880898, + "weighted_rkd_loss": 0.01032306468114257 + }, + { + "epoch": 5.12656386383474, + "flow_matching_loss": 0.013183626439422369, + "grad_norm": 0.15618866682052612, + "learning_rate": 2.5388709447731696e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005423186672851443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002359561249613762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01320593450218439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8019909739494324, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17620, + "weighted_flow_matching_loss": 0.013183626439422369, + "weighted_rkd_loss": 0.01059103487059474 + }, + { + "epoch": 5.129473377945883, + "flow_matching_loss": 0.013612647633999586, + "grad_norm": 0.17427733540534973, + "learning_rate": 2.5382744296663068e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00520290844142437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022233555908314885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012629172392189502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8017822742462158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17630, + "weighted_flow_matching_loss": 0.013612647633999586, + "weighted_rkd_loss": 0.01012581242248416 + }, + { + "epoch": 5.132382892057026, + "flow_matching_loss": 0.013356698118150235, + "grad_norm": 0.17135170102119446, + "learning_rate": 2.5376775991590865e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054722405970096585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002474821940995753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013419303204864263, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8015734851360321, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17640, + "weighted_flow_matching_loss": 0.013356698118150235, + "weighted_rkd_loss": 0.010756567306816577 + }, + { + "epoch": 5.13529240616817, + "flow_matching_loss": 0.0155546342022717, + "grad_norm": 0.14951805770397186, + "learning_rate": 2.5370804534328097e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005277833784930408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022463603992946444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0128020279109478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8013646304607391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17650, + "weighted_flow_matching_loss": 0.0155546342022717, + "weighted_rkd_loss": 0.010259074997156859 + }, + { + "epoch": 5.138201920279314, + "flow_matching_loss": 0.01298888511955738, + "grad_norm": 0.13837173581123352, + "learning_rate": 2.5364829926688736e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058471304830163716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024596985545940696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014153959508985282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8011556804180145, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17660, + "weighted_flow_matching_loss": 0.01298888511955738, + "weighted_rkd_loss": 0.011339480336755515 + }, + { + "epoch": 5.1411114343904565, + "flow_matching_loss": 0.014728105906397105, + "grad_norm": 0.1829495131969452, + "learning_rate": 2.5358852170487713e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005801467737182975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002661044581327587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01426397990435362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8009466469287873, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17670, + "weighted_flow_matching_loss": 0.014728105906397105, + "weighted_rkd_loss": 0.011424613278359174 + }, + { + "epoch": 5.1440209485016, + "flow_matching_loss": 0.01471112547442317, + "grad_norm": 0.15200670063495636, + "learning_rate": 2.5352871267540906e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004921047110110521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020973030128516256, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01193939708173275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8007375419139862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17680, + "weighted_flow_matching_loss": 0.01471112547442317, + "weighted_rkd_loss": 0.009560342039912939 + }, + { + "epoch": 5.146930462612744, + "flow_matching_loss": 0.013468233775347472, + "grad_norm": 0.15532918274402618, + "learning_rate": 2.534688721966516e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004807946691289544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002125690574757755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01174158388748765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8005283534526825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17690, + "weighted_flow_matching_loss": 0.013468233775347472, + "weighted_rkd_loss": 0.00939944814890623 + }, + { + "epoch": 5.149839976723887, + "flow_matching_loss": 0.012908658571541309, + "grad_norm": 0.17043980956077576, + "learning_rate": 2.5340900028678276e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005662165256217122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002340274630114436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013664605095982552, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8003190755844116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17700, + "weighted_flow_matching_loss": 0.012908658571541309, + "weighted_rkd_loss": 0.01093600452877581 + }, + { + "epoch": 5.15274949083503, + "flow_matching_loss": 0.012674551270902158, + "grad_norm": 0.1588570773601532, + "learning_rate": 2.5334909696398998e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005534003954380751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024209878989495336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013488995749503374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8001097083091736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17710, + "weighted_flow_matching_loss": 0.012674551270902158, + "weighted_rkd_loss": 0.010792660247534513 + }, + { + "epoch": 5.155659004946174, + "flow_matching_loss": 0.01277663828805089, + "grad_norm": 0.16073717176914215, + "learning_rate": 2.5328916224647027e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005117162689566612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001997244090307504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012231569457799196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7999002695083618, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17720, + "weighted_flow_matching_loss": 0.01277663828805089, + "weighted_rkd_loss": 0.00978409736417234 + }, + { + "epoch": 5.158568519057318, + "flow_matching_loss": 0.01459071384742856, + "grad_norm": 0.16430360078811646, + "learning_rate": 2.532291961524303e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005789685621857643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002647255815099925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014226627070456743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7996907413005829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17730, + "weighted_flow_matching_loss": 0.01459071384742856, + "weighted_rkd_loss": 0.01137687275186181 + }, + { + "epoch": 5.161478033168461, + "flow_matching_loss": 0.012850914569571614, + "grad_norm": 0.12666183710098267, + "learning_rate": 2.531691987000861e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055351056391373275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00261001200415194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013680223375558853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7994811356067657, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17740, + "weighted_flow_matching_loss": 0.012850914569571614, + "weighted_rkd_loss": 0.010937143675982952 + }, + { + "epoch": 5.164387547279604, + "flow_matching_loss": 0.014312360156327485, + "grad_norm": 0.21371877193450928, + "learning_rate": 2.5310916990766343e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005188194639049471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024365652934648097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01281295446678996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7992714524269104, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17750, + "weighted_flow_matching_loss": 0.014312360156327485, + "weighted_rkd_loss": 0.010240935301408172 + }, + { + "epoch": 5.167297061390748, + "flow_matching_loss": 0.01383068049326539, + "grad_norm": 0.182195782661438, + "learning_rate": 2.5304910979339734e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005683549027889967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025549915852025153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013922089617699385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7990616858005524, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17760, + "weighted_flow_matching_loss": 0.01383068049326539, + "weighted_rkd_loss": 0.01112462729215622 + }, + { + "epoch": 5.170206575501891, + "flow_matching_loss": 0.013389503769576549, + "grad_norm": 0.1677216738462448, + "learning_rate": 2.5298901837553255e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005648717563599348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002341566176619381, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01363900126889348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7988518297672271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17770, + "weighted_flow_matching_loss": 0.013389503769576549, + "weighted_rkd_loss": 0.010895539913326502 + }, + { + "epoch": 5.173116089613035, + "flow_matching_loss": 0.013751755189150571, + "grad_norm": 0.17602944374084473, + "learning_rate": 2.5292889567232326e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005704364483244717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00253187125781551, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013940600026398898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7986419022083282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17780, + "weighted_flow_matching_loss": 0.013751755189150571, + "weighted_rkd_loss": 0.011133548943325878 + }, + { + "epoch": 5.176025603724178, + "flow_matching_loss": 0.013324832823127509, + "grad_norm": 0.17882892489433289, + "learning_rate": 2.528687417020331e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005274221254512668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023916824138723314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012940125074237585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7984318852424621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17790, + "weighted_flow_matching_loss": 0.013324832823127509, + "weighted_rkd_loss": 0.010331743909046055 + }, + { + "epoch": 5.178935117835321, + "flow_matching_loss": 0.014692158903926611, + "grad_norm": 0.18745233118534088, + "learning_rate": 2.5280855648293536e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005136910802684724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022270227200351655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012500844430178404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7982217907905579, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17800, + "weighted_flow_matching_loss": 0.014692158903926611, + "weighted_rkd_loss": 0.00997841339558363 + }, + { + "epoch": 5.181844631946465, + "flow_matching_loss": 0.01405738415196538, + "grad_norm": 0.17263032495975494, + "learning_rate": 2.5274834003331266e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005025715217925608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002391153061762452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01244258340448141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7980116128921508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17810, + "weighted_flow_matching_loss": 0.01405738415196538, + "weighted_rkd_loss": 0.009929376933723689 + }, + { + "epoch": 5.1847541460576085, + "flow_matching_loss": 0.013336289487779141, + "grad_norm": 0.22802212834358215, + "learning_rate": 2.5268809237145717e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005774030904285609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002591756114270538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014139818027615547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7978013455867767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17820, + "weighted_flow_matching_loss": 0.013336289487779141, + "weighted_rkd_loss": 0.01128058391623199 + }, + { + "epoch": 5.187663660168752, + "flow_matching_loss": 0.011874900944530965, + "grad_norm": 0.1566820591688156, + "learning_rate": 2.5262781351567052e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005747857969254255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002561092330142856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014056808315217495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7975910007953644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17830, + "weighted_flow_matching_loss": 0.011874900944530965, + "weighted_rkd_loss": 0.011211591120809317 + }, + { + "epoch": 5.190573174279895, + "flow_matching_loss": 0.013198948185890912, + "grad_norm": 0.1914517730474472, + "learning_rate": 2.525675034842638e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0049287216970697045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021819087909534574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012039352301508188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7973805844783783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17840, + "weighted_flow_matching_loss": 0.013198948185890912, + "weighted_rkd_loss": 0.009599902154877781 + }, + { + "epoch": 5.193482688391039, + "flow_matching_loss": 0.012480553332716226, + "grad_norm": 0.20034678280353546, + "learning_rate": 2.5250716229555774e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005485746869817376, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002421487821266055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013392981328070163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7971700727939606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17850, + "weighted_flow_matching_loss": 0.012480553332716226, + "weighted_rkd_loss": 0.01067642024718225 + }, + { + "epoch": 5.1963922025021825, + "flow_matching_loss": 0.012455856706947089, + "grad_norm": 0.14953935146331787, + "learning_rate": 2.524467899678823e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005107683013193309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002175140054896474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012390506081283092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7969594895839691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17860, + "weighted_flow_matching_loss": 0.012455856706947089, + "weighted_rkd_loss": 0.009874813817441463 + }, + { + "epoch": 5.199301716613325, + "flow_matching_loss": 0.013901007734239102, + "grad_norm": 0.1756390780210495, + "learning_rate": 2.52386386519577e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005277509335428476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002445467549841851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01300048604607582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7967488169670105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17870, + "weighted_flow_matching_loss": 0.013901007734239102, + "weighted_rkd_loss": 0.010358099127188325 + }, + { + "epoch": 5.202211230724469, + "flow_matching_loss": 0.013047449849545955, + "grad_norm": 0.1486428827047348, + "learning_rate": 2.5232595196899086e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005542610911652446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025521078961901365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013637329824268818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7965380668640136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17880, + "weighted_flow_matching_loss": 0.013047449849545955, + "weighted_rkd_loss": 0.010862633120268583 + }, + { + "epoch": 5.205120744835613, + "flow_matching_loss": 0.012858987972140312, + "grad_norm": 0.16442087292671204, + "learning_rate": 2.5226548633448224e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00583838983438909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002670006942935288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014346786867827177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7963272392749786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17890, + "weighted_flow_matching_loss": 0.012858987972140312, + "weighted_rkd_loss": 0.01142477043904364 + }, + { + "epoch": 5.2080302589467555, + "flow_matching_loss": 0.013005403243005276, + "grad_norm": 0.1376558244228363, + "learning_rate": 2.5220498963441906e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005102040921337902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002171542448922992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01237562419846654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7961163341999054, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17900, + "weighted_flow_matching_loss": 0.013005403243005276, + "weighted_rkd_loss": 0.009852430503815412 + }, + { + "epoch": 5.210939773057899, + "flow_matching_loss": 0.012854571361094714, + "grad_norm": 0.15978427231311798, + "learning_rate": 2.521444618871786e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005235111620277166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021379630779847503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012608186341822147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.795905339717865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17910, + "weighted_flow_matching_loss": 0.012854571361094714, + "weighted_rkd_loss": 0.010034970007836819 + }, + { + "epoch": 5.213849287169043, + "flow_matching_loss": 0.013605640735477209, + "grad_norm": 0.1794656664133072, + "learning_rate": 2.5208390311114758e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005573278199881316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025069542694836854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013653510715812445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7956942617893219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17920, + "weighted_flow_matching_loss": 0.013605640735477209, + "weighted_rkd_loss": 0.01086404682137072 + }, + { + "epoch": 5.216758801280186, + "flow_matching_loss": 0.013927933014929294, + "grad_norm": 0.15354007482528687, + "learning_rate": 2.520233133247221e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005761733883991838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002440048928838223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013963516615331174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7954831004142762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17930, + "weighted_flow_matching_loss": 0.013927933014929294, + "weighted_rkd_loss": 0.01110762800090015 + }, + { + "epoch": 5.2196683153913295, + "flow_matching_loss": 0.013538658805191516, + "grad_norm": 0.18280571699142456, + "learning_rate": 2.519626925463079e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006041886983439326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002613587712403387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0146973617374897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7952718675136566, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17940, + "weighted_flow_matching_loss": 0.013538658805191516, + "weighted_rkd_loss": 0.011688347859308123 + }, + { + "epoch": 5.222577829502473, + "flow_matching_loss": 0.012970003020018338, + "grad_norm": 0.1849897801876068, + "learning_rate": 2.519020407943198e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062755719292908905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002801273576915264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01535241724923253, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7950605511665344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17950, + "weighted_flow_matching_loss": 0.012970003020018338, + "weighted_rkd_loss": 0.012206141464412212 + }, + { + "epoch": 5.225487343613617, + "flow_matching_loss": 0.014681264292448759, + "grad_norm": 0.18094290792942047, + "learning_rate": 2.518413580871823e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057324391789734365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023863006732426583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01385117918252945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7948491513729096, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17960, + "weighted_flow_matching_loss": 0.014681264292448759, + "weighted_rkd_loss": 0.01100951712578535 + }, + { + "epoch": 5.22839685772476, + "flow_matching_loss": 0.013655302673578262, + "grad_norm": 0.17198191583156586, + "learning_rate": 2.5178064444332922e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005732604162767529, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025574193568900228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014022627845406532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7946376740932465, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17970, + "weighted_flow_matching_loss": 0.013655302673578262, + "weighted_rkd_loss": 0.011142834089696407 + }, + { + "epoch": 5.231306371835903, + "flow_matching_loss": 0.012714559771120548, + "grad_norm": 0.17030632495880127, + "learning_rate": 2.5171989988120368e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004366101231426001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002020108257420361, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010752310696989297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7944261133670807, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17980, + "weighted_flow_matching_loss": 0.012714559771120548, + "weighted_rkd_loss": 0.008541891491040588 + }, + { + "epoch": 5.234215885947047, + "flow_matching_loss": 0.014386261347681284, + "grad_norm": 0.18586400151252747, + "learning_rate": 2.5165912441925832e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005807502567768097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025156747316941617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01413067989051342, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7942144751548768, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17990, + "weighted_flow_matching_loss": 0.014386261347681284, + "weighted_rkd_loss": 0.011222808808088302 + }, + { + "epoch": 5.23712540005819, + "flow_matching_loss": 0.01331392815336585, + "grad_norm": 0.14956042170524597, + "learning_rate": 2.515983180759551e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004950365191325546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020331259234808385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01193385636433959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7940027475357055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18000, + "weighted_flow_matching_loss": 0.01331392815336585, + "weighted_rkd_loss": 0.009475614968687296 + }, + { + "epoch": 5.240034914169334, + "flow_matching_loss": 0.014330426324158907, + "grad_norm": 0.1809799075126648, + "learning_rate": 2.515374808697654e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00702778990380466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003231390821747482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017286970745772123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7937909483909606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18010, + "weighted_flow_matching_loss": 0.014330426324158907, + "weighted_rkd_loss": 0.013722119480371475 + }, + { + "epoch": 5.242944428280477, + "flow_matching_loss": 0.013384791929274798, + "grad_norm": 0.12386291474103928, + "learning_rate": 2.5147661281916996e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005042320839129389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024030060390941798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012487647868692876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7935790657997132, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18020, + "weighted_flow_matching_loss": 0.013384791929274798, + "weighted_rkd_loss": 0.009910002583637834 + }, + { + "epoch": 5.24585394239162, + "flow_matching_loss": 0.013839082699269057, + "grad_norm": 0.1436949521303177, + "learning_rate": 2.5141571394265892e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046868178527802225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020296604139730333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011403296142816544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7933670997619628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18030, + "weighted_flow_matching_loss": 0.013839082699269057, + "weighted_rkd_loss": 0.009046949539333582 + }, + { + "epoch": 5.248763456502764, + "flow_matching_loss": 0.012739633955061436, + "grad_norm": 0.15595102310180664, + "learning_rate": 2.513547842587317e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005533199710771442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002340785868000239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01340718511492014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7931550562381744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18040, + "weighted_flow_matching_loss": 0.012739633955061436, + "weighted_rkd_loss": 0.010634021833539009 + }, + { + "epoch": 5.251672970613908, + "flow_matching_loss": 0.012320217676460743, + "grad_norm": 0.14775897562503815, + "learning_rate": 2.5129382378589708e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005302399606443941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024733685539104046, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013078167662024498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7929429411888123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18050, + "weighted_flow_matching_loss": 0.012320217676460743, + "weighted_rkd_loss": 0.010370333306491375 + }, + { + "epoch": 5.254582484725051, + "flow_matching_loss": 0.014322473946958781, + "grad_norm": 0.1489073932170868, + "learning_rate": 2.512328325426733e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053618585923686625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023272670223377645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013050984311848878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7927307307720184, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18060, + "weighted_flow_matching_loss": 0.014322473946958781, + "weighted_rkd_loss": 0.010345969023182989 + }, + { + "epoch": 5.257491998836194, + "flow_matching_loss": 0.013651485182344913, + "grad_norm": 0.15807387232780457, + "learning_rate": 2.5117181054758798e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005095528112724424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002346280193887651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012537336442619562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7925184488296508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18070, + "weighted_flow_matching_loss": 0.013651485182344913, + "weighted_rkd_loss": 0.009936073562130332 + }, + { + "epoch": 5.260401512947338, + "flow_matching_loss": 0.013722768612205982, + "grad_norm": 0.1259385645389557, + "learning_rate": 2.5111075781917783e-05, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004511065967381001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019673849921673535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010989516973495483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7923060774803161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18080, + "weighted_flow_matching_loss": 0.013722768612205982, + "weighted_rkd_loss": 0.008707167254760862 + }, + { + "epoch": 5.2633110270584815, + "flow_matching_loss": 0.013565519545227289, + "grad_norm": 0.13981835544109344, + "learning_rate": 2.510496743759892e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004975151154212654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002117667905986309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012067970260977745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7920936346054077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18090, + "weighted_flow_matching_loss": 0.013565519545227289, + "weighted_rkd_loss": 0.00955907697789371 + }, + { + "epoch": 5.266220541169624, + "flow_matching_loss": 0.011932837404310703, + "grad_norm": 0.21601398289203644, + "learning_rate": 2.509885602365775e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005300882435403764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022432623780332507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012845027260482311, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7918811142444611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18100, + "weighted_flow_matching_loss": 0.011932837404310703, + "weighted_rkd_loss": 0.010171633213758469 + }, + { + "epoch": 5.269130055280768, + "flow_matching_loss": 0.015284962393343449, + "grad_norm": 0.15468034148216248, + "learning_rate": 2.509274154195076e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005779501190409065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026409060228616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014199908450245857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7916685044765472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18110, + "weighted_flow_matching_loss": 0.015284962393343449, + "weighted_rkd_loss": 0.011241643410176038 + }, + { + "epoch": 5.272039569391912, + "flow_matching_loss": 0.014088049624115228, + "grad_norm": 0.18496935069561005, + "learning_rate": 2.5086623994335383e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005390126374550164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002175112848635763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012955365609377623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7914558231830597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18120, + "weighted_flow_matching_loss": 0.014088049624115228, + "weighted_rkd_loss": 0.010253447806462646 + }, + { + "epoch": 5.274949083503055, + "flow_matching_loss": 0.012854194454848766, + "grad_norm": 0.16980662941932678, + "learning_rate": 2.5080503382669953e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056640340480953455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023744836333207788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01370255183428526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7912430465221405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18130, + "weighted_flow_matching_loss": 0.012854194454848766, + "weighted_rkd_loss": 0.010842036735266448 + }, + { + "epoch": 5.277858597614198, + "flow_matching_loss": 0.014505577832460403, + "grad_norm": 0.147157222032547, + "learning_rate": 2.507437970881376e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005247434205375612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00247245286591351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012967321090400219, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7910302162170411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18140, + "weighted_flow_matching_loss": 0.014505577832460403, + "weighted_rkd_loss": 0.01025754832662642 + }, + { + "epoch": 5.280768111725342, + "flow_matching_loss": 0.012021555844694377, + "grad_norm": 0.20086011290550232, + "learning_rate": 2.5068252974627003e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005180849926546216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023007332463748755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012662433087825775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7908172905445099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18150, + "weighted_flow_matching_loss": 0.012021555844694377, + "weighted_rkd_loss": 0.010013553546741605 + }, + { + "epoch": 5.283677625836486, + "flow_matching_loss": 0.015520573128014803, + "grad_norm": 0.20501059293746948, + "learning_rate": 2.5062123181970834e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005215476639568806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002499907254241407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012930860742926598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7906042814254761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18160, + "weighted_flow_matching_loss": 0.015520573128014803, + "weighted_rkd_loss": 0.010223142802715302 + }, + { + "epoch": 5.2865871399476285, + "flow_matching_loss": 0.0171262351796031, + "grad_norm": 0.1430453062057495, + "learning_rate": 2.5055990332707316e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004972222936339676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021217759582214057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012066221982240676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7903912007808686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18170, + "weighted_flow_matching_loss": 0.0171262351796031, + "weighted_rkd_loss": 0.00953709646128118 + }, + { + "epoch": 5.289496654058772, + "flow_matching_loss": 0.012883063685148955, + "grad_norm": 0.162403866648674, + "learning_rate": 2.5049854428699444e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005333793722093106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002399482298642397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01306706964969635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7901780366897583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18180, + "weighted_flow_matching_loss": 0.012883063685148955, + "weighted_rkd_loss": 0.010325241880491377 + }, + { + "epoch": 5.292406168169916, + "flow_matching_loss": 0.013521150778979063, + "grad_norm": 0.1267024427652359, + "learning_rate": 2.5043715471811158e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057234561536461115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025609220378100873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014007834438234567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7899647951126099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18190, + "weighted_flow_matching_loss": 0.013521150778979063, + "weighted_rkd_loss": 0.011065821908414365 + }, + { + "epoch": 5.295315682281059, + "flow_matching_loss": 0.014335987158119678, + "grad_norm": 0.15612857043743134, + "learning_rate": 2.5037573463907296e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00562675108667463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026634983834810553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013917000219225883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7897514700889587, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18200, + "weighted_flow_matching_loss": 0.014335987158119678, + "weighted_rkd_loss": 0.010991070372983813 + }, + { + "epoch": 5.2982251963922025, + "flow_matching_loss": 0.015222297701984643, + "grad_norm": 0.15315940976142883, + "learning_rate": 2.5031428406853637e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006062335334718227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026995170977897943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014824187848716974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7895380675792694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18210, + "weighted_flow_matching_loss": 0.015222297701984643, + "weighted_rkd_loss": 0.011704383324831725 + }, + { + "epoch": 5.301134710503346, + "flow_matching_loss": 0.014393100887537003, + "grad_norm": 0.16763345897197723, + "learning_rate": 2.5025280302516897e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005278801079839468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024544887244701387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013012090977281333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7893245875835418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18220, + "weighted_flow_matching_loss": 0.014393100887537003, + "weighted_rkd_loss": 0.010270806308835746 + }, + { + "epoch": 5.304044224614489, + "flow_matching_loss": 0.01344819599762559, + "grad_norm": 0.16539599001407623, + "learning_rate": 2.5019129152764698e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005812934553250671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00241933292709291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014045201987028123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7891110241413116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18230, + "weighted_flow_matching_loss": 0.01344819599762559, + "weighted_rkd_loss": 0.011083326302468777 + }, + { + "epoch": 5.306953738725633, + "flow_matching_loss": 0.014012191165238618, + "grad_norm": 0.13621431589126587, + "learning_rate": 2.5012974959465612e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005499115493148565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00243479359196499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013433024566620589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7888973951339722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18240, + "weighted_flow_matching_loss": 0.014012191165238618, + "weighted_rkd_loss": 0.010597359342500567 + }, + { + "epoch": 5.309863252836776, + "flow_matching_loss": 0.013408224377781153, + "grad_norm": 0.18022669851779938, + "learning_rate": 2.5006817724489105e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004955320339649916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023186796461232007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012229320220649242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7886836647987365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18250, + "weighted_flow_matching_loss": 0.013408224377781153, + "weighted_rkd_loss": 0.009644984919577836 + }, + { + "epoch": 5.31277276694792, + "flow_matching_loss": 0.014251247141510248, + "grad_norm": 0.1440785974264145, + "learning_rate": 2.500065744970559e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057566300965845585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025289997109211983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014042259939014912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7884698688983918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18260, + "weighted_flow_matching_loss": 0.014251247141510248, + "weighted_rkd_loss": 0.011071895249187946 + }, + { + "epoch": 5.315682281059063, + "flow_matching_loss": 0.012837554421275854, + "grad_norm": 0.18269291520118713, + "learning_rate": 2.499449413698639e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005765352863818407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002617969491984695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014148674998432397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7882559955120086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18270, + "weighted_flow_matching_loss": 0.012837554421275854, + "weighted_rkd_loss": 0.011152757424861192 + }, + { + "epoch": 5.318591795170207, + "flow_matching_loss": 0.01403694348409772, + "grad_norm": 0.1872728317975998, + "learning_rate": 2.4988327788203764e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0050162946106866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002074550616089255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012107139732688666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7880420386791229, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18280, + "weighted_flow_matching_loss": 0.01403694348409772, + "weighted_rkd_loss": 0.009540935279801488 + }, + { + "epoch": 5.32150130928135, + "flow_matching_loss": 0.014795887656509877, + "grad_norm": 0.1724988967180252, + "learning_rate": 2.498215840523088e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00534520922228694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002568730677012354, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013259149063378573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7878279983997345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18290, + "weighted_flow_matching_loss": 0.014795887656509877, + "weighted_rkd_loss": 0.01044579865410924 + }, + { + "epoch": 5.324410823392493, + "flow_matching_loss": 0.013322021625936032, + "grad_norm": 0.1451381891965866, + "learning_rate": 2.4975985989941837e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005947565520182252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002689164981711656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014584296010434627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7876138865947724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18300, + "weighted_flow_matching_loss": 0.013322021625936032, + "weighted_rkd_loss": 0.011486800480633974 + }, + { + "epoch": 5.327320337503637, + "flow_matching_loss": 0.012662121374160052, + "grad_norm": 0.1787356436252594, + "learning_rate": 2.4969810544211652e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005454369122162461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024012661189772187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013310004491358995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7873996913433075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18310, + "weighted_flow_matching_loss": 0.012662121374160052, + "weighted_rkd_loss": 0.010480267461389303 + }, + { + "epoch": 5.330229851614781, + "flow_matching_loss": 0.014418733213096857, + "grad_norm": 0.16260701417922974, + "learning_rate": 2.4963632069916258e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005105508305132389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002448723034467548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012659739516675473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7871854245662689, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18320, + "weighted_flow_matching_loss": 0.014418733213096857, + "weighted_rkd_loss": 0.009965619072318077 + }, + { + "epoch": 5.333139365725923, + "flow_matching_loss": 0.013167665805667639, + "grad_norm": 0.16249269247055054, + "learning_rate": 2.495745056893252e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005700283823534847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024451847188174725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013845752459019423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7869710743427276, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18330, + "weighted_flow_matching_loss": 0.013167665805667639, + "weighted_rkd_loss": 0.010896231373772025 + }, + { + "epoch": 5.336048879837067, + "flow_matching_loss": 0.013843031786382198, + "grad_norm": 0.22530296444892883, + "learning_rate": 2.495126604313821e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005254888557828963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022671217331662772, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01277689877897501, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7867566406726837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18340, + "weighted_flow_matching_loss": 0.013843031786382198, + "weighted_rkd_loss": 0.010052282642573119 + }, + { + "epoch": 5.338958393948211, + "flow_matching_loss": 0.012435850407928228, + "grad_norm": 0.19639356434345245, + "learning_rate": 2.4945078494412023e-05, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005092221871018409, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002292003924958408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012476447504013777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7865421175956726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18350, + "weighted_flow_matching_loss": 0.012435850407928228, + "weighted_rkd_loss": 0.00981321237049997 + }, + { + "epoch": 5.3418679080593545, + "flow_matching_loss": 0.013865058217197657, + "grad_norm": 0.16460339725017548, + "learning_rate": 2.493888792463357e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005513044586405158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002327288768719882, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013353377766907214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7863275408744812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18360, + "weighted_flow_matching_loss": 0.013865058217197657, + "weighted_rkd_loss": 0.010500069987028838 + }, + { + "epoch": 5.344777422170497, + "flow_matching_loss": 0.011539196968078614, + "grad_norm": 0.1821148842573166, + "learning_rate": 2.4932694335683395e-05, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046374562429264184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019948941306211056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01126980660483241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7861128866672515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18370, + "weighted_flow_matching_loss": 0.011539196968078614, + "weighted_rkd_loss": 0.008859350671991706 + }, + { + "epoch": 5.347686936281641, + "flow_matching_loss": 0.013982224185019732, + "grad_norm": 0.19098913669586182, + "learning_rate": 2.4926497729442932e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051718601956963536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002368639106862247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012712359614670276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7858981370925904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18380, + "weighted_flow_matching_loss": 0.013982224185019732, + "weighted_rkd_loss": 0.0099905866663903 + }, + { + "epoch": 5.350596450392785, + "flow_matching_loss": 0.013056739326566457, + "grad_norm": 0.178262859582901, + "learning_rate": 2.4920298107794555e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005991244828328491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028049296233803034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014787419326603412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7856833159923553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18390, + "weighted_flow_matching_loss": 0.013056739326566457, + "weighted_rkd_loss": 0.011618105322122574 + }, + { + "epoch": 5.353505964503928, + "flow_matching_loss": 0.014394402131438255, + "grad_norm": 0.18875738978385925, + "learning_rate": 2.4914095472621544e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005378671921789646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022529866895638407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01301033077761531, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7854684114456176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18400, + "weighted_flow_matching_loss": 0.014394402131438255, + "weighted_rkd_loss": 0.010219220165163278 + }, + { + "epoch": 5.356415478615071, + "flow_matching_loss": 0.014151305239647628, + "grad_norm": 0.1553405225276947, + "learning_rate": 2.4907889825808093e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006252031633630395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027366479276679455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015240711253136396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7852534353733063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18410, + "weighted_flow_matching_loss": 0.014151305239647628, + "weighted_rkd_loss": 0.01196784721687436 + }, + { + "epoch": 5.359324992726215, + "flow_matching_loss": 0.014367901254445314, + "grad_norm": 0.18662233650684357, + "learning_rate": 2.4901681169239314e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005461022327654064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002448550844565034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013370595406740904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7850383758544922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18420, + "weighted_flow_matching_loss": 0.014367901254445314, + "weighted_rkd_loss": 0.010496393032371997 + }, + { + "epoch": 5.362234506837358, + "flow_matching_loss": 0.013021718803793192, + "grad_norm": 0.17281360924243927, + "learning_rate": 2.4895469504801236e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053616099059581755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002491205115802586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013214424997568131, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7848232686519623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18430, + "weighted_flow_matching_loss": 0.013021718803793192, + "weighted_rkd_loss": 0.010371009586378932 + }, + { + "epoch": 5.3651440209485015, + "flow_matching_loss": 0.012966876197606326, + "grad_norm": 0.1594315469264984, + "learning_rate": 2.4889254834380802e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005080067529343068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021405268576927482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012300661858171224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7846080243587494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18440, + "weighted_flow_matching_loss": 0.012966876197606326, + "weighted_rkd_loss": 0.009651120286434889 + }, + { + "epoch": 5.368053535059645, + "flow_matching_loss": 0.013147742860019207, + "grad_norm": 0.15646058320999146, + "learning_rate": 2.488303715986585e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005305970320478082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002184763131663203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012796703726053238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7843927443027496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18450, + "weighted_flow_matching_loss": 0.013147742860019207, + "weighted_rkd_loss": 0.010037656500935554 + }, + { + "epoch": 5.370963049170788, + "flow_matching_loss": 0.013741484750062228, + "grad_norm": 0.17914316058158875, + "learning_rate": 2.4876816483145166e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005824220599606633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024691595695912837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01411760076880455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7841773748397827, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18460, + "weighted_flow_matching_loss": 0.013741484750062228, + "weighted_rkd_loss": 0.011070618592202663 + }, + { + "epoch": 5.373872563281932, + "flow_matching_loss": 0.012088405713438988, + "grad_norm": 0.1575964093208313, + "learning_rate": 2.487059280610841e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005419267248362303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002558855037204921, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013397389370948077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7839619219303131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18470, + "weighted_flow_matching_loss": 0.012088405713438988, + "weighted_rkd_loss": 0.010503138974308967 + }, + { + "epoch": 5.3767820773930755, + "flow_matching_loss": 0.013183113280683756, + "grad_norm": 0.21278467774391174, + "learning_rate": 2.4864366130646178e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052684277528896924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023816479719243945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01291850358247757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7837464034557342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18480, + "weighted_flow_matching_loss": 0.013183113280683756, + "weighted_rkd_loss": 0.010124743077903986 + }, + { + "epoch": 5.379691591504219, + "flow_matching_loss": 0.01555206747725606, + "grad_norm": 0.18399664759635925, + "learning_rate": 2.485813645864997e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004940612660720945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002192650316283107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012073875591158866, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7835307955741883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18490, + "weighted_flow_matching_loss": 0.01555206747725606, + "weighted_rkd_loss": 0.009460249077528715 + }, + { + "epoch": 5.382601105615362, + "flow_matching_loss": 0.01565149538218975, + "grad_norm": 0.18347015976905823, + "learning_rate": 2.4851903792012198e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005184060521423816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002330328105017543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01269844900816679, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.783315122127533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18500, + "weighted_flow_matching_loss": 0.01565149538218975, + "weighted_rkd_loss": 0.009946885891258717 + }, + { + "epoch": 5.385510619726506, + "flow_matching_loss": 0.012770815473049878, + "grad_norm": 0.16084225475788116, + "learning_rate": 2.4845668132626173e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051231029443442825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023470959858968854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012593301851302386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.783099353313446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18510, + "weighted_flow_matching_loss": 0.012770815473049878, + "weighted_rkd_loss": 0.00986188929527998 + }, + { + "epoch": 5.388420133837649, + "flow_matching_loss": 0.012762334384024143, + "grad_norm": 0.12914451956748962, + "learning_rate": 2.483942948238613e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004708781861700118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018988160882145166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011316379997879267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7828835248947144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18520, + "weighted_flow_matching_loss": 0.012762334384024143, + "weighted_rkd_loss": 0.008859359752386808 + }, + { + "epoch": 5.391329647948792, + "flow_matching_loss": 0.014616053458303213, + "grad_norm": 0.1621561497449875, + "learning_rate": 2.4833187843187207e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00563686485402286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002682437573093921, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01395616726949811, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7826676189899444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18530, + "weighted_flow_matching_loss": 0.014616053458303213, + "weighted_rkd_loss": 0.01092316946014762 + }, + { + "epoch": 5.394239162059936, + "flow_matching_loss": 0.013246619421988726, + "grad_norm": 0.1802680790424347, + "learning_rate": 2.4826943216925448e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004985922737978399, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021585542941465973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012130399700254202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7824516236782074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18540, + "weighted_flow_matching_loss": 0.013246619421988726, + "weighted_rkd_loss": 0.009491439629346132 + }, + { + "epoch": 5.39714867617108, + "flow_matching_loss": 0.012937465030699969, + "grad_norm": 0.15631453692913055, + "learning_rate": 2.4820695605497807e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005641977651976049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00238848973531276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013672445062547922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7822355508804322, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18550, + "weighted_flow_matching_loss": 0.012937465030699969, + "weighted_rkd_loss": 0.010695086512714624 + }, + { + "epoch": 5.400058190282223, + "flow_matching_loss": 0.013960608839988708, + "grad_norm": 0.17523179948329926, + "learning_rate": 2.481444501080214e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005562798120081425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024276574258692563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013553253654390574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7820194125175476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18560, + "weighted_flow_matching_loss": 0.013960608839988708, + "weighted_rkd_loss": 0.010598861798644065 + }, + { + "epoch": 5.402967704393366, + "flow_matching_loss": 0.01554051199927926, + "grad_norm": 0.1731898933649063, + "learning_rate": 2.4808191434737217e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005380712752230466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024422852555289866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013203710783272982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7818031907081604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18570, + "weighted_flow_matching_loss": 0.01554051199927926, + "weighted_rkd_loss": 0.010322725865989923 + }, + { + "epoch": 5.40587721850451, + "flow_matching_loss": 0.011951501667499542, + "grad_norm": 0.14728544652462006, + "learning_rate": 2.4801934879202696e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005248880106955766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021856520092114807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012683412339538336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7815868973731994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18580, + "weighted_flow_matching_loss": 0.011951501667499542, + "weighted_rkd_loss": 0.00991311757825315 + }, + { + "epoch": 5.408786732615654, + "flow_matching_loss": 0.013323656097054482, + "grad_norm": 0.1571427583694458, + "learning_rate": 2.4795675346099172e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004900332703255117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022027099039405584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012003375310450792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7813705086708069, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18590, + "weighted_flow_matching_loss": 0.013323656097054482, + "weighted_rkd_loss": 0.009379083337262274 + }, + { + "epoch": 5.411696246726796, + "flow_matching_loss": 0.014386238437145949, + "grad_norm": 0.1715613603591919, + "learning_rate": 2.4789412837328114e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005659433640539646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025051277480088174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013823994994163513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7811540603637696, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18600, + "weighted_flow_matching_loss": 0.014386238437145949, + "weighted_rkd_loss": 0.010798610839992761 + }, + { + "epoch": 5.41460576083794, + "flow_matching_loss": 0.014321433380246162, + "grad_norm": 0.15438656508922577, + "learning_rate": 2.4783147354791915e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004825639934279024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002153150155209005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011804430093616248, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7809375286102295, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18610, + "weighted_flow_matching_loss": 0.014321433380246162, + "weighted_rkd_loss": 0.009218536084517837 + }, + { + "epoch": 5.417515274949084, + "flow_matching_loss": 0.015459595806896686, + "grad_norm": 0.16368670761585236, + "learning_rate": 2.4776878900393858e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005469905585050583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022168768919073045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013156688120216132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7807209193706512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18620, + "weighted_flow_matching_loss": 0.015459595806896686, + "weighted_rkd_loss": 0.010271760448813439 + }, + { + "epoch": 5.420424789060227, + "flow_matching_loss": 0.012766253016889096, + "grad_norm": 0.18493355810642242, + "learning_rate": 2.477060747603813e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005522375414147973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023520545451901852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013396805338561534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7805042326450348, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18630, + "weighted_flow_matching_loss": 0.012766253016889096, + "weighted_rkd_loss": 0.010456230258569121 + }, + { + "epoch": 5.42333430317137, + "flow_matching_loss": 0.0143208434805274, + "grad_norm": 0.1782565414905548, + "learning_rate": 2.4764333083629833e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053702750708907844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025105128646828235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013251063227653504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7802874684333801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18640, + "weighted_flow_matching_loss": 0.0143208434805274, + "weighted_rkd_loss": 0.010339767765253782 + }, + { + "epoch": 5.426243817282514, + "flow_matching_loss": 0.01261799931526184, + "grad_norm": 0.16110004484653473, + "learning_rate": 2.4758055725074954e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005104309972375631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021498932503163813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012358513381332158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7800706326961517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18650, + "weighted_flow_matching_loss": 0.01261799931526184, + "weighted_rkd_loss": 0.009640527796000243 + }, + { + "epoch": 5.429153331393657, + "flow_matching_loss": 0.012213654443621635, + "grad_norm": 0.16658452153205872, + "learning_rate": 2.4751775402280396e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004665738111361862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019463021657429636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011277778446674347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7798537015914917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18660, + "weighted_flow_matching_loss": 0.012213654443621635, + "weighted_rkd_loss": 0.008795012347400188 + }, + { + "epoch": 5.432062845504801, + "flow_matching_loss": 0.0143201963044703, + "grad_norm": 0.19918793439865112, + "learning_rate": 2.4745492117153947e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005286582931876183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024524110602214934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013025576900690793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7796367228031158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18670, + "weighted_flow_matching_loss": 0.0143201963044703, + "weighted_rkd_loss": 0.010155200120061636 + }, + { + "epoch": 5.434972359615944, + "flow_matching_loss": 0.014020508155226707, + "grad_norm": 0.15912604331970215, + "learning_rate": 2.473920587160431e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057986432686448095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024431457160972057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014040432404726744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7794196546077728, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18680, + "weighted_flow_matching_loss": 0.014020508155226707, + "weighted_rkd_loss": 0.010943426564335824 + }, + { + "epoch": 5.437881873727088, + "flow_matching_loss": 0.013358015473932028, + "grad_norm": 0.14119289815425873, + "learning_rate": 2.473291666754108e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005610206187702716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002738908678293228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013959321100264788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7792025029659271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18690, + "weighted_flow_matching_loss": 0.013358015473932028, + "weighted_rkd_loss": 0.01087727751582861 + }, + { + "epoch": 5.440791387838231, + "flow_matching_loss": 0.014148211199790239, + "grad_norm": 0.17521478235721588, + "learning_rate": 2.4726624506874748e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005044225603342056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002225831849500537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012314283102750779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7789852917194366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18700, + "weighted_flow_matching_loss": 0.014148211199790239, + "weighted_rkd_loss": 0.009592592436820268 + }, + { + "epoch": 5.4437009019493745, + "flow_matching_loss": 0.013334641139954329, + "grad_norm": 0.16903157532215118, + "learning_rate": 2.4720329391516708e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005062768212519586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002083701197989285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012209237646311521, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.77876797914505, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18710, + "weighted_flow_matching_loss": 0.013334641139954329, + "weighted_rkd_loss": 0.009508104715496301 + }, + { + "epoch": 5.446610416060518, + "flow_matching_loss": 0.015135439112782478, + "grad_norm": 0.16383399069309235, + "learning_rate": 2.4714031323379248e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00503109497949481, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002188398223370314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012250588182359934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7785506069660186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18720, + "weighted_flow_matching_loss": 0.015135439112782478, + "weighted_rkd_loss": 0.009537753975018859 + }, + { + "epoch": 5.449519930171661, + "flow_matching_loss": 0.014366428460925817, + "grad_norm": 0.18768690526485443, + "learning_rate": 2.4707730304375556e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005327060935087502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022806557128205894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012934777606278657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7783331513404846, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18730, + "weighted_flow_matching_loss": 0.014366428460925817, + "weighted_rkd_loss": 0.010067701619118453 + }, + { + "epoch": 5.452429444282805, + "flow_matching_loss": 0.014975227136164903, + "grad_norm": 0.2052505612373352, + "learning_rate": 2.4701426336419713e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005938171362504363, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002615860663354397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014492203574627638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7781156361103058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18740, + "weighted_flow_matching_loss": 0.014975227136164903, + "weighted_rkd_loss": 0.011276596877723932 + }, + { + "epoch": 5.4553389583939484, + "flow_matching_loss": 0.01333257695659995, + "grad_norm": 0.1508951485157013, + "learning_rate": 2.4695119421426707e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005125114391557872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002357634494546801, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01260786335915327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7778980255126953, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18750, + "weighted_flow_matching_loss": 0.01333257695659995, + "weighted_rkd_loss": 0.009807672863826156 + }, + { + "epoch": 5.458248472505091, + "flow_matching_loss": 0.01450979821383953, + "grad_norm": 0.17831778526306152, + "learning_rate": 2.46888095613124e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006601964402943849, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028719033696688713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016075832303613424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7776803493499755, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18760, + "weighted_flow_matching_loss": 0.01450979821383953, + "weighted_rkd_loss": 0.012501869816333055 + }, + { + "epoch": 5.461157986616235, + "flow_matching_loss": 0.014869169145822526, + "grad_norm": 0.23394463956356049, + "learning_rate": 2.468249675799357e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005203821253962815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002283661405090243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012691303808242082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7774625897407532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18770, + "weighted_flow_matching_loss": 0.014869169145822526, + "weighted_rkd_loss": 0.0098669758066535 + }, + { + "epoch": 5.464067500727379, + "flow_matching_loss": 0.014048900827765465, + "grad_norm": 0.16991503536701202, + "learning_rate": 2.467618101338787e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005625739833340049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002468762034550309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013720241654664279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7772447466850281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18780, + "weighted_flow_matching_loss": 0.014048900827765465, + "weighted_rkd_loss": 0.010664054052904248 + }, + { + "epoch": 5.4669770148385215, + "flow_matching_loss": 0.013581638038158417, + "grad_norm": 0.13161084055900574, + "learning_rate": 2.466986232941387e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005553648667410016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024070136598311363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013514310959726572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7770268440246582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18790, + "weighted_flow_matching_loss": 0.013581638038158417, + "weighted_rkd_loss": 0.01050102305598557 + }, + { + "epoch": 5.469886528949665, + "flow_matching_loss": 0.013729181233793497, + "grad_norm": 0.15886665880680084, + "learning_rate": 2.466354070799101e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005349349370226264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002267863869201392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012966562528163195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7768088638782501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18800, + "weighted_flow_matching_loss": 0.013729181233793497, + "weighted_rkd_loss": 0.010072533879429102 + }, + { + "epoch": 5.472796043060809, + "flow_matching_loss": 0.01200804766267538, + "grad_norm": 0.16757145524024963, + "learning_rate": 2.4657216151039634e-05, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054758102633059025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023877658997662364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013339386414736509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7765908062458038, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18810, + "weighted_flow_matching_loss": 0.01200804766267538, + "weighted_rkd_loss": 0.010359326843172312 + }, + { + "epoch": 5.475705557171953, + "flow_matching_loss": 0.01431539049372077, + "grad_norm": 0.1569131463766098, + "learning_rate": 2.4650888660480976e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00582717233337462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026242970954626797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01427864171564579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7763726711273193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18820, + "weighted_flow_matching_loss": 0.01431539049372077, + "weighted_rkd_loss": 0.011085559893399478 + }, + { + "epoch": 5.478615071283095, + "flow_matching_loss": 0.013245524186640979, + "grad_norm": 0.16868534684181213, + "learning_rate": 2.464455823823716e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005495842173695564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025268687284551563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013518553040921689, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7761544525623322, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18830, + "weighted_flow_matching_loss": 0.013245524186640979, + "weighted_rkd_loss": 0.010492504807189107 + }, + { + "epoch": 5.481524585394239, + "flow_matching_loss": 0.016020228154957294, + "grad_norm": 0.16231060028076172, + "learning_rate": 2.4638224886231196e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005550069618038833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024881329387426375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013588272081688047, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7759361684322357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18840, + "weighted_flow_matching_loss": 0.016020228154957294, + "weighted_rkd_loss": 0.010543650574982166 + }, + { + "epoch": 5.484434099505383, + "flow_matching_loss": 0.013692237716168166, + "grad_norm": 0.1607344150543213, + "learning_rate": 2.4631888606387e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005341467447578907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002290416054893285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012973351031541824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.775717806816101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18850, + "weighted_flow_matching_loss": 0.013692237716168166, + "weighted_rkd_loss": 0.010063665686175227 + }, + { + "epoch": 5.487343613616526, + "flow_matching_loss": 0.014080309495329856, + "grad_norm": 0.15971609950065613, + "learning_rate": 2.4625549400629356e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005867682513780892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002648600854445249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01438396591693163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7754993677139282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18860, + "weighted_flow_matching_loss": 0.014080309495329856, + "weighted_rkd_loss": 0.011154850525781513 + }, + { + "epoch": 5.490253127727669, + "flow_matching_loss": 0.013883247599005699, + "grad_norm": 0.14529794454574585, + "learning_rate": 2.4619207270883958e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005651978077366948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002556647441815585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013860603701323271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7752808511257172, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18870, + "weighted_flow_matching_loss": 0.013883247599005699, + "weighted_rkd_loss": 0.010745870135724544 + }, + { + "epoch": 5.493162641838813, + "flow_matching_loss": 0.013607772625982761, + "grad_norm": 0.15366309881210327, + "learning_rate": 2.461286221907737e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005423654755577445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023676833719946445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013214993011206388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7750622570514679, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18880, + "weighted_flow_matching_loss": 0.013607772625982761, + "weighted_rkd_loss": 0.0102423963136971 + }, + { + "epoch": 5.496072155949957, + "flow_matching_loss": 0.014363639801740647, + "grad_norm": 0.14043140411376953, + "learning_rate": 2.460651424713705e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005371190095320344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028145578340627252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013556937873363494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7748435974121094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18890, + "weighted_flow_matching_loss": 0.014363639801740647, + "weighted_rkd_loss": 0.010504579450935126 + }, + { + "epoch": 5.4989816700611, + "flow_matching_loss": 0.01191713074222207, + "grad_norm": 0.153829425573349, + "learning_rate": 2.4600163356991347e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004833887796849012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020039768307469783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011671752389520406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7746248543262482, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18900, + "weighted_flow_matching_loss": 0.01191713074222207, + "weighted_rkd_loss": 0.009041182370856405 + }, + { + "epoch": 5.501891184172243, + "flow_matching_loss": 0.013805901166051626, + "grad_norm": 0.1715102642774582, + "learning_rate": 2.4593809550569498e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005335780326277018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002315400808583945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012986961659044028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7744060337543488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18910, + "weighted_flow_matching_loss": 0.013805901166051626, + "weighted_rkd_loss": 0.010057147778570651 + }, + { + "epoch": 5.504800698283387, + "flow_matching_loss": 0.013105537928640842, + "grad_norm": 0.18313172459602356, + "learning_rate": 2.4587452829801614e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005163258383981883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002304499049205333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01263101575896144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7741871416568756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18920, + "weighted_flow_matching_loss": 0.013105537928640842, + "weighted_rkd_loss": 0.009778833156451583 + }, + { + "epoch": 5.50771021239453, + "flow_matching_loss": 0.015842126961797477, + "grad_norm": 0.1258547306060791, + "learning_rate": 2.45810931966187e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005427895928733051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023519984330050646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013207790348678828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7739681780338288, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18930, + "weighted_flow_matching_loss": 0.015842126961797477, + "weighted_rkd_loss": 0.0102224197704345 + }, + { + "epoch": 5.5106197265056736, + "flow_matching_loss": 0.014073380827903747, + "grad_norm": 0.16267937421798706, + "learning_rate": 2.457473065295265e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00526641528122127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022999233799055217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012832754012197255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7737491369247437, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18940, + "weighted_flow_matching_loss": 0.014073380827903747, + "weighted_rkd_loss": 0.009929277095943689 + }, + { + "epoch": 5.513529240616817, + "flow_matching_loss": 0.013828863762319088, + "grad_norm": 0.16728895902633667, + "learning_rate": 2.4568365200736232e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005594928888604045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002461618452798575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013651476055383683, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7735300183296203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18950, + "weighted_flow_matching_loss": 0.013828863762319088, + "weighted_rkd_loss": 0.010559926694259047 + }, + { + "epoch": 5.51643875472796, + "flow_matching_loss": 0.013286172598600387, + "grad_norm": 0.13683341443538666, + "learning_rate": 2.4561996841903104e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005300390906631947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002516785601619631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013117567542940379, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7733108282089234, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18960, + "weighted_flow_matching_loss": 0.013286172598600387, + "weighted_rkd_loss": 0.010143967624753714 + }, + { + "epoch": 5.519348268839104, + "flow_matching_loss": 0.011709375865757465, + "grad_norm": 0.16258837282657623, + "learning_rate": 2.455562557838781e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005404610652476549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024509258335456254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013260147161781788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7730915546417236, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18970, + "weighted_flow_matching_loss": 0.011709375865757465, + "weighted_rkd_loss": 0.010251281689852476 + }, + { + "epoch": 5.5222577829502475, + "flow_matching_loss": 0.012814025115221738, + "grad_norm": 0.16433851420879364, + "learning_rate": 2.4549251412125762e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005145498621277511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022109546116553247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012501951865851879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7728722095489502, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18980, + "weighted_flow_matching_loss": 0.012814025115221738, + "weighted_rkd_loss": 0.009662394365295767 + }, + { + "epoch": 5.52516729706139, + "flow_matching_loss": 0.012991401506587863, + "grad_norm": 0.15272079408168793, + "learning_rate": 2.454287434505327e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005135067808441818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002359866921324283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012630002666264772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.772652804851532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18990, + "weighted_flow_matching_loss": 0.012991401506587863, + "weighted_rkd_loss": 0.009758575493469835 + }, + { + "epoch": 5.528076811172534, + "flow_matching_loss": 0.014172946009784937, + "grad_norm": 0.18472795188426971, + "learning_rate": 2.4536494379107514e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005080452072434127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002282948524225503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012443852610886097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7724333107471466, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19000, + "weighted_flow_matching_loss": 0.014172946009784937, + "weighted_rkd_loss": 0.009612116264179349 + }, + { + "epoch": 5.530986325283678, + "flow_matching_loss": 0.014220930263400077, + "grad_norm": 0.16647525131702423, + "learning_rate": 2.453011151622657e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052187366411089895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023459379328414796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012783411145210265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.772213751077652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19010, + "weighted_flow_matching_loss": 0.014220930263400077, + "weighted_rkd_loss": 0.009871467109769582 + }, + { + "epoch": 5.533895839394821, + "flow_matching_loss": 0.01204069945961237, + "grad_norm": 0.15786652266979218, + "learning_rate": 2.452372575834937e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005221085785888135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002410304709337652, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0128524761646986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7719941079616547, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19020, + "weighted_flow_matching_loss": 0.01204069945961237, + "weighted_rkd_loss": 0.009922051057219505 + }, + { + "epoch": 5.536805353505964, + "flow_matching_loss": 0.012842008285224438, + "grad_norm": 0.14660398662090302, + "learning_rate": 2.4517337107415742e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005442222533747554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025367754860781133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013421220518648625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7717743933200836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19030, + "weighted_flow_matching_loss": 0.012842008285224438, + "weighted_rkd_loss": 0.010358176613226532 + }, + { + "epoch": 5.539714867617108, + "flow_matching_loss": 0.012861193716526031, + "grad_norm": 0.15437385439872742, + "learning_rate": 2.4510945565366397e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005355396377854049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002406866196542978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013117658998817205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7715546011924743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19040, + "weighted_flow_matching_loss": 0.012861193716526031, + "weighted_rkd_loss": 0.010121061000972986 + }, + { + "epoch": 5.542624381728252, + "flow_matching_loss": 0.01350928759202361, + "grad_norm": 0.1510983109474182, + "learning_rate": 2.4504551134142905e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005008472176268697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002179274393711239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012196218781173229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7713347375392914, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19050, + "weighted_flow_matching_loss": 0.01350928759202361, + "weighted_rkd_loss": 0.009407383529469371 + }, + { + "epoch": 5.5455338958393945, + "flow_matching_loss": 0.013993003126233815, + "grad_norm": 0.15558260679244995, + "learning_rate": 2.4498153815687738e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004669456905685365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021021705353632568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011441084370017052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7711147964000702, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19060, + "weighted_flow_matching_loss": 0.013993003126233815, + "weighted_rkd_loss": 0.008822349645197392 + }, + { + "epoch": 5.548443409950538, + "flow_matching_loss": 0.013544460851699113, + "grad_norm": 0.17403335869312286, + "learning_rate": 2.4491753611944224e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005001689773052931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023027192219160497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012306098733097315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7708947837352753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19070, + "weighted_flow_matching_loss": 0.013544460851699113, + "weighted_rkd_loss": 0.00948668601922691 + }, + { + "epoch": 5.551352924061682, + "flow_matching_loss": 0.013182212319225072, + "grad_norm": 0.18242640793323517, + "learning_rate": 2.4485350524856577e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005321672884747386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002393609075807035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01303695486858487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7706746995449066, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19080, + "weighted_flow_matching_loss": 0.013182212319225072, + "weighted_rkd_loss": 0.0100472881924361 + }, + { + "epoch": 5.554262438172826, + "flow_matching_loss": 0.011304201651364565, + "grad_norm": 0.12165901809930801, + "learning_rate": 2.4478944556369886e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004961754335090518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00206736964173615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011990878265351057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7704545438289643, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19090, + "weighted_flow_matching_loss": 0.011304201651364565, + "weighted_rkd_loss": 0.009238470159471036 + }, + { + "epoch": 5.557171952283968, + "flow_matching_loss": 0.012489100452512503, + "grad_norm": 0.19810347259044647, + "learning_rate": 2.4472535708430116e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005350901046767831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022423040703870357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012944106105715036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7702343046665192, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19100, + "weighted_flow_matching_loss": 0.012489100452512503, + "weighted_rkd_loss": 0.009969961410388351 + }, + { + "epoch": 5.560081466395112, + "flow_matching_loss": 0.012454477325081825, + "grad_norm": 0.16181959211826324, + "learning_rate": 2.4466123982984103e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005363984708674252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023711370886303484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013099106680601835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7700139939785003, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19110, + "weighted_flow_matching_loss": 0.012454477325081825, + "weighted_rkd_loss": 0.010086479131132364 + }, + { + "epoch": 5.562990980506255, + "flow_matching_loss": 0.014894196018576621, + "grad_norm": 0.16817833483219147, + "learning_rate": 2.445970938197957e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005128718889318406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023423919337801635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012599829770624637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7697936117649078, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19120, + "weighted_flow_matching_loss": 0.014894196018576621, + "weighted_rkd_loss": 0.009699217090383173 + }, + { + "epoch": 5.565900494617399, + "flow_matching_loss": 0.01317557692527771, + "grad_norm": 0.167373389005661, + "learning_rate": 2.4453291907365092e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005737878661602735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002582772565074265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014058529958128929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7695731520652771, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19130, + "weighted_flow_matching_loss": 0.01317557692527771, + "weighted_rkd_loss": 0.010818931506946683 + }, + { + "epoch": 5.568810008728542, + "flow_matching_loss": 0.013601016998291016, + "grad_norm": 0.16595308482646942, + "learning_rate": 2.444687156109013e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005500163254328072, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025110924849286675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013511419016867875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7693526148796082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19140, + "weighted_flow_matching_loss": 0.013601016998291016, + "weighted_rkd_loss": 0.010395078035071493 + }, + { + "epoch": 5.571719522839686, + "flow_matching_loss": 0.012326624151319266, + "grad_norm": 0.140447735786438, + "learning_rate": 2.444044834510502e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005455055087804794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022977061104029416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013207816239446401, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7691320180892944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19150, + "weighted_flow_matching_loss": 0.012326624151319266, + "weighted_rkd_loss": 0.010158611321821808 + }, + { + "epoch": 5.574629036950829, + "flow_matching_loss": 0.011530610639601946, + "grad_norm": 0.1508587896823883, + "learning_rate": 2.443402226136096e-05, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005373615026473999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002414272306486964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013161502312868834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7689113438129425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19160, + "weighted_flow_matching_loss": 0.011530610639601946, + "weighted_rkd_loss": 0.0101200420409441 + }, + { + "epoch": 5.577538551061973, + "flow_matching_loss": 0.013775300607085228, + "grad_norm": 0.13706468045711517, + "learning_rate": 2.442759331181003e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005108793545514345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002232365729287267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012449952866882085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7686905860900879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19170, + "weighted_flow_matching_loss": 0.013775300607085228, + "weighted_rkd_loss": 0.009570200368762017 + }, + { + "epoch": 5.580448065173116, + "flow_matching_loss": 0.01264448333531618, + "grad_norm": 0.19992396235466003, + "learning_rate": 2.4421161498405176e-05, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051538543310016395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021030972944572567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012410805933177472, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.768469762802124, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19180, + "weighted_flow_matching_loss": 0.01264448333531618, + "weighted_rkd_loss": 0.009537352295592428 + }, + { + "epoch": 5.583357579284259, + "flow_matching_loss": 0.013690508529543876, + "grad_norm": 0.17166022956371307, + "learning_rate": 2.4414726823100207e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005332262115553021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002462188946083188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013126713410019874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7682488560676575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19190, + "weighted_flow_matching_loss": 0.013690508529543876, + "weighted_rkd_loss": 0.01008459022268653 + }, + { + "epoch": 5.586267093395403, + "flow_matching_loss": 0.014591455459594727, + "grad_norm": 0.1876498907804489, + "learning_rate": 2.4408289287849813e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004757412592880428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002035896445158869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011550721572712063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7680278897285462, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19200, + "weighted_flow_matching_loss": 0.014591455459594727, + "weighted_rkd_loss": 0.008871182752773165 + }, + { + "epoch": 5.5891766075065465, + "flow_matching_loss": 0.013392658531665802, + "grad_norm": 0.1577954739332199, + "learning_rate": 2.4401848894609543e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005211462778970599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002147690998390317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012570616789162159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7678068459033967, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19210, + "weighted_flow_matching_loss": 0.013392658531665802, + "weighted_rkd_loss": 0.009651711164042353 + }, + { + "epoch": 5.59208612161769, + "flow_matching_loss": 0.012837100867182016, + "grad_norm": 0.1491457223892212, + "learning_rate": 2.439540564533582e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006090618716552853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002860185015015304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015041422378271819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7675857245922089, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19220, + "weighted_flow_matching_loss": 0.012837100867182016, + "weighted_rkd_loss": 0.011545613408088684 + }, + { + "epoch": 5.594995635728833, + "flow_matching_loss": 0.014812830183655024, + "grad_norm": 0.1525060087442398, + "learning_rate": 2.4388959541985935e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053807795513421295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002301357930991799, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013062916882336139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7673645257949829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19230, + "weighted_flow_matching_loss": 0.014812830183655024, + "weighted_rkd_loss": 0.010024027805775404 + }, + { + "epoch": 5.597905149839977, + "flow_matching_loss": 0.012868628930300474, + "grad_norm": 0.1481487900018692, + "learning_rate": 2.438251058651804e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004957314464263618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020197335630655288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011934362445026636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7671432614326477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19240, + "weighted_flow_matching_loss": 0.012868628930300474, + "weighted_rkd_loss": 0.009155408246442676 + }, + { + "epoch": 5.6008146639511205, + "flow_matching_loss": 0.014105404634028674, + "grad_norm": 0.15911494195461273, + "learning_rate": 2.437605878089116e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004939199634827673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021710728644393383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012049472285434603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7669219195842742, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19250, + "weighted_flow_matching_loss": 0.014105404634028674, + "weighted_rkd_loss": 0.009241021471098065 + }, + { + "epoch": 5.603724178062263, + "flow_matching_loss": 0.01413688249886036, + "grad_norm": 0.17067696154117584, + "learning_rate": 2.4369604127065183e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005744088953360915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00269548911601305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01418366702273488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7667005062103271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19260, + "weighted_flow_matching_loss": 0.01413688249886036, + "weighted_rkd_loss": 0.01087466343306005 + }, + { + "epoch": 5.606633692173407, + "flow_matching_loss": 0.01432068208232522, + "grad_norm": 0.15417860448360443, + "learning_rate": 2.4363146627000862e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00470453838352114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019973250455223024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011406402010470629, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7664790213108063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19270, + "weighted_flow_matching_loss": 0.01432068208232522, + "weighted_rkd_loss": 0.00874284515157342 + }, + { + "epoch": 5.609543206284551, + "flow_matching_loss": 0.014350783172994852, + "grad_norm": 0.1403433382511139, + "learning_rate": 2.435668628265982e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004984056088142097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021604570909403266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012128569278866052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7662574589252472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19280, + "weighted_flow_matching_loss": 0.014350783172994852, + "weighted_rkd_loss": 0.009293551417067647 + }, + { + "epoch": 5.6124527203956935, + "flow_matching_loss": 0.011747314594686031, + "grad_norm": 0.2007618248462677, + "learning_rate": 2.4350223096004533e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005981846805661916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002824436197988689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014788130018860101, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7660358309745788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19290, + "weighted_flow_matching_loss": 0.011747314594686031, + "weighted_rkd_loss": 0.011328064510598779 + }, + { + "epoch": 5.615362234506837, + "flow_matching_loss": 0.013050493318587542, + "grad_norm": 0.17523305118083954, + "learning_rate": 2.4343757068998343e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005992793268524111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027894793660379947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014775065798312425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7658141255378723, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19300, + "weighted_flow_matching_loss": 0.013050493318587542, + "weighted_rkd_loss": 0.01131499488838017 + }, + { + "epoch": 5.618271748617981, + "flow_matching_loss": 0.013764634728431702, + "grad_norm": 0.18663446605205536, + "learning_rate": 2.4337288203605465e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005559264961630106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023712404654361308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013489770237356424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7655923426151275, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19310, + "weighted_flow_matching_loss": 0.013764634728431702, + "weighted_rkd_loss": 0.010327624622732401 + }, + { + "epoch": 5.621181262729124, + "flow_matching_loss": 0.013228446245193481, + "grad_norm": 0.15226396918296814, + "learning_rate": 2.433081650179097e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005053797480650246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002236347389407456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012343942373991012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7653704941272735, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19320, + "weighted_flow_matching_loss": 0.013228446245193481, + "weighted_rkd_loss": 0.00944770616479218 + }, + { + "epoch": 5.6240907768402675, + "flow_matching_loss": 0.012531034741550684, + "grad_norm": 0.1555829644203186, + "learning_rate": 2.4324341965520787e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0047936949646100405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002081904816441238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011669294629245996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7651485621929168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19330, + "weighted_flow_matching_loss": 0.012531034741550684, + "weighted_rkd_loss": 0.00892877634614706 + }, + { + "epoch": 5.627000290951411, + "flow_matching_loss": 0.013520735315978527, + "grad_norm": 0.12634439766407013, + "learning_rate": 2.4317864596761706e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005043562571518123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002158072357997298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01224519731476903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7649265825748444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19340, + "weighted_flow_matching_loss": 0.013520735315978527, + "weighted_rkd_loss": 0.00936673223040998 + }, + { + "epoch": 5.629909805062555, + "flow_matching_loss": 0.012548881396651269, + "grad_norm": 0.1342710256576538, + "learning_rate": 2.4311384397481387e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005106135364621877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021851190831512213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012397389952093362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7647045016288757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19350, + "weighted_flow_matching_loss": 0.012548881396651269, + "weighted_rkd_loss": 0.009480349253863096 + }, + { + "epoch": 5.632819319173698, + "flow_matching_loss": 0.012929858546704054, + "grad_norm": 0.15788546204566956, + "learning_rate": 2.4304901369648344e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054900154238566755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022778842016123236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013257915154099464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7644823610782623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19360, + "weighted_flow_matching_loss": 0.012929858546704054, + "weighted_rkd_loss": 0.010135391587391495 + }, + { + "epoch": 5.635728833284841, + "flow_matching_loss": 0.012094027083367109, + "grad_norm": 0.17607423663139343, + "learning_rate": 2.429841551523194e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005631240922957659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024302951525896786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013692777138203382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7642601430416107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19370, + "weighted_flow_matching_loss": 0.012094027083367109, + "weighted_rkd_loss": 0.010464816121384501 + }, + { + "epoch": 5.638638347395985, + "flow_matching_loss": 0.013441562838852405, + "grad_norm": 0.1557706743478775, + "learning_rate": 2.4291926836202416e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005496292281895876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023846492753364147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01337723359465599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7640378773212433, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19380, + "weighted_flow_matching_loss": 0.013441562838852405, + "weighted_rkd_loss": 0.010220774030312896 + }, + { + "epoch": 5.641547861507128, + "flow_matching_loss": 0.012991633545607329, + "grad_norm": 0.1420510858297348, + "learning_rate": 2.428543533453086e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005176612338982522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002320431882981211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012673656456172466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7638155043125152, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19390, + "weighted_flow_matching_loss": 0.012991633545607329, + "weighted_rkd_loss": 0.009680408239364623 + }, + { + "epoch": 5.644457375618272, + "flow_matching_loss": 0.012670303462073207, + "grad_norm": 0.1541522741317749, + "learning_rate": 2.4278941012189215e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004715479095466435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020345121854916217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011465470492839813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.763593065738678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19400, + "weighted_flow_matching_loss": 0.012670303462073207, + "weighted_rkd_loss": 0.0087549296207726 + }, + { + "epoch": 5.647366889729415, + "flow_matching_loss": 0.013711991813033819, + "grad_norm": 0.16288118064403534, + "learning_rate": 2.427244387115029e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005097851995378733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00223709101555869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01243279492482543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.763370567560196, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19410, + "weighted_flow_matching_loss": 0.013711991813033819, + "weighted_rkd_loss": 0.009490873711183668 + }, + { + "epoch": 5.650276403840559, + "flow_matching_loss": 0.013412455189973116, + "grad_norm": 0.14504261314868927, + "learning_rate": 2.4265943913387738e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00519243897870183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002308559895027429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012693437654525042, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7631479799747467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19420, + "weighted_flow_matching_loss": 0.013412455189973116, + "weighted_rkd_loss": 0.009687018487602473 + }, + { + "epoch": 5.653185917951702, + "flow_matching_loss": 0.013348280172795057, + "grad_norm": 0.1612953394651413, + "learning_rate": 2.425944114087608e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057855604216456415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00253869432490319, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014109815005213022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7629253387451171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19430, + "weighted_flow_matching_loss": 0.013348280172795057, + "weighted_rkd_loss": 0.010764785250648857 + }, + { + "epoch": 5.656095432062846, + "flow_matching_loss": 0.013655075524002313, + "grad_norm": 0.15251418948173523, + "learning_rate": 2.4252935555590684e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005022348975762725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023118646582588553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012356562726199626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7627026259899139, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19440, + "weighted_flow_matching_loss": 0.013655075524002313, + "weighted_rkd_loss": 0.009424427011981606 + }, + { + "epoch": 5.659004946173989, + "flow_matching_loss": 0.012044369429349899, + "grad_norm": 0.1557103544473648, + "learning_rate": 2.4246427159507772e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005120611353777349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023150772554799913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012556299846619368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7624798238277435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19450, + "weighted_flow_matching_loss": 0.012044369429349899, + "weighted_rkd_loss": 0.009573893016204238 + }, + { + "epoch": 5.661914460285132, + "flow_matching_loss": 0.011418160423636437, + "grad_norm": 0.18718774616718292, + "learning_rate": 2.4239915954604433e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004950993927195668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002126420382410288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012028408236801624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7622569739818573, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19460, + "weighted_flow_matching_loss": 0.011418160423636437, + "weighted_rkd_loss": 0.009168727975338697 + }, + { + "epoch": 5.664823974396276, + "flow_matching_loss": 0.014462783094495535, + "grad_norm": 0.18987926840782166, + "learning_rate": 2.4233401942858595e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005592072126455605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024739809217862786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013658125139772891, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7620340287685394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19470, + "weighted_flow_matching_loss": 0.014462783094495535, + "weighted_rkd_loss": 0.010408041067421436 + }, + { + "epoch": 5.6677334885074195, + "flow_matching_loss": 0.012266181409358978, + "grad_norm": 0.13396699726581573, + "learning_rate": 2.4226885126249033e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005353830405510962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002419166313484311, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013126827403903007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7618110239505768, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19480, + "weighted_flow_matching_loss": 0.012266181409358978, + "weighted_rkd_loss": 0.010000321688130498 + }, + { + "epoch": 5.670643002618562, + "flow_matching_loss": 0.013768524955958128, + "grad_norm": 0.18087700009346008, + "learning_rate": 2.4220365506755394e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048299582675099375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021784678800031545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011838384252041579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.761587941646576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19490, + "weighted_flow_matching_loss": 0.013768524955958128, + "weighted_rkd_loss": 0.009015881037339569 + }, + { + "epoch": 5.673552516729706, + "flow_matching_loss": 0.012634160462766886, + "grad_norm": 0.15639500319957733, + "learning_rate": 2.4213843086358166e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052497042808681725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024531146162189545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01295252311974764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7613647818565369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19500, + "weighted_flow_matching_loss": 0.012634160462766886, + "weighted_rkd_loss": 0.009861536137759686 + }, + { + "epoch": 5.67646203084085, + "flow_matching_loss": 0.012275764625519514, + "grad_norm": 0.17441613972187042, + "learning_rate": 2.4207317867038684e-05, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005042249080725014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023567174444906413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012441215571016073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.761141562461853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19510, + "weighted_flow_matching_loss": 0.012275764625519514, + "weighted_rkd_loss": 0.00946949077770114 + }, + { + "epoch": 5.679371544951993, + "flow_matching_loss": 0.012045525200664996, + "grad_norm": 0.16886857151985168, + "learning_rate": 2.4200789850779137e-05, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004880308546125889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022016841801814737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011962301097810268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7609182715415954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19520, + "weighted_flow_matching_loss": 0.012045525200664996, + "weighted_rkd_loss": 0.009102353500202299 + }, + { + "epoch": 5.682281059063136, + "flow_matching_loss": 0.013050381746143102, + "grad_norm": 0.16808821260929108, + "learning_rate": 2.419425903956257e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005126768606714904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002495170768816024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01274870801717043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7606949090957642, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19530, + "weighted_flow_matching_loss": 0.013050381746143102, + "weighted_rkd_loss": 0.00969785605557263 + }, + { + "epoch": 5.68519057317428, + "flow_matching_loss": 0.013810688350349664, + "grad_norm": 0.14763596653938293, + "learning_rate": 2.4187725435372865e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005664754658937454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024206201080232857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013750129472464323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7604714691638946, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19540, + "weighted_flow_matching_loss": 0.013810688350349664, + "weighted_rkd_loss": 0.010456554451957345 + }, + { + "epoch": 5.688100087285424, + "flow_matching_loss": 0.013477382622659207, + "grad_norm": 0.16334977746009827, + "learning_rate": 2.418118904019476e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005113519448786974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002187350729946047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012414389755576849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7602479696273804, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19550, + "weighted_flow_matching_loss": 0.013477382622659207, + "weighted_rkd_loss": 0.00943799247033894 + }, + { + "epoch": 5.6910096013965665, + "flow_matching_loss": 0.013798964023590089, + "grad_norm": 0.16878458857536316, + "learning_rate": 2.4174649856013842e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005447802413254976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024978979723528027, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013393502868711948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7600243747234344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19560, + "weighted_flow_matching_loss": 0.013798964023590089, + "weighted_rkd_loss": 0.01017938768491149 + }, + { + "epoch": 5.69391911550771, + "flow_matching_loss": 0.013761053700000048, + "grad_norm": 0.1679280549287796, + "learning_rate": 2.416810788481654e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0050109252100810405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022772553027607502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012299105711281299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7598007261753082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19570, + "weighted_flow_matching_loss": 0.013761053700000048, + "weighted_rkd_loss": 0.009344899980351328 + }, + { + "epoch": 5.696828629618854, + "flow_matching_loss": 0.013480752147734166, + "grad_norm": 0.19315190613269806, + "learning_rate": 2.416156312859013e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005936791282147169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002542884321883321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014416466653347015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7595770001411438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19580, + "weighted_flow_matching_loss": 0.013480752147734166, + "weighted_rkd_loss": 0.010950419493019581 + }, + { + "epoch": 5.699738143729997, + "flow_matching_loss": 0.012644321378320455, + "grad_norm": 0.14242780208587646, + "learning_rate": 2.4155015589322742e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005017224815674126, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002172296040225774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012206745613366366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7593532145023346, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19590, + "weighted_flow_matching_loss": 0.012644321378320455, + "weighted_rkd_loss": 0.009269231325015425 + }, + { + "epoch": 5.7026476578411405, + "flow_matching_loss": 0.014106580428779126, + "grad_norm": 0.19113963842391968, + "learning_rate": 2.4148465269003338e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005276672332547605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021167066995985808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012670051213353873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7591293394565582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19600, + "weighted_flow_matching_loss": 0.014106580428779126, + "weighted_rkd_loss": 0.009618338011205197 + }, + { + "epoch": 5.705557171952284, + "flow_matching_loss": 0.013583325408399105, + "grad_norm": 0.15274499356746674, + "learning_rate": 2.4141912169621742e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005184369487687945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020853187306784093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012454057577997446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7589054107666016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19610, + "weighted_flow_matching_loss": 0.013583325408399105, + "weighted_rkd_loss": 0.009451508754864335 + }, + { + "epoch": 5.708466686063428, + "flow_matching_loss": 0.013305506017059088, + "grad_norm": 0.2084663063287735, + "learning_rate": 2.4135356293168602e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005734733957797289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024178626132197677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013887330330908299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7586813986301422, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19620, + "weighted_flow_matching_loss": 0.013305506017059088, + "weighted_rkd_loss": 0.010536037012934684 + }, + { + "epoch": 5.711376200174571, + "flow_matching_loss": 0.014623877592384815, + "grad_norm": 0.14343740046024323, + "learning_rate": 2.4128797641635427e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004715797724202275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001975403830874711, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011406999174505472, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7584573030471802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19630, + "weighted_flow_matching_loss": 0.014623877592384815, + "weighted_rkd_loss": 0.008651732420548796 + }, + { + "epoch": 5.714285714285714, + "flow_matching_loss": 0.013584632147103547, + "grad_norm": 0.20874521136283875, + "learning_rate": 2.412223621701456e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005284924292936921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021432648412883283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01271311342716217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7582331717014312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19640, + "weighted_flow_matching_loss": 0.013584632147103547, + "weighted_rkd_loss": 0.009639495797455312 + }, + { + "epoch": 5.717195228396857, + "flow_matching_loss": 0.01465593371540308, + "grad_norm": 0.13747771084308624, + "learning_rate": 2.4115672021299193e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005375747988000512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021826958982273936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012934191804379224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7580089569091797, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19650, + "weighted_flow_matching_loss": 0.01465593371540308, + "weighted_rkd_loss": 0.009804262546822429 + }, + { + "epoch": 5.720104742508001, + "flow_matching_loss": 0.014729792438447475, + "grad_norm": 0.19191282987594604, + "learning_rate": 2.4109105056483345e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005955430120229721, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002855686005204916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014766546059399843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7577846586704254, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19660, + "weighted_flow_matching_loss": 0.014729792438447475, + "weighted_rkd_loss": 0.011189835239201784 + }, + { + "epoch": 5.723014256619145, + "flow_matching_loss": 0.012790380232036113, + "grad_norm": 0.17840372025966644, + "learning_rate": 2.4102535324561898e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005018996074795723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002193690068088472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012231682240962983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7575603067874909, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19670, + "weighted_flow_matching_loss": 0.012790380232036113, + "weighted_rkd_loss": 0.009266244806349277 + }, + { + "epoch": 5.725923770730288, + "flow_matching_loss": 0.01426217295229435, + "grad_norm": 0.1287832260131836, + "learning_rate": 2.4095962827530556e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005370638309977948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002340145781636238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013081422680988908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7573358654975891, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19680, + "weighted_flow_matching_loss": 0.01426217295229435, + "weighted_rkd_loss": 0.009906988451257348 + }, + { + "epoch": 5.728833284841431, + "flow_matching_loss": 0.013136059697717428, + "grad_norm": 0.16945938766002655, + "learning_rate": 2.408938756738587e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057893948629498485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002605682599823922, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01418447233736515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7571113646030426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19690, + "weighted_flow_matching_loss": 0.013136059697717428, + "weighted_rkd_loss": 0.010739304544404148 + }, + { + "epoch": 5.731742798952575, + "flow_matching_loss": 0.014010188076645136, + "grad_norm": 0.18368595838546753, + "learning_rate": 2.4082809546125235e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005269356863573193, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022666265023872255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012805340252816677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7568867862224579, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19700, + "weighted_flow_matching_loss": 0.014010188076645136, + "weighted_rkd_loss": 0.009692200645804405 + }, + { + "epoch": 5.734652313063719, + "flow_matching_loss": 0.012320189923048019, + "grad_norm": 0.13259492814540863, + "learning_rate": 2.4076228765746876e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004487143247388303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018192785093560815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010793564841151237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7566621601581573, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19710, + "weighted_flow_matching_loss": 0.012320189923048019, + "weighted_rkd_loss": 0.008167121279984712 + }, + { + "epoch": 5.737561827174861, + "flow_matching_loss": 0.012175183929502963, + "grad_norm": 0.15416933596134186, + "learning_rate": 2.4069645228249864e-05, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005189666640944779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022953562322072683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012674689665436745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7564374387264252, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19720, + "weighted_flow_matching_loss": 0.012175183929502963, + "weighted_rkd_loss": 0.009587689815089107 + }, + { + "epoch": 5.740471341286005, + "flow_matching_loss": 0.013661934714764356, + "grad_norm": 0.1579919308423996, + "learning_rate": 2.40630589356341e-05, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004703811998479068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022023751400411127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011609998997300863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7562126517295837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19730, + "weighted_flow_matching_loss": 0.013661934714764356, + "weighted_rkd_loss": 0.008779575675725937 + }, + { + "epoch": 5.743380855397149, + "flow_matching_loss": 0.013493560627102852, + "grad_norm": 0.16430425643920898, + "learning_rate": 2.4056469889900327e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005326314410194754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002178127190563828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012830755859613418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.755987799167633, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19740, + "weighted_flow_matching_loss": 0.013493560627102852, + "weighted_rkd_loss": 0.009699930530041456 + }, + { + "epoch": 5.7462903695082925, + "flow_matching_loss": 0.012984142638742924, + "grad_norm": 0.16233325004577637, + "learning_rate": 2.404987809305012e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005426720599643886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025187521358020605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013372193276882171, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7557628691196442, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19750, + "weighted_flow_matching_loss": 0.012984142638742924, + "weighted_rkd_loss": 0.010106079559773207 + }, + { + "epoch": 5.749199883619435, + "flow_matching_loss": 0.013507030718028546, + "grad_norm": 0.13904309272766113, + "learning_rate": 2.4043283547085903e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005398633843287826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023454644600860775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013142732065171003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7555378794670105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19760, + "weighted_flow_matching_loss": 0.013507030718028546, + "weighted_rkd_loss": 0.009929846785962582 + }, + { + "epoch": 5.752109397730579, + "flow_matching_loss": 0.012931023724377155, + "grad_norm": 0.14915651082992554, + "learning_rate": 2.403668625401092e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005743921315297484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026426377007737756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014130480121821165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7553128123283386, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19770, + "weighted_flow_matching_loss": 0.012931023724377155, + "weighted_rkd_loss": 0.010672958567738533 + }, + { + "epoch": 5.755018911841723, + "flow_matching_loss": 0.013393490575253963, + "grad_norm": 0.1510768085718155, + "learning_rate": 2.4030086215829247e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004747978621162474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002047628501895815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011543585546314716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7550876796245575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19780, + "weighted_flow_matching_loss": 0.013393490575253963, + "weighted_rkd_loss": 0.008716383296996356 + }, + { + "epoch": 5.757928425952866, + "flow_matching_loss": 0.01287637772038579, + "grad_norm": 0.14050906896591187, + "learning_rate": 2.4023483434545806e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053262949455529455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022826784756034614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01293526841327548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7548624753952027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19790, + "weighted_flow_matching_loss": 0.01287637772038579, + "weighted_rkd_loss": 0.009764301870018244 + }, + { + "epoch": 5.760837940064009, + "flow_matching_loss": 0.014916992280632257, + "grad_norm": 0.16182886064052582, + "learning_rate": 2.4016877912166356e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005002474412322044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00226613231934607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012271081283688545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7546371817588806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19800, + "weighted_flow_matching_loss": 0.014916992280632257, + "weighted_rkd_loss": 0.00926023251377046 + }, + { + "epoch": 5.763747454175153, + "flow_matching_loss": 0.014299446158111096, + "grad_norm": 0.14075753092765808, + "learning_rate": 2.401026965069747e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057240386959165335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002424935984890908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013873013388365506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7544118523597717, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19810, + "weighted_flow_matching_loss": 0.014299446158111096, + "weighted_rkd_loss": 0.010465915407985448 + }, + { + "epoch": 5.766656968286296, + "flow_matching_loss": 0.015033893380314111, + "grad_norm": 0.13517791032791138, + "learning_rate": 2.4003658652146564e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004785469477064907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021002839668653904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011671222746372223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7541864395141602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19820, + "weighted_flow_matching_loss": 0.015033893380314111, + "weighted_rkd_loss": 0.00880229203030467 + }, + { + "epoch": 5.7695664823974395, + "flow_matching_loss": 0.013590184226632118, + "grad_norm": 0.16869962215423584, + "learning_rate": 2.3997044918521896e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00547641827724874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021911061252467335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013143942784518003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7539609551429749, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19830, + "weighted_flow_matching_loss": 0.013590184226632118, + "weighted_rkd_loss": 0.009910054039210081 + }, + { + "epoch": 5.772475996508583, + "flow_matching_loss": 0.013482456095516682, + "grad_norm": 0.14551129937171936, + "learning_rate": 2.3990428451832524e-05, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004837030079215765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002160111046396196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011834171414375306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7537353992462158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19840, + "weighted_flow_matching_loss": 0.013482456095516682, + "weighted_rkd_loss": 0.008919905731454491 + }, + { + "epoch": 5.775385510619726, + "flow_matching_loss": 0.012515105679631234, + "grad_norm": 0.13047797977924347, + "learning_rate": 2.398380925408837e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004756764322519302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00212286637397483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011636395007371902, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.753509783744812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19850, + "weighted_flow_matching_loss": 0.012515105679631234, + "weighted_rkd_loss": 0.00876811663620174 + }, + { + "epoch": 5.77829502473087, + "flow_matching_loss": 0.013431811518967152, + "grad_norm": 0.1265474408864975, + "learning_rate": 2.3977187327300174e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005149172618985176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002152326609939337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012450671754777431, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.75328409075737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19860, + "weighted_flow_matching_loss": 0.013431811518967152, + "weighted_rkd_loss": 0.009379017679020763 + }, + { + "epoch": 5.7812045388420135, + "flow_matching_loss": 0.014346953062340618, + "grad_norm": 0.17083483934402466, + "learning_rate": 2.397056267347949e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005486517562530935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024667566642165182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013439791649580002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7530583202838897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19870, + "weighted_flow_matching_loss": 0.014346953062340618, + "weighted_rkd_loss": 0.01012080698274076 + }, + { + "epoch": 5.784114052953157, + "flow_matching_loss": 0.013825184386223554, + "grad_norm": 0.15803076326847076, + "learning_rate": 2.3963935294638725e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004874077788554132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020089234109036624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01175707895308733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7528325021266937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19880, + "weighted_flow_matching_loss": 0.013825184386223554, + "weighted_rkd_loss": 0.00885111871175468 + }, + { + "epoch": 5.7870235670643, + "flow_matching_loss": 0.013080287631601096, + "grad_norm": 0.17080406844615936, + "learning_rate": 2.395730519279109e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0049187344498932365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00216513064224273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012002599518746137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.752606600522995, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19890, + "weighted_flow_matching_loss": 0.013080287631601096, + "weighted_rkd_loss": 0.009033298259600997 + }, + { + "epoch": 5.789933081175444, + "flow_matching_loss": 0.012662669736891985, + "grad_norm": 0.14914196729660034, + "learning_rate": 2.3950672369950646e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004532950930297375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019896841840818524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011055586114525794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.752380621433258, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19900, + "weighted_flow_matching_loss": 0.012662669736891985, + "weighted_rkd_loss": 0.00831803111359477 + }, + { + "epoch": 5.792842595286587, + "flow_matching_loss": 0.013612690754234791, + "grad_norm": 0.1987670660018921, + "learning_rate": 2.3944036828132266e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00514701814390719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021450926666148006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012439128942787647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7521545827388764, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19910, + "weighted_flow_matching_loss": 0.013612690754234791, + "weighted_rkd_loss": 0.009356129635125398 + }, + { + "epoch": 5.79575210939773, + "flow_matching_loss": 0.013922927901148796, + "grad_norm": 0.14550545811653137, + "learning_rate": 2.3937398569351647e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005055556679144502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002230934996623546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012342048436403274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7519284784793854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19920, + "weighted_flow_matching_loss": 0.013922927901148796, + "weighted_rkd_loss": 0.009280451247468591 + }, + { + "epoch": 5.798661623508874, + "flow_matching_loss": 0.011266840808093547, + "grad_norm": 0.16064980626106262, + "learning_rate": 2.3930757595625326e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005146059114485979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002090478129684925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012382596358656883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7517023086547852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19930, + "weighted_flow_matching_loss": 0.011266840808093547, + "weighted_rkd_loss": 0.009307930618524552 + }, + { + "epoch": 5.801571137620018, + "flow_matching_loss": 0.012225812021642923, + "grad_norm": 0.1102820336818695, + "learning_rate": 2.3924113908970645e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004644917999394238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019666670938022434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011256503220647574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7514760553836822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19940, + "weighted_flow_matching_loss": 0.012225812021642923, + "weighted_rkd_loss": 0.008459028741344809 + }, + { + "epoch": 5.804480651731161, + "flow_matching_loss": 0.012781005166471005, + "grad_norm": 0.15490858256816864, + "learning_rate": 2.391746751140579e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004446095344610513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001930565608199686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010822756309062242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7512497425079345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19950, + "weighted_flow_matching_loss": 0.012781005166471005, + "weighted_rkd_loss": 0.008130547916516661 + }, + { + "epoch": 5.807390165842304, + "flow_matching_loss": 0.014334164839237929, + "grad_norm": 0.16911517083644867, + "learning_rate": 2.3910818404949756e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058948618825525045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027986689819954336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01458839289844036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7510233521461487, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19960, + "weighted_flow_matching_loss": 0.014334164839237929, + "weighted_rkd_loss": 0.010956302424892784 + }, + { + "epoch": 5.810299679953448, + "flow_matching_loss": 0.014060529880225658, + "grad_norm": 0.1553398072719574, + "learning_rate": 2.3904166591622367e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004836657922714949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002176478039473295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011849793791770934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.750796914100647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19970, + "weighted_flow_matching_loss": 0.014060529880225658, + "weighted_rkd_loss": 0.00889679896645248 + }, + { + "epoch": 5.813209194064592, + "flow_matching_loss": 0.01370185511186719, + "grad_norm": 0.14861224591732025, + "learning_rate": 2.3897512073444267e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005546856950968504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002262843865901232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013356557860970498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7505703806877136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19980, + "weighted_flow_matching_loss": 0.01370185511186719, + "weighted_rkd_loss": 0.010024986881762742 + }, + { + "epoch": 5.816118708175734, + "flow_matching_loss": 0.014543966390192509, + "grad_norm": 0.13680791854858398, + "learning_rate": 2.389085485243693e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004789209668524563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021386007661931216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011717020161449909, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7503437936306, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19990, + "weighted_flow_matching_loss": 0.014543966390192509, + "weighted_rkd_loss": 0.008791802078485489 + }, + { + "epoch": 5.819028222286878, + "flow_matching_loss": 0.014065312780439853, + "grad_norm": 0.1449681967496872, + "learning_rate": 2.3884194930622632e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00498334439471364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002209158253390342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012175847124308348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7501171350479126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20000, + "weighted_flow_matching_loss": 0.014065312780439853, + "weighted_rkd_loss": 0.009133358858525753 + }, + { + "epoch": 5.821937736398022, + "flow_matching_loss": 0.01524203596636653, + "grad_norm": 0.16281920671463013, + "learning_rate": 2.3877532310024488e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005208627879619598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002303971187211573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012721226830035449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7498904049396515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20010, + "weighted_flow_matching_loss": 0.01524203596636653, + "weighted_rkd_loss": 0.009539552638307214 + }, + { + "epoch": 5.824847250509165, + "flow_matching_loss": 0.014263860322535038, + "grad_norm": 0.13862864673137665, + "learning_rate": 2.3870866992666428e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005461149103939533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002493070182390511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01341536846011877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7496636092662812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20020, + "weighted_flow_matching_loss": 0.014263860322535038, + "weighted_rkd_loss": 0.010057068010792137 + }, + { + "epoch": 5.827756764620308, + "flow_matching_loss": 0.012134916055947543, + "grad_norm": 0.17364434897899628, + "learning_rate": 2.3864198980573196e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005043230880983174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002174934290815145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01226139608770609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7494367361068726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20030, + "weighted_flow_matching_loss": 0.012134916055947543, + "weighted_rkd_loss": 0.00918911574408412 + }, + { + "epoch": 5.830666278731452, + "flow_matching_loss": 0.013307103607803582, + "grad_norm": 0.16153143346309662, + "learning_rate": 2.385752827577036e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005395167227834463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00253596423426643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013326298631727695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7492098093032837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20040, + "weighted_flow_matching_loss": 0.013307103607803582, + "weighted_rkd_loss": 0.009984180890023708 + }, + { + "epoch": 5.833575792842595, + "flow_matching_loss": 0.013701267912983895, + "grad_norm": 0.16909414529800415, + "learning_rate": 2.38508548802843e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005117932450957596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023004051530733705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012536270078271627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7489828109741211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20050, + "weighted_flow_matching_loss": 0.013701267912983895, + "weighted_rkd_loss": 0.009389445651322603 + }, + { + "epoch": 5.836485306953739, + "flow_matching_loss": 0.012397577706724405, + "grad_norm": 0.15951870381832123, + "learning_rate": 2.3844178796142226e-05, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005111380154266954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002330991020426154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012553751189261674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7487557232379913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20060, + "weighted_flow_matching_loss": 0.012397577706724405, + "weighted_rkd_loss": 0.009399649873375892 + }, + { + "epoch": 5.839394821064882, + "flow_matching_loss": 0.012620222102850676, + "grad_norm": 0.17723435163497925, + "learning_rate": 2.383750002537215e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005303345061838627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002261334168724716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012868024129420518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7485286056995392, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20070, + "weighted_flow_matching_loss": 0.012620222102850676, + "weighted_rkd_loss": 0.009631966426968574 + }, + { + "epoch": 5.842304335176026, + "flow_matching_loss": 0.012611347064375878, + "grad_norm": 0.1642204225063324, + "learning_rate": 2.383081857000291e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005799640761688351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002753763517830521, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014353045076131821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7483013927936554, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20080, + "weighted_flow_matching_loss": 0.012611347064375878, + "weighted_rkd_loss": 0.010740301851183177 + }, + { + "epoch": 5.845213849287169, + "flow_matching_loss": 0.01381416218355298, + "grad_norm": 0.13042496144771576, + "learning_rate": 2.3824134432064145e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005096803815104067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002169242827221751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012362850457429886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7480741202831268, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20090, + "weighted_flow_matching_loss": 0.01381416218355298, + "weighted_rkd_loss": 0.009248345484957099 + }, + { + "epoch": 5.8481233633983125, + "flow_matching_loss": 0.012728171423077583, + "grad_norm": 0.22747093439102173, + "learning_rate": 2.3817447613586332e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005010475893504918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002197065460495651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012218017224222421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.747846782207489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20100, + "weighted_flow_matching_loss": 0.012728171423077583, + "weighted_rkd_loss": 0.009137295000255109 + }, + { + "epoch": 5.851032877509456, + "flow_matching_loss": 0.013964951783418656, + "grad_norm": 0.14414435625076294, + "learning_rate": 2.381075811660074e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005224815034307539, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023335946607403457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012783224787563085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.747619366645813, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20110, + "weighted_flow_matching_loss": 0.013964951783418656, + "weighted_rkd_loss": 0.009557022387161851 + }, + { + "epoch": 5.853942391620599, + "flow_matching_loss": 0.014768268447369337, + "grad_norm": 0.14320406317710876, + "learning_rate": 2.3804065943139473e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005273027461953461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020579288015142084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012603983702138066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7473918855190277, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20120, + "weighted_flow_matching_loss": 0.014768268447369337, + "weighted_rkd_loss": 0.009420277271419763 + }, + { + "epoch": 5.856851905731743, + "flow_matching_loss": 0.011688006017357111, + "grad_norm": 0.15941207110881805, + "learning_rate": 2.379737109523543e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004954005335457623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021736951428465547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012081705778837205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7471643447875976, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20130, + "weighted_flow_matching_loss": 0.011688006017357111, + "weighted_rkd_loss": 0.009027019329369067 + }, + { + "epoch": 5.8597614198428865, + "flow_matching_loss": 0.012983298674225808, + "grad_norm": 0.162258118391037, + "learning_rate": 2.3790673574922324e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004946283134631813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021111547132022677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012003720924258231, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7469367384910583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20140, + "weighted_flow_matching_loss": 0.012983298674225808, + "weighted_rkd_loss": 0.008965968294069171 + }, + { + "epoch": 5.86267093395403, + "flow_matching_loss": 0.014620037097483874, + "grad_norm": 0.16961225867271423, + "learning_rate": 2.3783973384234695e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053855150006711485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002304108371026814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013075138349086047, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7467090487480164, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20150, + "weighted_flow_matching_loss": 0.014620037097483874, + "weighted_rkd_loss": 0.009763325983658432 + }, + { + "epoch": 5.865580448065173, + "flow_matching_loss": 0.012964253593236209, + "grad_norm": 0.1707250326871872, + "learning_rate": 2.3777270525207876e-05, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004901198716834187, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020183758577331902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011820773128420115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.746481305360794, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20160, + "weighted_flow_matching_loss": 0.012964253593236209, + "weighted_rkd_loss": 0.008824067749083042 + }, + { + "epoch": 5.868489962176317, + "flow_matching_loss": 0.013146120496094227, + "grad_norm": 0.16152964532375336, + "learning_rate": 2.3770564999878024e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004303711443208158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018497042008675634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010457127122208476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.746253490447998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20170, + "weighted_flow_matching_loss": 0.013146120496094227, + "weighted_rkd_loss": 0.00780368996784091 + }, + { + "epoch": 5.8713994762874595, + "flow_matching_loss": 0.013506533950567246, + "grad_norm": 0.18346618115901947, + "learning_rate": 2.37638568102821e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005102641624398529, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002391878457274288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012597161624580622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7460256040096283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20180, + "weighted_flow_matching_loss": 0.013506533950567246, + "weighted_rkd_loss": 0.00939783719368279 + }, + { + "epoch": 5.874308990398603, + "flow_matching_loss": 0.013510940037667752, + "grad_norm": 0.14762729406356812, + "learning_rate": 2.3757145958457868e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00471968031488359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002047492505516857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011486853193491697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7457976579666138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20190, + "weighted_flow_matching_loss": 0.013510940037667752, + "weighted_rkd_loss": 0.008566891681402922 + }, + { + "epoch": 5.877218504509747, + "flow_matching_loss": 0.012657259497791529, + "grad_norm": 0.16327109932899475, + "learning_rate": 2.375043244644392e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005423640413209796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002600860642269254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013448141608387231, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7455696344375611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20200, + "weighted_flow_matching_loss": 0.012657259497791529, + "weighted_rkd_loss": 0.01002645273692906 + }, + { + "epoch": 5.880128018620891, + "flow_matching_loss": 0.013928319420665502, + "grad_norm": 0.163408562541008, + "learning_rate": 2.374371627627963e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004728667926974595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020063598756678404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011463695578277111, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7453415513038635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20210, + "weighted_flow_matching_loss": 0.013928319420665502, + "weighted_rkd_loss": 0.008544349763542413 + }, + { + "epoch": 5.8830375327320334, + "flow_matching_loss": 0.013036615308374166, + "grad_norm": 0.13913612067699432, + "learning_rate": 2.37369974500052e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005838691582903266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024144850322045386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014091868046671152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7451133906841279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20220, + "weighted_flow_matching_loss": 0.013036615308374166, + "weighted_rkd_loss": 0.010499901417642832 + }, + { + "epoch": 5.885947046843177, + "flow_matching_loss": 0.012513627391308546, + "grad_norm": 0.14150266349315643, + "learning_rate": 2.373027596966164e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004875841923058033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021796445245854555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011931328289210796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7448851764202118, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20230, + "weighted_flow_matching_loss": 0.012513627391308546, + "weighted_rkd_loss": 0.008887490537017585 + }, + { + "epoch": 5.888856560954321, + "flow_matching_loss": 0.012605374213308096, + "grad_norm": 0.1618567556142807, + "learning_rate": 2.3723551837290744e-05, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004659617622382939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002097141626290977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011416377034038305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.744656890630722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20240, + "weighted_flow_matching_loss": 0.012605374213308096, + "weighted_rkd_loss": 0.008501279726624488 + }, + { + "epoch": 5.891766075065464, + "flow_matching_loss": 0.014639812335371972, + "grad_norm": 0.17409545183181763, + "learning_rate": 2.3716825054935133e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005550956446677446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002446371573023498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013548284396529197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.744428527355194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20250, + "weighted_flow_matching_loss": 0.014639812335371972, + "weighted_rkd_loss": 0.010085651138797402 + }, + { + "epoch": 5.894675589176607, + "flow_matching_loss": 0.012755702808499336, + "grad_norm": 0.15727773308753967, + "learning_rate": 2.371009562463822e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004955116682685912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019491540384478866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011859387438744307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7442001044750214, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20260, + "weighted_flow_matching_loss": 0.012755702808499336, + "weighted_rkd_loss": 0.008825832279399037 + }, + { + "epoch": 5.897585103287751, + "flow_matching_loss": 0.013288384303450584, + "grad_norm": 0.15434175729751587, + "learning_rate": 2.3703363548444238e-05, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004664363898336887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019546128925867377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011283340584486722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7439716219902038, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20270, + "weighted_flow_matching_loss": 0.013288384303450584, + "weighted_rkd_loss": 0.008394576003775 + }, + { + "epoch": 5.900494617398895, + "flow_matching_loss": 0.014669043663889169, + "grad_norm": 0.16799770295619965, + "learning_rate": 2.3696628828398208e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005297786975279451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002333007811103016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012928581610322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7437430620193481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20280, + "weighted_flow_matching_loss": 0.014669043663889169, + "weighted_rkd_loss": 0.009615595312789082 + }, + { + "epoch": 5.903404131510038, + "flow_matching_loss": 0.013080681953579188, + "grad_norm": 0.17860889434814453, + "learning_rate": 2.368989146654596e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052580771734938025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024132249876856805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012929379288107157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7435144424438477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20290, + "weighted_flow_matching_loss": 0.013080681953579188, + "weighted_rkd_loss": 0.009613144863396882 + }, + { + "epoch": 5.906313645621181, + "flow_matching_loss": 0.012319776695221663, + "grad_norm": 0.15870417654514313, + "learning_rate": 2.368315146493413e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004936857800930739, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025993661605753005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012473081704229116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7432857573032379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20300, + "weighted_flow_matching_loss": 0.012319776695221663, + "weighted_rkd_loss": 0.009271146170794964 + }, + { + "epoch": 5.909223159732325, + "flow_matching_loss": 0.013973489683121443, + "grad_norm": 0.13718518614768982, + "learning_rate": 2.367640882561015e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00484128191601485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020676013897173106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011750165186822415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7430569946765899, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20310, + "weighted_flow_matching_loss": 0.013973489683121443, + "weighted_rkd_loss": 0.008731045294553041 + }, + { + "epoch": 5.912132673843468, + "flow_matching_loss": 0.013668443635106086, + "grad_norm": 0.1676899641752243, + "learning_rate": 2.3669663550622255e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0049363524885848165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022660913760773837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012138796225190162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7428281903266907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20320, + "weighted_flow_matching_loss": 0.013668443635106086, + "weighted_rkd_loss": 0.009016976179555058 + }, + { + "epoch": 5.915042187954612, + "flow_matching_loss": 0.01335065858438611, + "grad_norm": 0.1736045777797699, + "learning_rate": 2.3662915642019482e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004820868535898626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019286003196612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011570337507873774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7425992906093597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20330, + "weighted_flow_matching_loss": 0.01335065858438611, + "weighted_rkd_loss": 0.008592096529901028 + }, + { + "epoch": 5.917951702065755, + "flow_matching_loss": 0.013401609938591718, + "grad_norm": 0.12327531725168228, + "learning_rate": 2.365616510185167e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004976402618922293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002253543981350958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012206349149346351, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7423703372478485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20340, + "weighted_flow_matching_loss": 0.013401609938591718, + "weighted_rkd_loss": 0.009061646042391657 + }, + { + "epoch": 5.920861216176898, + "flow_matching_loss": 0.014113605301827193, + "grad_norm": 0.15351471304893494, + "learning_rate": 2.364941193216945e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004798269621096552, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002089183812495321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011685723252594472, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7421413123607635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20350, + "weighted_flow_matching_loss": 0.014113605301827193, + "weighted_rkd_loss": 0.008672478096559643 + }, + { + "epoch": 5.923770730288042, + "flow_matching_loss": 0.013220111094415187, + "grad_norm": 0.12691745162010193, + "learning_rate": 2.364265613502426e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004795371810905636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021192891988903286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011710032820701599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7419122219085693, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20360, + "weighted_flow_matching_loss": 0.013220111094415187, + "weighted_rkd_loss": 0.008687876304611564 + }, + { + "epoch": 5.9266802443991855, + "flow_matching_loss": 0.012795033119618892, + "grad_norm": 0.15228399634361267, + "learning_rate": 2.3635897712468333e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005584183614701033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024243030231446026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013592670299112796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7416830658912659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20370, + "weighted_flow_matching_loss": 0.012795033119618892, + "weighted_rkd_loss": 0.010081461211666465 + }, + { + "epoch": 5.929589758510328, + "flow_matching_loss": 0.013081392087042332, + "grad_norm": 0.15799711644649506, + "learning_rate": 2.36291366665547e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004558913689106703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019948140368796887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011112641356885434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7414538443088532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20380, + "weighted_flow_matching_loss": 0.013081392087042332, + "weighted_rkd_loss": 0.008239457150921226 + }, + { + "epoch": 5.932499272621472, + "flow_matching_loss": 0.013073344063013793, + "grad_norm": 0.1727941483259201, + "learning_rate": 2.362237299933719e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005109325400553644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002262863307259977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012481514271348715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7412245571613312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20390, + "weighted_flow_matching_loss": 0.013073344063013793, + "weighted_rkd_loss": 0.009251601994037628 + }, + { + "epoch": 5.935408786732616, + "flow_matching_loss": 0.013622258231043816, + "grad_norm": 0.1827448159456253, + "learning_rate": 2.3615606712870414e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005038353474810719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002335042168851942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012411748990416527, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7409952104091644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20400, + "weighted_flow_matching_loss": 0.013622258231043816, + "weighted_rkd_loss": 0.009197037713602186 + }, + { + "epoch": 5.9383183008437594, + "flow_matching_loss": 0.01408994449302554, + "grad_norm": 0.29952272772789, + "learning_rate": 2.3608837809209807e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00490453839302063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002033326355740428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011842403188347816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.740765780210495, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20410, + "weighted_flow_matching_loss": 0.01408994449302554, + "weighted_rkd_loss": 0.00877236477099359 + }, + { + "epoch": 5.941227814954902, + "flow_matching_loss": 0.012971487827599049, + "grad_norm": 0.13742156326770782, + "learning_rate": 2.360206629041157e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004582264786586165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00206753391539678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011232063546776772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7405363023281097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20420, + "weighted_flow_matching_loss": 0.012971487827599049, + "weighted_rkd_loss": 0.008317736117169261 + }, + { + "epoch": 5.944137329066046, + "flow_matching_loss": 0.01279569547623396, + "grad_norm": 0.16085997223854065, + "learning_rate": 2.359529215853273e-05, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053699272219091656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022843559156171976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013024210371077061, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7403067469596862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20430, + "weighted_flow_matching_loss": 0.01279569547623396, + "weighted_rkd_loss": 0.009641875559464098 + }, + { + "epoch": 5.94704684317719, + "flow_matching_loss": 0.013977851159870625, + "grad_norm": 0.1533249467611313, + "learning_rate": 2.358851541563107e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004828350385650992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022225902299396693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011879291199147702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7400771319866181, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20440, + "weighted_flow_matching_loss": 0.013977851159870625, + "weighted_rkd_loss": 0.008791614044457673 + }, + { + "epoch": 5.9499563572883325, + "flow_matching_loss": 0.012920169346034526, + "grad_norm": 0.1678624302148819, + "learning_rate": 2.3581736063765193e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004806666565127671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020976455300115047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011710978671908378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7398474514484406, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20450, + "weighted_flow_matching_loss": 0.012920169346034526, + "weighted_rkd_loss": 0.008664411678910255 + }, + { + "epoch": 5.952865871399476, + "flow_matching_loss": 0.013295167032629252, + "grad_norm": 0.1642119288444519, + "learning_rate": 2.3574954104994492e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00444883331656456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002077827404718846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010975493956357241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7396177053451538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20460, + "weighted_flow_matching_loss": 0.013295167032629252, + "weighted_rkd_loss": 0.008117668982595205 + }, + { + "epoch": 5.95577538551062, + "flow_matching_loss": 0.0133049669675529, + "grad_norm": 0.13295100629329681, + "learning_rate": 2.3568169541379137e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005175011325627565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00230203487444669, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012652057502418757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7393878936767578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20470, + "weighted_flow_matching_loss": 0.0133049669675529, + "weighted_rkd_loss": 0.009354819962754845 + }, + { + "epoch": 5.958684899621764, + "flow_matching_loss": 0.012257182598114013, + "grad_norm": 0.1482134908437729, + "learning_rate": 2.356138237498011e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004782059439457953, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020874136011116208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011651532631367445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7391580045223236, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20480, + "weighted_flow_matching_loss": 0.012257182598114013, + "weighted_rkd_loss": 0.008612178638577462 + }, + { + "epoch": 5.961594413732906, + "flow_matching_loss": 0.013189211580902338, + "grad_norm": 0.15517844259738922, + "learning_rate": 2.3554592607859166e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004940130771137774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00212178643560037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01200204798951745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7389280557632446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20490, + "weighted_flow_matching_loss": 0.013189211580902338, + "weighted_rkd_loss": 0.008868672139942645 + }, + { + "epoch": 5.96450392784405, + "flow_matching_loss": 0.012473736796528101, + "grad_norm": 0.20111492276191711, + "learning_rate": 2.3547800242078857e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005694829672574997, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025805989280343056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013970258319750429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7386980593204499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20500, + "weighted_flow_matching_loss": 0.012473736796528101, + "weighted_rkd_loss": 0.010319768823683262 + }, + { + "epoch": 5.967413441955194, + "flow_matching_loss": 0.01309699546545744, + "grad_norm": 0.13434304296970367, + "learning_rate": 2.3541005279702528e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004714618250727653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002038157684728503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01146739423274994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7384679794311524, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20510, + "weighted_flow_matching_loss": 0.01309699546545744, + "weighted_rkd_loss": 0.00846830876544118 + }, + { + "epoch": 5.970322956066337, + "flow_matching_loss": 0.013877223152667284, + "grad_norm": 0.16282199323177338, + "learning_rate": 2.3534207722794307e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005153843248263001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021454652887769043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01245315158739686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7382378339767456, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20520, + "weighted_flow_matching_loss": 0.013877223152667284, + "weighted_rkd_loss": 0.009193412307649852 + }, + { + "epoch": 5.97323247017748, + "flow_matching_loss": 0.013017244730144739, + "grad_norm": 0.1397716999053955, + "learning_rate": 2.352740757341911e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004451044951565564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018327344441786408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010734824184328317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7380076289176941, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20530, + "weighted_flow_matching_loss": 0.013017244730144739, + "weighted_rkd_loss": 0.007922434713691473 + }, + { + "epoch": 5.976141984288624, + "flow_matching_loss": 0.013755477871745825, + "grad_norm": 0.1859978884458542, + "learning_rate": 2.3520604833642645e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004932874185033143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022389525547623634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012104700971394777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7377773523330688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20540, + "weighted_flow_matching_loss": 0.013755477871745825, + "weighted_rkd_loss": 0.008930571004748344 + }, + { + "epoch": 5.979051498399767, + "flow_matching_loss": 0.012409307295456529, + "grad_norm": 0.16977588832378387, + "learning_rate": 2.3513799505531407e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004730425775051117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020892067113891244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011550058145076036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7375470101833344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20550, + "weighted_flow_matching_loss": 0.012409307295456529, + "weighted_rkd_loss": 0.008518826216459274 + }, + { + "epoch": 5.981961012510911, + "flow_matching_loss": 0.014175507891923188, + "grad_norm": 0.12756392359733582, + "learning_rate": 2.350699159115268e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005278858565725386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022797200828790663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01283743716776371, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7373166143894195, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20560, + "weighted_flow_matching_loss": 0.014175507891923188, + "weighted_rkd_loss": 0.009465363901108503 + }, + { + "epoch": 5.984870526622054, + "flow_matching_loss": 0.014015078824013472, + "grad_norm": 0.15702825784683228, + "learning_rate": 2.350018109257451e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005170168704353273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022706744144670664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012611011788249017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.737086147069931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20570, + "weighted_flow_matching_loss": 0.014015078824013472, + "weighted_rkd_loss": 0.00929544228129089 + }, + { + "epoch": 5.987780040733197, + "flow_matching_loss": 0.012856981251388788, + "grad_norm": 0.1537819802761078, + "learning_rate": 2.3493368011865766e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005170251289382577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021962961996905505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012536798790097236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7368556141853333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20580, + "weighted_flow_matching_loss": 0.012856981251388788, + "weighted_rkd_loss": 0.009237822145223618 + }, + { + "epoch": 5.990689554844341, + "flow_matching_loss": 0.012692191172391176, + "grad_norm": 0.1884811520576477, + "learning_rate": 2.348655235109607e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005427611665800214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024186950293369593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013273918442428112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7366250157356262, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20590, + "weighted_flow_matching_loss": 0.012692191172391176, + "weighted_rkd_loss": 0.009777711471542716 + }, + { + "epoch": 5.9935990689554846, + "flow_matching_loss": 0.01253961380571127, + "grad_norm": 0.17775604128837585, + "learning_rate": 2.3479734112335845e-05, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00535894613713026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002401137026026845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013119029253721238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7363943576812744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20600, + "weighted_flow_matching_loss": 0.01253961380571127, + "weighted_rkd_loss": 0.009660696750506759 + }, + { + "epoch": 5.996508583066628, + "flow_matching_loss": 0.013353666104376317, + "grad_norm": 0.13651418685913086, + "learning_rate": 2.347291329765629e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00477493773214519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020627355203032493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01161261098459363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7361636221408844, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20610, + "weighted_flow_matching_loss": 0.013353666104376317, + "weighted_rkd_loss": 0.00854881787672639 + }, + { + "epoch": 5.999418097177771, + "flow_matching_loss": 0.012560865981504321, + "grad_norm": 0.14218586683273315, + "learning_rate": 2.3466089909129396e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005739491945132613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002624816435854882, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014103800058364868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7359328269958496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20620, + "weighted_flow_matching_loss": 0.012560865981504321, + "weighted_rkd_loss": 0.010379481827840209 + }, + { + "epoch": 6.002327611288915, + "flow_matching_loss": 0.013315926864743232, + "grad_norm": 0.15852919220924377, + "learning_rate": 2.3459263948827916e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045526054920628665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002006291155703366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01111150225624442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7357019782066345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20630, + "weighted_flow_matching_loss": 0.013315926864743232, + "weighted_rkd_loss": 0.008174752537161111 + }, + { + "epoch": 6.0052371254000585, + "flow_matching_loss": 0.012914715800434351, + "grad_norm": 0.18536078929901123, + "learning_rate": 2.3452435418825404e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005281771440058947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002208968985360116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01277251187711954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7354710519313812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20640, + "weighted_flow_matching_loss": 0.012914715800434351, + "weighted_rkd_loss": 0.009393750922754407 + }, + { + "epoch": 6.008146639511201, + "flow_matching_loss": 0.014519147016108036, + "grad_norm": 0.15300355851650238, + "learning_rate": 2.344560432119618e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005158289801329374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002290418918710202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01260699862614274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7352400600910187, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20650, + "weighted_flow_matching_loss": 0.014519147016108036, + "weighted_rkd_loss": 0.009269138611853123 + }, + { + "epoch": 6.011056153622345, + "flow_matching_loss": 0.013173816539347172, + "grad_norm": 0.14044740796089172, + "learning_rate": 2.3438770658015368e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004266291507519782, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001842736778780818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010375319793820382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7350090205669403, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20660, + "weighted_flow_matching_loss": 0.013173816539347172, + "weighted_rkd_loss": 0.00762589480727911 + }, + { + "epoch": 6.013965667733489, + "flow_matching_loss": 0.014686116576194763, + "grad_norm": 0.15443633496761322, + "learning_rate": 2.3431934431358835e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005280251475051045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002369794144760817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012930297013372182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7347778975963593, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20670, + "weighted_flow_matching_loss": 0.014686116576194763, + "weighted_rkd_loss": 0.009500858373939991 + }, + { + "epoch": 6.0168751818446315, + "flow_matching_loss": 0.011882215831428766, + "grad_norm": 0.14559690654277802, + "learning_rate": 2.342509564330325e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004623237950727344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019259496941231192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01117242556065321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7345467209815979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20680, + "weighted_flow_matching_loss": 0.011882215831428766, + "weighted_rkd_loss": 0.00820665149949491 + }, + { + "epoch": 6.019784695955775, + "flow_matching_loss": 0.014162594731897116, + "grad_norm": 0.17780235409736633, + "learning_rate": 2.3418254295926068e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005189232667908073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002149579243268818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01252804473042488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7343154549598694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20690, + "weighted_flow_matching_loss": 0.014162594731897116, + "weighted_rkd_loss": 0.009199451934546233 + }, + { + "epoch": 6.022694210066919, + "flow_matching_loss": 0.012712740618735553, + "grad_norm": 0.15915724635124207, + "learning_rate": 2.34114103913055e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005404577753506601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026812672847881915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013490422908216716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7340841710567474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20700, + "weighted_flow_matching_loss": 0.012712740618735553, + "weighted_rkd_loss": 0.009903247514739633 + }, + { + "epoch": 6.025603724178063, + "flow_matching_loss": 0.013347579538822174, + "grad_norm": 0.13201670348644257, + "learning_rate": 2.3404563931520545e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057441147044301035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024437821819446983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013932011649012565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7338527977466583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20710, + "weighted_flow_matching_loss": 0.013347579538822174, + "weighted_rkd_loss": 0.010224025277420879 + }, + { + "epoch": 6.0285132382892055, + "flow_matching_loss": 0.012284539826214313, + "grad_norm": 0.1483333557844162, + "learning_rate": 2.3397714918650982e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048626626143231984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021516803186386823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01187700554728508, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.73362135887146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20720, + "weighted_flow_matching_loss": 0.012284539826214313, + "weighted_rkd_loss": 0.008713212329894304 + }, + { + "epoch": 6.031422752400349, + "flow_matching_loss": 0.013448209688067436, + "grad_norm": 0.11637214571237564, + "learning_rate": 2.3390863354777346e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005344930524006486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023133751237764955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013003236427903176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7333898603916168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20730, + "weighted_flow_matching_loss": 0.013448209688067436, + "weighted_rkd_loss": 0.009536524256691336 + }, + { + "epoch": 6.034332266511493, + "flow_matching_loss": 0.01258912468329072, + "grad_norm": 0.14824914932250977, + "learning_rate": 2.338400924198098e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00487475770059973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021119915531016887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011861507035791873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7331582963466644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20740, + "weighted_flow_matching_loss": 0.01258912468329072, + "weighted_rkd_loss": 0.008696435065940023 + }, + { + "epoch": 6.037241780622636, + "flow_matching_loss": 0.01348623689264059, + "grad_norm": 0.13678638637065887, + "learning_rate": 2.3377152582343963e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005275433999486268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024230081820860506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01297387620434165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7329266667366028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20750, + "weighted_flow_matching_loss": 0.01348623689264059, + "weighted_rkd_loss": 0.009508930100128054 + }, + { + "epoch": 6.040151294733779, + "flow_matching_loss": 0.013671832531690598, + "grad_norm": 0.13399766385555267, + "learning_rate": 2.3370293377949184e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005110398819670081, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002374983381014317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012595781031996012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7326949775218964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20760, + "weighted_flow_matching_loss": 0.013671832531690598, + "weighted_rkd_loss": 0.00922894962131977 + }, + { + "epoch": 6.043060808844923, + "flow_matching_loss": 0.01323337322100997, + "grad_norm": 0.1518498808145523, + "learning_rate": 2.3363431630880282e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005334235355257988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022561507765203716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012924621533602475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7324632227420806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20770, + "weighted_flow_matching_loss": 0.01323337322100997, + "weighted_rkd_loss": 0.009466777229681611 + }, + { + "epoch": 6.045970322956066, + "flow_matching_loss": 0.01348843714222312, + "grad_norm": 0.16624954342842102, + "learning_rate": 2.3356567343221676e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005351739842444659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024435534840449693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013147033285349607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7322314083576202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20780, + "weighted_flow_matching_loss": 0.01348843714222312, + "weighted_rkd_loss": 0.009626546874642372 + }, + { + "epoch": 6.04887983706721, + "flow_matching_loss": 0.012104096449911594, + "grad_norm": 0.19730927050113678, + "learning_rate": 2.3349700517058548e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044972655363380905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019212225335650146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010915753711014987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.731999522447586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20790, + "weighted_flow_matching_loss": 0.012104096449911594, + "weighted_rkd_loss": 0.007990265358239412 + }, + { + "epoch": 6.051789351178353, + "flow_matching_loss": 0.01301945261657238, + "grad_norm": 0.16748932003974915, + "learning_rate": 2.3342831154476866e-05, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0049377442337572575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002133000106550753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012008488457649947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7317675709724426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20800, + "weighted_flow_matching_loss": 0.01301945261657238, + "weighted_rkd_loss": 0.008787333918735384 + }, + { + "epoch": 6.054698865289497, + "flow_matching_loss": 0.012249605171382428, + "grad_norm": 0.15256136655807495, + "learning_rate": 2.3335959257563362e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004326328355818987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00197644418803975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010629100911319255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7315355598926544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20810, + "weighted_flow_matching_loss": 0.012249605171382428, + "weighted_rkd_loss": 0.0077755628153681755 + }, + { + "epoch": 6.05760837940064, + "flow_matching_loss": 0.013337655272334815, + "grad_norm": 0.14391744136810303, + "learning_rate": 2.332908482840553e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004281415184959769, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017864395631477237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010349270049482584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.731303483247757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20820, + "weighted_flow_matching_loss": 0.013337655272334815, + "weighted_rkd_loss": 0.0075684547889977695 + }, + { + "epoch": 6.060517893511784, + "flow_matching_loss": 0.012448331248015166, + "grad_norm": 0.1505843847990036, + "learning_rate": 2.3322207869091648e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005299153318628669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002360258647240698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012958565074950457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7310713529586792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20830, + "weighted_flow_matching_loss": 0.012448331248015166, + "weighted_rkd_loss": 0.009473674092441797 + }, + { + "epoch": 6.063427407622927, + "flow_matching_loss": 0.012808089051395655, + "grad_norm": 0.12661369144916534, + "learning_rate": 2.3315328381710743e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004456179868429899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002017046033870429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010929405875504016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7308391451835632, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20840, + "weighted_flow_matching_loss": 0.012808089051395655, + "weighted_rkd_loss": 0.007987600611522793 + }, + { + "epoch": 6.06633692173407, + "flow_matching_loss": 0.013520421460270882, + "grad_norm": 0.16290825605392456, + "learning_rate": 2.3308446368352634e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004845139291137457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002110696828458458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011800975259393453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.730606883764267, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20850, + "weighted_flow_matching_loss": 0.013520421460270882, + "weighted_rkd_loss": 0.008621858432888985 + }, + { + "epoch": 6.069246435845214, + "flow_matching_loss": 0.014064181223511696, + "grad_norm": 0.1277722865343094, + "learning_rate": 2.3301561831107888e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005304240342229605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024445544579066334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01305303517729044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.730374550819397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20860, + "weighted_flow_matching_loss": 0.014064181223511696, + "weighted_rkd_loss": 0.009533598972484469 + }, + { + "epoch": 6.0721559499563575, + "flow_matching_loss": 0.012955545354634524, + "grad_norm": 0.16135196387767792, + "learning_rate": 2.3294674772067853e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004733896441757679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002098830125760287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01156662292778492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7301421701908112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20870, + "weighted_flow_matching_loss": 0.012955545354634524, + "weighted_rkd_loss": 0.008445223839953542 + }, + { + "epoch": 6.0750654640675, + "flow_matching_loss": 0.011637436598539353, + "grad_norm": 0.15820933878421783, + "learning_rate": 2.3287785193324628e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005133003927767277, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002274254139047116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012540261913090944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7299097180366516, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20880, + "weighted_flow_matching_loss": 0.011637436598539353, + "weighted_rkd_loss": 0.009153300896286964 + }, + { + "epoch": 6.077974978178644, + "flow_matching_loss": 0.01450389577075839, + "grad_norm": 0.17580394446849823, + "learning_rate": 2.3280893096971087e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004914053063839674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022375581320375204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012065664306282997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7296771943569184, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20890, + "weighted_flow_matching_loss": 0.01450389577075839, + "weighted_rkd_loss": 0.008803985919803381 + }, + { + "epoch": 6.080884492289788, + "flow_matching_loss": 0.012980814278125762, + "grad_norm": 0.17850062251091003, + "learning_rate": 2.3273998485100867e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005131127731874585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002269301901105791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012531557399779559, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7294446229934692, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20900, + "weighted_flow_matching_loss": 0.012980814278125762, + "weighted_rkd_loss": 0.00914100268855691 + }, + { + "epoch": 6.083794006400931, + "flow_matching_loss": 0.013779950514435767, + "grad_norm": 0.1602669358253479, + "learning_rate": 2.326710135980837e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004657924291677773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020629177102819084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01137876622378826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.729211974143982, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20910, + "weighted_flow_matching_loss": 0.013779950514435767, + "weighted_rkd_loss": 0.008297544857487082 + }, + { + "epoch": 6.086703520512074, + "flow_matching_loss": 0.014443797525018454, + "grad_norm": 0.1420721858739853, + "learning_rate": 2.326020172318876e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005030787386931479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022412503021769226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012302825134247542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7289792597293854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20920, + "weighted_flow_matching_loss": 0.014443797525018454, + "weighted_rkd_loss": 0.008968550944700837 + }, + { + "epoch": 6.089613034623218, + "flow_matching_loss": 0.011609302833676338, + "grad_norm": 0.15385308861732483, + "learning_rate": 2.3253299577337966e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004613609658554197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020826637279242276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011309883184731006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.728746497631073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20930, + "weighted_flow_matching_loss": 0.011609302833676338, + "weighted_rkd_loss": 0.008242001570761204 + }, + { + "epoch": 6.092522548734362, + "flow_matching_loss": 0.01410785224288702, + "grad_norm": 0.13606616854667664, + "learning_rate": 2.324639492435267e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054700958076864484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002395216759759933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013335408177226782, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7285136759281159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20940, + "weighted_flow_matching_loss": 0.01410785224288702, + "weighted_rkd_loss": 0.009715058607980609 + }, + { + "epoch": 6.0954320628455045, + "flow_matching_loss": 0.014496305491775274, + "grad_norm": 0.15428052842617035, + "learning_rate": 2.3239487766330336e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0050675167702138426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002097154618240893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012232188135385513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7282807767391205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20950, + "weighted_flow_matching_loss": 0.014496305491775274, + "weighted_rkd_loss": 0.008908447762951254 + }, + { + "epoch": 6.098341576956648, + "flow_matching_loss": 0.012768293078988791, + "grad_norm": 0.14647863805294037, + "learning_rate": 2.323257810536917e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005160749610513449, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002301257033832371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012622756138443946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7280478119850159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20960, + "weighted_flow_matching_loss": 0.012768293078988791, + "weighted_rkd_loss": 0.009190011536702514 + }, + { + "epoch": 6.101251091067792, + "flow_matching_loss": 0.012098558712750673, + "grad_norm": 0.16006386280059814, + "learning_rate": 2.3225665943568138e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004669136088341475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019217700697481631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01126004233956337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7278148233890533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20970, + "weighted_flow_matching_loss": 0.012098558712750673, + "weighted_rkd_loss": 0.008195202611386777 + }, + { + "epoch": 6.104160605178935, + "flow_matching_loss": 0.013432728499174118, + "grad_norm": 0.15857355296611786, + "learning_rate": 2.3218751283026984e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005622009187936783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025653911172412334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013809409458190203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7275817215442657, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20980, + "weighted_flow_matching_loss": 0.013432728499174118, + "weighted_rkd_loss": 0.010047489684075117 + }, + { + "epoch": 6.1070701192900785, + "flow_matching_loss": 0.012542765028774739, + "grad_norm": 0.14647118747234344, + "learning_rate": 2.321183412584619e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005142608401365578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021930497256107627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012478266470134258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7273485779762268, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20990, + "weighted_flow_matching_loss": 0.012542765028774739, + "weighted_rkd_loss": 0.009076058398932219 + }, + { + "epoch": 6.109979633401222, + "flow_matching_loss": 0.014855332672595978, + "grad_norm": 0.15734684467315674, + "learning_rate": 2.320491447412701e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004881488136015832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002353007311467081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012115983571857215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7271153688430786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21000, + "weighted_flow_matching_loss": 0.014855332672595978, + "weighted_rkd_loss": 0.00880978354252875 + }, + { + "epoch": 6.112889147512365, + "flow_matching_loss": 0.014034582022577523, + "grad_norm": 0.15013419091701508, + "learning_rate": 2.3197992329971455e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00520593801047653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002101846050936729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012513722106814385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7268821060657501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21010, + "weighted_flow_matching_loss": 0.014034582022577523, + "weighted_rkd_loss": 0.009096033545210958 + }, + { + "epoch": 6.115798661623509, + "flow_matching_loss": 0.011589732393622399, + "grad_norm": 0.13955754041671753, + "learning_rate": 2.3191067695482285e-05, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053661963902413845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023293884936720134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013061781134456395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7266487777233124, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21020, + "weighted_flow_matching_loss": 0.011589732393622399, + "weighted_rkd_loss": 0.009491282701492309 + }, + { + "epoch": 6.118708175734652, + "flow_matching_loss": 0.013424677401781082, + "grad_norm": 0.16025210916996002, + "learning_rate": 2.3184140572763018e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005434042285196483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024096092558465896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01327769411727786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7264153778553009, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21030, + "weighted_flow_matching_loss": 0.013424677401781082, + "weighted_rkd_loss": 0.009645139751955866 + }, + { + "epoch": 6.121617689845796, + "flow_matching_loss": 0.012700676266103984, + "grad_norm": 0.14953863620758057, + "learning_rate": 2.3177210963917937e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005536578036844731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002476676064543426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013549832347780465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7261819303035736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21040, + "weighted_flow_matching_loss": 0.012700676266103984, + "weighted_rkd_loss": 0.009839660953730345 + }, + { + "epoch": 6.124527203956939, + "flow_matching_loss": 0.014561595674604178, + "grad_norm": 0.15688292682170868, + "learning_rate": 2.317027887105207e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005036341119557619, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021533698891289532, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012226052023470402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7259484112262726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21050, + "weighted_flow_matching_loss": 0.014561595674604178, + "weighted_rkd_loss": 0.008875507907941938 + }, + { + "epoch": 6.127436718068083, + "flow_matching_loss": 0.013514979928731918, + "grad_norm": 0.14404214918613434, + "learning_rate": 2.3163344296271206e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004216169891878963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018836082890629768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01031594816595316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7257148385047912, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21060, + "weighted_flow_matching_loss": 0.013514979928731918, + "weighted_rkd_loss": 0.007486367924138903 + }, + { + "epoch": 6.130346232179226, + "flow_matching_loss": 0.01296049477532506, + "grad_norm": 0.1275617927312851, + "learning_rate": 2.3156407241681888e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004537919373251498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019407494459301234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011016588192433119, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7254812061786652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21070, + "weighted_flow_matching_loss": 0.01296049477532506, + "weighted_rkd_loss": 0.007992385467514396 + }, + { + "epoch": 6.133255746290369, + "flow_matching_loss": 0.012780133448541165, + "grad_norm": 0.13006635010242462, + "learning_rate": 2.3149467709391407e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005203784187324345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022654974018223584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012673065811395646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7252474844455719, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21080, + "weighted_flow_matching_loss": 0.012780133448541165, + "weighted_rkd_loss": 0.009191118879243731 + }, + { + "epoch": 6.136165260401513, + "flow_matching_loss": 0.012108213920146226, + "grad_norm": 0.1346357762813568, + "learning_rate": 2.3142525701507807e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004661087901331485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002002374187577516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011324550025165081, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7250137329101562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21090, + "weighted_flow_matching_loss": 0.012108213920146226, + "weighted_rkd_loss": 0.008210464240983128 + }, + { + "epoch": 6.139074774512657, + "flow_matching_loss": 0.014123446401208639, + "grad_norm": 0.1552387773990631, + "learning_rate": 2.3135581220139893e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005508792353793979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002422237803693861, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013439822662621736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7247798979282379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21100, + "weighted_flow_matching_loss": 0.014123446401208639, + "weighted_rkd_loss": 0.009740824345499277 + }, + { + "epoch": 6.141984288623799, + "flow_matching_loss": 0.01352758901193738, + "grad_norm": 0.16462984681129456, + "learning_rate": 2.3128634267397213e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004777994798496365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020770927192643287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011633082386106252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7245460152626038, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21110, + "weighted_flow_matching_loss": 0.01352758901193738, + "weighted_rkd_loss": 0.008428646670654415 + }, + { + "epoch": 6.144893802734943, + "flow_matching_loss": 0.013024357985705138, + "grad_norm": 0.17120693624019623, + "learning_rate": 2.3121684845390065e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004884406481869519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020142631372436883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011783076263964177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7243120610713959, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21120, + "weighted_flow_matching_loss": 0.013024357985705138, + "weighted_rkd_loss": 0.00853458298370242 + }, + { + "epoch": 6.147803316846087, + "flow_matching_loss": 0.013511977717280388, + "grad_norm": 0.13565003871917725, + "learning_rate": 2.3114732956229496e-05, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004679338238202035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002107447141315788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01146612367592752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7240780591964722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21130, + "weighted_flow_matching_loss": 0.013511977717280388, + "weighted_rkd_loss": 0.008302326733246445 + }, + { + "epoch": 6.1507128309572305, + "flow_matching_loss": 0.012347211595624685, + "grad_norm": 0.12876760959625244, + "learning_rate": 2.3107778602027318e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004810780542902649, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002071994647849351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011693555768579245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7238439857959748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21140, + "weighted_flow_matching_loss": 0.012347211595624685, + "weighted_rkd_loss": 0.00846433830447495 + }, + { + "epoch": 6.153622345068373, + "flow_matching_loss": 0.01275376882404089, + "grad_norm": 0.14709585905075073, + "learning_rate": 2.3100821784896072e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004623455880209804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001977268955670297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01122418073937297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.723609858751297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21150, + "weighted_flow_matching_loss": 0.01275376882404089, + "weighted_rkd_loss": 0.008121929410845042 + }, + { + "epoch": 6.156531859179517, + "flow_matching_loss": 0.014848017785698175, + "grad_norm": 0.13809292018413544, + "learning_rate": 2.309386250694906e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004639085638336838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020372062688693406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01131537752225995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7233756482601166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21160, + "weighted_flow_matching_loss": 0.014848017785698175, + "weighted_rkd_loss": 0.008185278857126832 + }, + { + "epoch": 6.159441373290661, + "flow_matching_loss": 0.01192432465031743, + "grad_norm": 0.12034593522548676, + "learning_rate": 2.3086900770300325e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045883134240284566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020557102863676844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011232337076216936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7231414020061493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21170, + "weighted_flow_matching_loss": 0.01192432465031743, + "weighted_rkd_loss": 0.008122484060004353 + }, + { + "epoch": 6.162350887401804, + "flow_matching_loss": 0.014113925863057375, + "grad_norm": 0.1362762749195099, + "learning_rate": 2.3079936577064656e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005157660832628608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002438490581698716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012753812316805124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7229070842266083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21180, + "weighted_flow_matching_loss": 0.014113925863057375, + "weighted_rkd_loss": 0.009219871507957578 + }, + { + "epoch": 6.165260401512947, + "flow_matching_loss": 0.01333373412489891, + "grad_norm": 0.1652764081954956, + "learning_rate": 2.3072969929357593e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046585410134866835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002133415499702096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011450497526675462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7226727128028869, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21190, + "weighted_flow_matching_loss": 0.01333373412489891, + "weighted_rkd_loss": 0.00827492242679 + }, + { + "epoch": 6.168169915624091, + "flow_matching_loss": 0.014035399537533522, + "grad_norm": 0.14603747427463531, + "learning_rate": 2.3066000829295422e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00484029627405107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021582977729849516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011838890239596367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7224382698535919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21200, + "weighted_flow_matching_loss": 0.014035399537533522, + "weighted_rkd_loss": 0.008552973018959164 + }, + { + "epoch": 6.171079429735234, + "flow_matching_loss": 0.014164528902620077, + "grad_norm": 0.15178750455379486, + "learning_rate": 2.3059029278995167e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004863576265051961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021443000878207386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011871452536433935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7222037613391876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21210, + "weighted_flow_matching_loss": 0.014164528902620077, + "weighted_rkd_loss": 0.008573615271598101 + }, + { + "epoch": 6.1739889438463775, + "flow_matching_loss": 0.01355907814577222, + "grad_norm": 0.1409389227628708, + "learning_rate": 2.3052055280574605e-05, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004349471419118345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018910481128841638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010589990764856339, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7219692051410675, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21220, + "weighted_flow_matching_loss": 0.01355907814577222, + "weighted_rkd_loss": 0.007645648904144764 + }, + { + "epoch": 6.176898457957521, + "flow_matching_loss": 0.01366234254091978, + "grad_norm": 0.12772490084171295, + "learning_rate": 2.3045078836152254e-05, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004885892290621996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021154530928470193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011887237802147866, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7217345893383026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21230, + "weighted_flow_matching_loss": 0.01366234254091978, + "weighted_rkd_loss": 0.008579442463815212 + }, + { + "epoch": 6.179807972068665, + "flow_matching_loss": 0.014240072574466467, + "grad_norm": 0.17356251180171967, + "learning_rate": 2.3038099947847367e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0050111219985410575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022500304388813675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012272274494171143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.721499902009964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21240, + "weighted_flow_matching_loss": 0.014240072574466467, + "weighted_rkd_loss": 0.008854449773207308 + }, + { + "epoch": 6.182717486179808, + "flow_matching_loss": 0.014342114329338074, + "grad_norm": 0.15131227672100067, + "learning_rate": 2.3031118617779946e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004812322044745088, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002055654744617641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011680298857390881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7212651669979095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21250, + "weighted_flow_matching_loss": 0.014342114329338074, + "weighted_rkd_loss": 0.008424572553485632 + }, + { + "epoch": 6.1856270002909515, + "flow_matching_loss": 0.013376989867538214, + "grad_norm": 0.1745452582836151, + "learning_rate": 2.3024134848070744e-05, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005084084696136415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002349369612056762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012517539132386447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7210303544998169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21260, + "weighted_flow_matching_loss": 0.013376989867538214, + "weighted_rkd_loss": 0.009025416616350413 + }, + { + "epoch": 6.188536514402095, + "flow_matching_loss": 0.013677745219320058, + "grad_norm": 0.16609878838062286, + "learning_rate": 2.301714864084124e-05, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004555336153134704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020341101451776923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011144782276824116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7207954943180084, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21270, + "weighted_flow_matching_loss": 0.013677745219320058, + "weighted_rkd_loss": 0.008033056510612369 + }, + { + "epoch": 6.191446028513238, + "flow_matching_loss": 0.012195553351193666, + "grad_norm": 0.11177773773670197, + "learning_rate": 2.3010159998213654e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004235676303505897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018902363372035325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010361588979139923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7205605685710907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21280, + "weighted_flow_matching_loss": 0.012195553351193666, + "weighted_rkd_loss": 0.007466222299262882 + }, + { + "epoch": 6.194355542624382, + "flow_matching_loss": 0.01355165969580412, + "grad_norm": 0.14291203022003174, + "learning_rate": 2.300316892231096e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004436898557469249, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001885663333814591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01075946046039462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7203255832195282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21290, + "weighted_flow_matching_loss": 0.01355165969580412, + "weighted_rkd_loss": 0.007750348560512066 + }, + { + "epoch": 6.197265056735525, + "flow_matching_loss": 0.014186262898147107, + "grad_norm": 0.13632076978683472, + "learning_rate": 2.299617541525686e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005419614072889089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002667559159453958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013506787363439798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7200905382633209, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21300, + "weighted_flow_matching_loss": 0.014186262898147107, + "weighted_rkd_loss": 0.009726246865466238 + }, + { + "epoch": 6.200174570846668, + "flow_matching_loss": 0.014082126319408417, + "grad_norm": 0.13097533583641052, + "learning_rate": 2.2989179479175794e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004824608098715544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021224597119726242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011771675944328309, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7198554277420044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21310, + "weighted_flow_matching_loss": 0.014082126319408417, + "weighted_rkd_loss": 0.008473997563123703 + }, + { + "epoch": 6.203084084957812, + "flow_matching_loss": 0.013044322654604912, + "grad_norm": 0.1294649988412857, + "learning_rate": 2.2982181116192944e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004731045523658395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002150323987007141, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01161241508089006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.719620269536972, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21320, + "weighted_flow_matching_loss": 0.013044322654604912, + "weighted_rkd_loss": 0.008356556994840503 + }, + { + "epoch": 6.205993599068956, + "flow_matching_loss": 0.011691128090023994, + "grad_norm": 0.15328744053840637, + "learning_rate": 2.297518032843423e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004726402182132006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002053110615815967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011505914945155382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7193850398063659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21330, + "weighted_flow_matching_loss": 0.011691128090023994, + "weighted_rkd_loss": 0.008277099765837192 + }, + { + "epoch": 6.208903113180099, + "flow_matching_loss": 0.013750800397247076, + "grad_norm": 0.15202927589416504, + "learning_rate": 2.2968177118026305e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00599298900924623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027149174828082325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014700895361602306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7191497504711151, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21340, + "weighted_flow_matching_loss": 0.013750800397247076, + "weighted_rkd_loss": 0.010572259407490492 + }, + { + "epoch": 6.211812627291242, + "flow_matching_loss": 0.012048145104199649, + "grad_norm": 0.1245342567563057, + "learning_rate": 2.296117148709656e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046825404278934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002117519418243319, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011482600308954715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7189144015312194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21350, + "weighted_flow_matching_loss": 0.012048145104199649, + "weighted_rkd_loss": 0.008255024161189794 + }, + { + "epoch": 6.214722141402386, + "flow_matching_loss": 0.013839787896722555, + "grad_norm": 0.1811354011297226, + "learning_rate": 2.295416343777312e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046584720490500334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021367268403992058, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011453670915216208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7186789810657501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21360, + "weighted_flow_matching_loss": 0.013839787896722555, + "weighted_rkd_loss": 0.008231505658477545 + }, + { + "epoch": 6.21763165551353, + "flow_matching_loss": 0.012312363367527723, + "grad_norm": 0.168420672416687, + "learning_rate": 2.2947152972184838e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004914149735122919, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020647915778681638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011893091211095452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7184435307979584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21370, + "weighted_flow_matching_loss": 0.012312363367527723, + "weighted_rkd_loss": 0.008544458635151386 + }, + { + "epoch": 6.220541169624672, + "flow_matching_loss": 0.012848368566483259, + "grad_norm": 0.13568663597106934, + "learning_rate": 2.294014009246132e-05, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005317526846192777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002426967793144286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013062021508812905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7182080030441285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21380, + "weighted_flow_matching_loss": 0.012848368566483259, + "weighted_rkd_loss": 0.009381402283906937 + }, + { + "epoch": 6.223450683735816, + "flow_matching_loss": 0.01460168305784464, + "grad_norm": 0.15806785225868225, + "learning_rate": 2.293312480073289e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005048006982542574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002162280527409166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012258294690400361, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7179724156856537, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21390, + "weighted_flow_matching_loss": 0.01460168305784464, + "weighted_rkd_loss": 0.008801116514950991 + }, + { + "epoch": 6.22636019784696, + "flow_matching_loss": 0.011348514817655087, + "grad_norm": 0.13873951137065887, + "learning_rate": 2.2926107099130607e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004876208864152432, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020259440527297555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011778361722826957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7177367687225342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21400, + "weighted_flow_matching_loss": 0.011348514817655087, + "weighted_rkd_loss": 0.00845374520868063 + }, + { + "epoch": 6.229269711958103, + "flow_matching_loss": 0.01406814269721508, + "grad_norm": 0.14313966035842896, + "learning_rate": 2.2919086989786267e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005130682280287147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022233068011701107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012484671361744405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7175010740756989, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21410, + "weighted_flow_matching_loss": 0.01406814269721508, + "weighted_rkd_loss": 0.00895789209753275 + }, + { + "epoch": 6.232179226069246, + "flow_matching_loss": 0.01352369338274002, + "grad_norm": 0.12100935727357864, + "learning_rate": 2.2912064474832384e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004770512669347226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020733963465318084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011614421661943198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7172652959823609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21420, + "weighted_flow_matching_loss": 0.01352369338274002, + "weighted_rkd_loss": 0.008330672513693571 + }, + { + "epoch": 6.23508874018039, + "flow_matching_loss": 0.01348575372248888, + "grad_norm": 0.13804686069488525, + "learning_rate": 2.2905039556402223e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005096271657384932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002153994864784181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012346538063138724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7170294761657715, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21430, + "weighted_flow_matching_loss": 0.01348575372248888, + "weighted_rkd_loss": 0.008852846827358008 + }, + { + "epoch": 6.237998254291533, + "flow_matching_loss": 0.011642898339778185, + "grad_norm": 0.14576874673366547, + "learning_rate": 2.2898012236629766e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004488478531129658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001880833541508764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010857790615409613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7167935907840729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21440, + "weighted_flow_matching_loss": 0.011642898339778185, + "weighted_rkd_loss": 0.00778282000683248 + }, + { + "epoch": 6.240907768402677, + "flow_matching_loss": 0.013332037348300218, + "grad_norm": 0.16353248059749603, + "learning_rate": 2.2890982517649728e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048374386969953775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022926385747268794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011967515852302314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7165576457977295, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21450, + "weighted_flow_matching_loss": 0.013332037348300218, + "weighted_rkd_loss": 0.008575378730893136 + }, + { + "epoch": 6.24381728251382, + "flow_matching_loss": 0.010899073258042336, + "grad_norm": 0.12477356940507889, + "learning_rate": 2.2883950401597548e-05, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004589946055784821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002195615426171571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01137550761923194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7163216471672058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21460, + "weighted_flow_matching_loss": 0.010899073258042336, + "weighted_rkd_loss": 0.008148469123989343 + }, + { + "epoch": 6.246726796624964, + "flow_matching_loss": 0.012804285902529956, + "grad_norm": 0.1410832554101944, + "learning_rate": 2.2876915890609396e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00480231773108244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021966859698295594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011801321525126696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7160855829715729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21470, + "weighted_flow_matching_loss": 0.012804285902529956, + "weighted_rkd_loss": 0.008450759761035442 + }, + { + "epoch": 6.249636310736107, + "flow_matching_loss": 0.014216919802129269, + "grad_norm": 0.14047692716121674, + "learning_rate": 2.2869878986822182e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004806536831893027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002220326301176101, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01183340004645288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7158494532108307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21480, + "weighted_flow_matching_loss": 0.014216919802129269, + "weighted_rkd_loss": 0.008470876608043909 + }, + { + "epoch": 6.2525458248472505, + "flow_matching_loss": 0.014598291739821434, + "grad_norm": 0.20600925385951996, + "learning_rate": 2.286283969237352e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005442340509034693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025858355103991924, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013470516446977854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7156132757663727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21490, + "weighted_flow_matching_loss": 0.014598291739821434, + "weighted_rkd_loss": 0.009639425529167056 + }, + { + "epoch": 6.255455338958394, + "flow_matching_loss": 0.013321963232010603, + "grad_norm": 0.13167937099933624, + "learning_rate": 2.2855798009401768e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005314226052723825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002411962777841836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01304041501134634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7153770327568054, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21500, + "weighted_flow_matching_loss": 0.013321963232010603, + "weighted_rkd_loss": 0.009328771336004137 + }, + { + "epoch": 6.258364853069537, + "flow_matching_loss": 0.014369025453925132, + "grad_norm": 0.1635400801897049, + "learning_rate": 2.2848753940045996e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004961881530471146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002103807928506285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012027570884674788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7151407301425934, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21510, + "weighted_flow_matching_loss": 0.014369025453925132, + "weighted_rkd_loss": 0.008601368078961969 + }, + { + "epoch": 6.261274367180681, + "flow_matching_loss": 0.012385872565209866, + "grad_norm": 0.15501831471920013, + "learning_rate": 2.284170748644602e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004579702136106789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019036569050513209, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011063061375170945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7149043798446655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21520, + "weighted_flow_matching_loss": 0.012385872565209866, + "weighted_rkd_loss": 0.007909058639779688 + }, + { + "epoch": 6.2641838812918245, + "flow_matching_loss": 0.012548449076712131, + "grad_norm": 0.14142407476902008, + "learning_rate": 2.283465865074235e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005055779265239835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023983416380360723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012509900331497192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7146679580211639, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21530, + "weighted_flow_matching_loss": 0.012548449076712131, + "weighted_rkd_loss": 0.008940430684015154 + }, + { + "epoch": 6.267093395402968, + "flow_matching_loss": 0.013468117732554675, + "grad_norm": 0.15618719160556793, + "learning_rate": 2.2827607435076245e-05, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051537174731493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002259446703828871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01256688153371215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7144314885139466, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21540, + "weighted_flow_matching_loss": 0.013468117732554675, + "weighted_rkd_loss": 0.008978094300255179 + }, + { + "epoch": 6.270002909514111, + "flow_matching_loss": 0.013350323401391506, + "grad_norm": 0.2089511603116989, + "learning_rate": 2.282055384158968e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004664711840450764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020296664326451717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011359089985489845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7141949534416199, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21550, + "weighted_flow_matching_loss": 0.013350323401391506, + "weighted_rkd_loss": 0.008112550945952535 + }, + { + "epoch": 6.272912423625255, + "flow_matching_loss": 0.014192756358534097, + "grad_norm": 0.14830756187438965, + "learning_rate": 2.2813497872425348e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004714843328110873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019007736700586975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011330460291355848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.713958352804184, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21560, + "weighted_flow_matching_loss": 0.014192756358534097, + "weighted_rkd_loss": 0.008089476870372892 + }, + { + "epoch": 6.275821937736398, + "flow_matching_loss": 0.013509661238640547, + "grad_norm": 0.1544671505689621, + "learning_rate": 2.2806439529726662e-05, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044276469852775335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019741375814191996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010829431656748056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7137217044830322, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21570, + "weighted_flow_matching_loss": 0.013509661238640547, + "weighted_rkd_loss": 0.0077291487250477076 + }, + { + "epoch": 6.278731451847541, + "flow_matching_loss": 0.014183923322707414, + "grad_norm": 0.17837689816951752, + "learning_rate": 2.2799378815637766e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00504669442307204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002201342221815139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012294731102883816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7134849905967713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21580, + "weighted_flow_matching_loss": 0.014183923322707414, + "weighted_rkd_loss": 0.008772027865052223 + }, + { + "epoch": 6.281640965958685, + "flow_matching_loss": 0.013651644065976142, + "grad_norm": 0.13653235137462616, + "learning_rate": 2.2792315732303517e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004859339678660035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021661909064278006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011884870287030936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.71324822306633, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21590, + "weighted_flow_matching_loss": 0.013651644065976142, + "weighted_rkd_loss": 0.008476918656378984 + }, + { + "epoch": 6.284550480069829, + "flow_matching_loss": 0.012652029190212489, + "grad_norm": 0.13750585913658142, + "learning_rate": 2.2785250281869492e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004384565050713718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001824414066504687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010593544226139784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7130113959312439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21600, + "weighted_flow_matching_loss": 0.012652029190212489, + "weighted_rkd_loss": 0.007553308131173253 + }, + { + "epoch": 6.2874599941809715, + "flow_matching_loss": 0.014697816409170627, + "grad_norm": 0.1605718582868576, + "learning_rate": 2.277818246648199e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004928260110318661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00231398205505684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01217050226405263, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7127745091915131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21610, + "weighted_flow_matching_loss": 0.014697816409170627, + "weighted_rkd_loss": 0.008674792200326919 + }, + { + "epoch": 6.290369508292115, + "flow_matching_loss": 0.013088114093989134, + "grad_norm": 0.14471477270126343, + "learning_rate": 2.2771112288288025e-05, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004724634974263608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019313485361635685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011380618531256914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7125375628471374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21620, + "weighted_flow_matching_loss": 0.013088114093989134, + "weighted_rkd_loss": 0.00810916069895029 + }, + { + "epoch": 6.293279022403259, + "flow_matching_loss": 0.013730761222541332, + "grad_norm": 0.1433587223291397, + "learning_rate": 2.276403974943533e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00467523701954633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020982481772080066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011448722146451474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.712300568819046, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21630, + "weighted_flow_matching_loss": 0.013730761222541332, + "weighted_rkd_loss": 0.00815493860282004 + }, + { + "epoch": 6.296188536514402, + "flow_matching_loss": 0.013570602145045995, + "grad_norm": 0.16453158855438232, + "learning_rate": 2.2756964852072357e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005311213992536068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023803856573067607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0130028136074543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7120635032653808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21640, + "weighted_flow_matching_loss": 0.013570602145045995, + "weighted_rkd_loss": 0.009258811501786113 + }, + { + "epoch": 6.299098050625545, + "flow_matching_loss": 0.012826602067798376, + "grad_norm": 0.13442394137382507, + "learning_rate": 2.2749887598348273e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045468100579455495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001889007759746164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010982627794146538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7118263721466065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21650, + "weighted_flow_matching_loss": 0.012826602067798376, + "weighted_rkd_loss": 0.007817710936069488 + }, + { + "epoch": 6.302007564736689, + "flow_matching_loss": 0.013891242444515228, + "grad_norm": 0.13242238759994507, + "learning_rate": 2.2742807990412963e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004414018825627863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018720420077443123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010700079705566167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7115892171859741, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21660, + "weighted_flow_matching_loss": 0.013891242444515228, + "weighted_rkd_loss": 0.007614090107381344 + }, + { + "epoch": 6.304917078847833, + "flow_matching_loss": 0.012014870159327984, + "grad_norm": 0.12714307010173798, + "learning_rate": 2.2735726030417022e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004545438010245562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001975932123605162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011066808085888623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7113519728183746, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21670, + "weighted_flow_matching_loss": 0.012014870159327984, + "weighted_rkd_loss": 0.007872375706210733 + }, + { + "epoch": 6.307826592958976, + "flow_matching_loss": 0.01197688803076744, + "grad_norm": 0.13476289808750153, + "learning_rate": 2.272864172051176e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004545296914875508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020223911618813873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011112984921783208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7111146748065948, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21680, + "weighted_flow_matching_loss": 0.01197688803076744, + "weighted_rkd_loss": 0.007902622641995549 + }, + { + "epoch": 6.310736107070119, + "flow_matching_loss": 0.013041857909411193, + "grad_norm": 0.12300495058298111, + "learning_rate": 2.272155506284921e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004473040741868317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019064644700847566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01085254605859518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7108773231506348, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21690, + "weighted_flow_matching_loss": 0.013041857909411193, + "weighted_rkd_loss": 0.0077148976270109415 + }, + { + "epoch": 6.313645621181263, + "flow_matching_loss": 0.011937447171658278, + "grad_norm": 0.11775640398263931, + "learning_rate": 2.2714466059582105e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004639509646221995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018976356019265949, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011176654789596795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7106399178504944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21700, + "weighted_flow_matching_loss": 0.011937447171658278, + "weighted_rkd_loss": 0.007942567532882094 + }, + { + "epoch": 6.316555135292406, + "flow_matching_loss": 0.012940176483243703, + "grad_norm": 0.13615253567695618, + "learning_rate": 2.2707374712863902e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004457403672859073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019189589889720081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01083376631140709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7104024469852448, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21710, + "weighted_flow_matching_loss": 0.012940176483243703, + "weighted_rkd_loss": 0.007696409337222576 + }, + { + "epoch": 6.31946464940355, + "flow_matching_loss": 0.013223097193986177, + "grad_norm": 0.16413240134716034, + "learning_rate": 2.270028102484876e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046951626660302285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019212197978049517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011311545222997665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7101649284362793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21720, + "weighted_flow_matching_loss": 0.013223097193986177, + "weighted_rkd_loss": 0.008033063914626836 + }, + { + "epoch": 6.322374163514693, + "flow_matching_loss": 0.012455113604664803, + "grad_norm": 0.14385348558425903, + "learning_rate": 2.2693184997691558e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004701142059639096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019203912583179773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011322675459086894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7099273383617402, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21730, + "weighted_flow_matching_loss": 0.012455113604664803, + "weighted_rkd_loss": 0.008038316015154123 + }, + { + "epoch": 6.325283677625836, + "flow_matching_loss": 0.013355191797018051, + "grad_norm": 0.16352961957454681, + "learning_rate": 2.268608663354788e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004605355928651989, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022094872430898247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01142019908875227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7096897065639496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21740, + "weighted_flow_matching_loss": 0.013355191797018051, + "weighted_rkd_loss": 0.008104765554890037 + }, + { + "epoch": 6.32819319173698, + "flow_matching_loss": 0.013851789943873882, + "grad_norm": 0.15899911522865295, + "learning_rate": 2.2678985934574022e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004747394402511418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002003878331743181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011498667020350694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7094520032405853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21750, + "weighted_flow_matching_loss": 0.013851789943873882, + "weighted_rkd_loss": 0.008157833246514201 + }, + { + "epoch": 6.3311027058481235, + "flow_matching_loss": 0.0132584935054183, + "grad_norm": 0.1349824070930481, + "learning_rate": 2.267188290292699e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004862994374707341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020752000040374697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011801188811659813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7092142403125763, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21760, + "weighted_flow_matching_loss": 0.0132584935054183, + "weighted_rkd_loss": 0.008369613019749523 + }, + { + "epoch": 6.334012219959266, + "flow_matching_loss": 0.01596396891400218, + "grad_norm": 0.1507236212491989, + "learning_rate": 2.2664777540764496e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005201724846847356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021758365212008355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012579286191612481, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7089764177799225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21770, + "weighted_flow_matching_loss": 0.01596396891400218, + "weighted_rkd_loss": 0.00891845915466547 + }, + { + "epoch": 6.33692173407041, + "flow_matching_loss": 0.012371616763994097, + "grad_norm": 0.15528379380702972, + "learning_rate": 2.265766985024496e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004607786796987057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002036391536239535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011251965118572117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7087385714054107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21780, + "weighted_flow_matching_loss": 0.012371616763994097, + "weighted_rkd_loss": 0.007974655088037252 + }, + { + "epoch": 6.339831248181554, + "flow_matching_loss": 0.014438333828002215, + "grad_norm": 0.17185154557228088, + "learning_rate": 2.2650559833527516e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005142450914718211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002177093341015279, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012461995147168637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7085006475448609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21790, + "weighted_flow_matching_loss": 0.014438333828002215, + "weighted_rkd_loss": 0.008829251583665609 + }, + { + "epoch": 6.3427407622926975, + "flow_matching_loss": 0.013000142388045787, + "grad_norm": 0.17583341896533966, + "learning_rate": 2.2643447492771998e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005282697058282793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023453836445696654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01291077770292759, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7082626640796661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21800, + "weighted_flow_matching_loss": 0.013000142388045787, + "weighted_rkd_loss": 0.009144135424867273 + }, + { + "epoch": 6.34565027640384, + "flow_matching_loss": 0.012123593688011169, + "grad_norm": 0.15727025270462036, + "learning_rate": 2.2636332830138943e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004482167446985841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020372200291603805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011001554876565933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7080246210098267, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21810, + "weighted_flow_matching_loss": 0.012123593688011169, + "weighted_rkd_loss": 0.007789331302046776 + }, + { + "epoch": 6.348559790514984, + "flow_matching_loss": 0.012949393037706614, + "grad_norm": 0.14384418725967407, + "learning_rate": 2.2629215847789596e-05, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00471728655975312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020663622068241238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01150093525648117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7077865302562714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21820, + "weighted_flow_matching_loss": 0.012949393037706614, + "weighted_rkd_loss": 0.008140205265954137 + }, + { + "epoch": 6.351469304626128, + "flow_matching_loss": 0.012488646432757378, + "grad_norm": 0.11421775072813034, + "learning_rate": 2.2622096547885912e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005357596976682544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022158331819809975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01293102717027068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7075483679771424, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21830, + "weighted_flow_matching_loss": 0.012488646432757378, + "weighted_rkd_loss": 0.009149398421868682 + }, + { + "epoch": 6.3543788187372705, + "flow_matching_loss": 0.01277965009212494, + "grad_norm": 0.14224125444889069, + "learning_rate": 2.261497493259055e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004678769549354911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020978608983568845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011455400008708238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7073101580142975, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21840, + "weighted_flow_matching_loss": 0.01277965009212494, + "weighted_rkd_loss": 0.008102462720125913 + }, + { + "epoch": 6.357288332848414, + "flow_matching_loss": 0.013730409555137157, + "grad_norm": 0.1648990362882614, + "learning_rate": 2.2607851004066863e-05, + "loss": 0.022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004827477713115513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019810508121736346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011636006366461515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7070718884468079, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21850, + "weighted_flow_matching_loss": 0.013730409555137157, + "weighted_rkd_loss": 0.008227556059136987 + }, + { + "epoch": 6.360197846959558, + "flow_matching_loss": 0.012408406380563975, + "grad_norm": 0.14446638524532318, + "learning_rate": 2.2600724764478915e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004995958786457777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020099896471947433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012001907173544168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.706833565235138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21860, + "weighted_flow_matching_loss": 0.012408406380563975, + "weighted_rkd_loss": 0.008483406156301498 + }, + { + "epoch": 6.363107361070702, + "flow_matching_loss": 0.01467130808159709, + "grad_norm": 0.1214461475610733, + "learning_rate": 2.2593596215991473e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043063184944912795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018820532248355448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010494690295308829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7065951824188232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21870, + "weighted_flow_matching_loss": 0.01467130808159709, + "weighted_rkd_loss": 0.007415526593104005 + }, + { + "epoch": 6.3660168751818444, + "flow_matching_loss": 0.013671283796429633, + "grad_norm": 0.12316619604825974, + "learning_rate": 2.2586465360769996e-05, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005010590399615466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020919315284118056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012113112257793546, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7063567459583282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21880, + "weighted_flow_matching_loss": 0.013671283796429633, + "weighted_rkd_loss": 0.008556334394961595 + }, + { + "epoch": 6.368926389292988, + "flow_matching_loss": 0.015260430425405503, + "grad_norm": 0.17745287716388702, + "learning_rate": 2.257933220098065e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005516866082325577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00231515261111781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013348884880542755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.706118243932724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21890, + "weighted_flow_matching_loss": 0.015260430425405503, + "weighted_rkd_loss": 0.009425830980762839 + }, + { + "epoch": 6.371835903404132, + "flow_matching_loss": 0.012727215234190226, + "grad_norm": 0.15584580600261688, + "learning_rate": 2.2572196738790298e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005235657328739762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002147098130080849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012618412915617227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7058796942234039, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21900, + "weighted_flow_matching_loss": 0.012727215234190226, + "weighted_rkd_loss": 0.008907091710716486 + }, + { + "epoch": 6.374745417515275, + "flow_matching_loss": 0.013104389049112796, + "grad_norm": 0.2604711353778839, + "learning_rate": 2.2565058976366517e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051563467597588895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023820955189876258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012694789096713066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7056410908699036, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21910, + "weighted_flow_matching_loss": 0.013104389049112796, + "weighted_rkd_loss": 0.00895780948922038 + }, + { + "epoch": 6.377654931626418, + "flow_matching_loss": 0.014598201029002666, + "grad_norm": 0.20651239156723022, + "learning_rate": 2.2557918915877558e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005528367450460791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025259620742872357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01358269713819027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7054024279117584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21920, + "weighted_flow_matching_loss": 0.014598201029002666, + "weighted_rkd_loss": 0.009581280779093504 + }, + { + "epoch": 6.380564445737562, + "flow_matching_loss": 0.01370034129358828, + "grad_norm": 0.16376161575317383, + "learning_rate": 2.2550776559492392e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00543250385671854, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002223971369676292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013088979199528694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7051637053489686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21930, + "weighted_flow_matching_loss": 0.01370034129358828, + "weighted_rkd_loss": 0.009229812351986767 + }, + { + "epoch": 6.383473959848705, + "flow_matching_loss": 0.012752567045390606, + "grad_norm": 0.15158697962760925, + "learning_rate": 2.254363190938067e-05, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005228891409933567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023397324723191558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012797515280544758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7049249231815338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21940, + "weighted_flow_matching_loss": 0.012752567045390606, + "weighted_rkd_loss": 0.009021324664354324 + }, + { + "epoch": 6.386383473959849, + "flow_matching_loss": 0.012647434696555137, + "grad_norm": 0.15201763808727264, + "learning_rate": 2.2536484967712753e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004736934718675911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002010719082318246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011484588589519262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7046860873699188, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21950, + "weighted_flow_matching_loss": 0.012647434696555137, + "weighted_rkd_loss": 0.008092961693182587 + }, + { + "epoch": 6.389292988070992, + "flow_matching_loss": 0.013984127808362246, + "grad_norm": 0.1549261212348938, + "learning_rate": 2.252933573665969e-05, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048221189761534335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022785463370382786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011922784429043532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.704447203874588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21960, + "weighted_flow_matching_loss": 0.013984127808362246, + "weighted_rkd_loss": 0.008399010729044675 + }, + { + "epoch": 6.392202502182135, + "flow_matching_loss": 0.012573495227843522, + "grad_norm": 0.14905838668346405, + "learning_rate": 2.2522184218393234e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0049480032874271275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020929078571498396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01198891457170248, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7042082607746124, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21970, + "weighted_flow_matching_loss": 0.012573495227843522, + "weighted_rkd_loss": 0.008442654507234693 + }, + { + "epoch": 6.395112016293279, + "flow_matching_loss": 0.012529324274510145, + "grad_norm": 0.1514357328414917, + "learning_rate": 2.2515030415085822e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004763304186053574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020927775534801186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011619385983794927, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.703969269990921, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21980, + "weighted_flow_matching_loss": 0.012529324274510145, + "weighted_rkd_loss": 0.008179689757525921 + }, + { + "epoch": 6.398021530404423, + "flow_matching_loss": 0.011469016410410404, + "grad_norm": 0.15740296244621277, + "learning_rate": 2.2507874328910592e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004502918012440205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019377227523364126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01094355881214142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7037301957607269, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21990, + "weighted_flow_matching_loss": 0.011469016410410404, + "weighted_rkd_loss": 0.007701361365616321 + }, + { + "epoch": 6.400931044515566, + "flow_matching_loss": 0.013857685308903456, + "grad_norm": 0.19116060435771942, + "learning_rate": 2.250071596204138e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005317362188361585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022415020037442447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012876226380467416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.703491085767746, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22000, + "weighted_flow_matching_loss": 0.013857685308903456, + "weighted_rkd_loss": 0.009058286668732762 + }, + { + "epoch": 6.403840558626709, + "flow_matching_loss": 0.013544932566583156, + "grad_norm": 0.20950177311897278, + "learning_rate": 2.2493555316652698e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00518224542029202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002366149565204978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012730640545487403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7032519161701203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22010, + "weighted_flow_matching_loss": 0.013544932566583156, + "weighted_rkd_loss": 0.00895281545817852 + }, + { + "epoch": 6.406750072737853, + "flow_matching_loss": 0.013950288482010364, + "grad_norm": 0.1392907053232193, + "learning_rate": 2.248639239491977e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0049977706046774985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00239708834560588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012392629776149989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7030126929283143, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22020, + "weighted_flow_matching_loss": 0.013950288482010364, + "weighted_rkd_loss": 0.008712128642946482 + }, + { + "epoch": 6.4096595868489965, + "flow_matching_loss": 0.013753575505688787, + "grad_norm": 0.14459142088890076, + "learning_rate": 2.24792271990185e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004456876846961677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019402128527872264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010853966511785985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7027734220027924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22030, + "weighted_flow_matching_loss": 0.013753575505688787, + "weighted_rkd_loss": 0.007627843134105206 + }, + { + "epoch": 6.412569100960139, + "flow_matching_loss": 0.013031282927840948, + "grad_norm": 0.13251687586307526, + "learning_rate": 2.2472059731125483e-05, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005067588086239994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021459497744217514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01228112606331706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7025340735912323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22040, + "weighted_flow_matching_loss": 0.013031282927840948, + "weighted_rkd_loss": 0.00862797535955906 + }, + { + "epoch": 6.415478615071283, + "flow_matching_loss": 0.01473670769482851, + "grad_norm": 0.14600449800491333, + "learning_rate": 2.2464889993418003e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004546492407098412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020701933070085944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011163178132846951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7022946774959564, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22050, + "weighted_flow_matching_loss": 0.01473670769482851, + "weighted_rkd_loss": 0.007839838089421392 + }, + { + "epoch": 6.418388129182427, + "flow_matching_loss": 0.014062322303652763, + "grad_norm": 0.14796994626522064, + "learning_rate": 2.2457717988074047e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004829371371306479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002029931405559182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011688674055039882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7020552396774292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22060, + "weighted_flow_matching_loss": 0.014062322303652763, + "weighted_rkd_loss": 0.00820609307847917 + }, + { + "epoch": 6.4212976432935704, + "flow_matching_loss": 0.010865774983540178, + "grad_norm": 0.1548335701227188, + "learning_rate": 2.245054371727228e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004872307344339788, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002201088855508715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011945703346282244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7018157422542572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22070, + "weighted_flow_matching_loss": 0.010865774983540178, + "weighted_rkd_loss": 0.00838370001874864 + }, + { + "epoch": 6.424207157404713, + "flow_matching_loss": 0.01306023383513093, + "grad_norm": 0.13987600803375244, + "learning_rate": 2.2443367183192046e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045815275516361, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002083773701451719, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011246828641742469, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7015761733055115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22080, + "weighted_flow_matching_loss": 0.01306023383513093, + "weighted_rkd_loss": 0.007890601083636284 + }, + { + "epoch": 6.427116671515857, + "flow_matching_loss": 0.015978183876723052, + "grad_norm": 0.15324291586875916, + "learning_rate": 2.2436188388013394e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004679572465829551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002036897325888276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011396042350679637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7013365626335144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22090, + "weighted_flow_matching_loss": 0.015978183876723052, + "weighted_rkd_loss": 0.00799244549125433 + }, + { + "epoch": 6.430026185627001, + "flow_matching_loss": 0.013740816339850425, + "grad_norm": 0.15786904096603394, + "learning_rate": 2.2429007333917048e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005464787827804685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002470091066788882, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013399666734039783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.701096898317337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22100, + "weighted_flow_matching_loss": 0.013740816339850425, + "weighted_rkd_loss": 0.009394398424774408 + }, + { + "epoch": 6.4329356997381435, + "flow_matching_loss": 0.013624203391373158, + "grad_norm": 0.12533599138259888, + "learning_rate": 2.242182402308443e-05, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004411505931057036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018940872978419065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01071709911338985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7008571743965148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22110, + "weighted_flow_matching_loss": 0.013624203391373158, + "weighted_rkd_loss": 0.007511171977967024 + }, + { + "epoch": 6.435845213849287, + "flow_matching_loss": 0.013084699772298336, + "grad_norm": 0.10523157566785812, + "learning_rate": 2.2414638457697632e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005149229313246906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002258421469014138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012556879967451095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7006174027919769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22120, + "weighted_flow_matching_loss": 0.013084699772298336, + "weighted_rkd_loss": 0.00879766196012497 + }, + { + "epoch": 6.438754727960431, + "flow_matching_loss": 0.01278962749056518, + "grad_norm": 0.1484394371509552, + "learning_rate": 2.2407450639939448e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005118805542588234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002283002785407007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012520613986998797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7003775596618652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22130, + "weighted_flow_matching_loss": 0.01278962749056518, + "weighted_rkd_loss": 0.00876908521167934 + }, + { + "epoch": 6.441664242071574, + "flow_matching_loss": 0.011693281866610051, + "grad_norm": 0.13652554154396057, + "learning_rate": 2.2400260571993333e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004670107597485184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018663915572687984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011206606775522232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7001376748085022, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22140, + "weighted_flow_matching_loss": 0.011693281866610051, + "weighted_rkd_loss": 0.00784618011675775 + }, + { + "epoch": 6.444573756182717, + "flow_matching_loss": 0.013630101922899485, + "grad_norm": 0.1205289214849472, + "learning_rate": 2.2393068256043457e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00501822205260396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020209201029501855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012057364359498024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6998977363109589, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22150, + "weighted_flow_matching_loss": 0.013630101922899485, + "weighted_rkd_loss": 0.008438948914408684 + }, + { + "epoch": 6.447483270293861, + "flow_matching_loss": 0.011801364179700614, + "grad_norm": 0.14893214404582977, + "learning_rate": 2.2385873694274646e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004769932338967919, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002070927433669567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011610792204737663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6996577322483063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22160, + "weighted_flow_matching_loss": 0.011801364179700614, + "weighted_rkd_loss": 0.00812362041324377 + }, + { + "epoch": 6.450392784405004, + "flow_matching_loss": 0.012665357813239098, + "grad_norm": 0.14371460676193237, + "learning_rate": 2.2378676888872418e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046771790366619825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020978036569431423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011452161613851786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6994176805019379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22170, + "weighted_flow_matching_loss": 0.012665357813239098, + "weighted_rkd_loss": 0.008009898848831654 + }, + { + "epoch": 6.453302298516148, + "flow_matching_loss": 0.013674639724195003, + "grad_norm": 0.14659859240055084, + "learning_rate": 2.2371477842022976e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004256522562354803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001914358395151794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010427403496578336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6991775751113891, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22180, + "weighted_flow_matching_loss": 0.013674639724195003, + "weighted_rkd_loss": 0.007290483498945832 + }, + { + "epoch": 6.456211812627291, + "flow_matching_loss": 0.01295180693268776, + "grad_norm": 0.15873803198337555, + "learning_rate": 2.2364276555913195e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004469788656570018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020031589432619513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010942736314609647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6989374041557312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22190, + "weighted_flow_matching_loss": 0.01295180693268776, + "weighted_rkd_loss": 0.00764828622341156 + }, + { + "epoch": 6.459121326738435, + "flow_matching_loss": 0.013330963626503944, + "grad_norm": 0.1566515564918518, + "learning_rate": 2.235707303273064e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004820883879438043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019199038622900843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011561671737581491, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6986971795558929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22200, + "weighted_flow_matching_loss": 0.013330963626503944, + "weighted_rkd_loss": 0.008078141417354346 + }, + { + "epoch": 6.462030840849578, + "flow_matching_loss": 0.012306770216673613, + "grad_norm": 0.14371515810489655, + "learning_rate": 2.234986727466355e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005013110605068505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021768878679722546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012203109078109264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6984569251537323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22210, + "weighted_flow_matching_loss": 0.012306770216673613, + "weighted_rkd_loss": 0.008523372374475003 + }, + { + "epoch": 6.464940354960722, + "flow_matching_loss": 0.01308154994621873, + "grad_norm": 0.1352451890707016, + "learning_rate": 2.2342659283900843e-05, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045993222622200845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002237576001789421, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011436220537871123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6982165992259979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22220, + "weighted_flow_matching_loss": 0.01308154994621873, + "weighted_rkd_loss": 0.007984980242326856 + }, + { + "epoch": 6.467849869071865, + "flow_matching_loss": 0.01271579172462225, + "grad_norm": 0.12580689787864685, + "learning_rate": 2.2335449062632117e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004766094824299216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021134336828254165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011645623296499253, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6979762196540833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22230, + "weighted_flow_matching_loss": 0.01271579172462225, + "weighted_rkd_loss": 0.008128439541906118 + }, + { + "epoch": 6.470759383183008, + "flow_matching_loss": 0.01310357842594385, + "grad_norm": 0.13933445513248444, + "learning_rate": 2.232823661304765e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004495724709704519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020380850066430865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011029534414410591, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6977357864379883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22240, + "weighted_flow_matching_loss": 0.01310357842594385, + "weighted_rkd_loss": 0.0076957007870078085 + }, + { + "epoch": 6.473668897294152, + "flow_matching_loss": 0.013847390376031398, + "grad_norm": 0.14591021835803986, + "learning_rate": 2.2321021937338392e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045178344706073405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020276498747989536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011063318885862827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6974952995777131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22250, + "weighted_flow_matching_loss": 0.013847390376031398, + "weighted_rkd_loss": 0.007716700853779912 + }, + { + "epoch": 6.4765784114052956, + "flow_matching_loss": 0.013368080230429768, + "grad_norm": 0.16384637355804443, + "learning_rate": 2.2313805037695965e-05, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00494765150360763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002061387151479721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011956690065562724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.697254753112793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22260, + "weighted_flow_matching_loss": 0.013368080230429768, + "weighted_rkd_loss": 0.00833673975430429 + }, + { + "epoch": 6.479487925516438, + "flow_matching_loss": 0.010345349414274096, + "grad_norm": 0.1344226598739624, + "learning_rate": 2.230658591631268e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004424437368288636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002012332249432802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010861206939443946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6970141589641571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22270, + "weighted_flow_matching_loss": 0.010345349414274096, + "weighted_rkd_loss": 0.007570477481931448 + }, + { + "epoch": 6.482397439627582, + "flow_matching_loss": 0.012991004530340433, + "grad_norm": 0.14080040156841278, + "learning_rate": 2.2299364575381512e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004476384725421667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018547338200733065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010807503294199706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6967735111713409, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22280, + "weighted_flow_matching_loss": 0.012991004530340433, + "weighted_rkd_loss": 0.007530347397550941 + }, + { + "epoch": 6.485306953738726, + "flow_matching_loss": 0.011870288150385022, + "grad_norm": 0.1514521688222885, + "learning_rate": 2.2292141017096115e-05, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00436936232727021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019279335159808398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01066665817052126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6965328097343445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22290, + "weighted_flow_matching_loss": 0.011870288150385022, + "weighted_rkd_loss": 0.007429703930392861 + }, + { + "epoch": 6.488216467849869, + "flow_matching_loss": 0.012289571668952703, + "grad_norm": 0.1157970130443573, + "learning_rate": 2.228491524365081e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004586433782242239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018177012214437127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010990568809211254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6962920427322388, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22300, + "weighted_flow_matching_loss": 0.012289571668952703, + "weighted_rkd_loss": 0.007652688585221767 + }, + { + "epoch": 6.491125981961012, + "flow_matching_loss": 0.015703382343053816, + "grad_norm": 0.1641281247138977, + "learning_rate": 2.227768725724061e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004561269353143871, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020423259818926455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01116486471146345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6960512399673462, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22310, + "weighted_flow_matching_loss": 0.015703382343053816, + "weighted_rkd_loss": 0.007771291397511959 + }, + { + "epoch": 6.494035496072156, + "flow_matching_loss": 0.013584522064775229, + "grad_norm": 0.14894796907901764, + "learning_rate": 2.227045706006117e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004902415536344051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021100114448927344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01191484248265624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6958103597164154, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22320, + "weighted_flow_matching_loss": 0.013584522064775229, + "weighted_rkd_loss": 0.008290441753342748 + }, + { + "epoch": 6.4969450101833, + "flow_matching_loss": 0.01427977131679654, + "grad_norm": 0.145734503865242, + "learning_rate": 2.2263224654308845e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004921220894902945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020936797722242773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01193612152710557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6955694496631623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22330, + "weighted_flow_matching_loss": 0.01427977131679654, + "weighted_rkd_loss": 0.008302423777058721 + }, + { + "epoch": 6.4998545242944425, + "flow_matching_loss": 0.013045026827603579, + "grad_norm": 0.140597403049469, + "learning_rate": 2.2255990042180646e-05, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005192230199463665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002158925333060324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01254338575527072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6953284740447998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22340, + "weighted_flow_matching_loss": 0.013045026827603579, + "weighted_rkd_loss": 0.008721674466505647 + }, + { + "epoch": 6.502764038405586, + "flow_matching_loss": 0.01214315313845873, + "grad_norm": 0.12428169697523117, + "learning_rate": 2.2248753225874252e-05, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004253219556994736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017386687686666846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010245107812806964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6950874507427216, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22350, + "weighted_flow_matching_loss": 0.01214315313845873, + "weighted_rkd_loss": 0.007121250312775373 + }, + { + "epoch": 6.50567355251673, + "flow_matching_loss": 0.01191629795357585, + "grad_norm": 0.11546061187982559, + "learning_rate": 2.2241514207588018e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038679254241287707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017716438975185155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0095074946526438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6948463618755341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22360, + "weighted_flow_matching_loss": 0.01191629795357585, + "weighted_rkd_loss": 0.006606285134330392 + }, + { + "epoch": 6.508583066627873, + "flow_matching_loss": 0.012144731357693672, + "grad_norm": 0.14708925783634186, + "learning_rate": 2.223427298952097e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004700545175001025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021386057371273635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011539696250110865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6946052372455597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22370, + "weighted_flow_matching_loss": 0.012144731357693672, + "weighted_rkd_loss": 0.008015442453324795 + }, + { + "epoch": 6.5114925807390165, + "flow_matching_loss": 0.012937870062887668, + "grad_norm": 0.14049164950847626, + "learning_rate": 2.2227029573872803e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004185813129879534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017907543340697884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010162380570545793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.694364047050476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22380, + "weighted_flow_matching_loss": 0.012937870062887668, + "weighted_rkd_loss": 0.0070564400404691694 + }, + { + "epoch": 6.51440209485016, + "flow_matching_loss": 0.013277421239763498, + "grad_norm": 0.13278605043888092, + "learning_rate": 2.2219783962843866e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003877456556074321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001673039048910141, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009427952067926525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6941228091716767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22390, + "weighted_flow_matching_loss": 0.013277421239763498, + "weighted_rkd_loss": 0.006544170947745442 + }, + { + "epoch": 6.517311608961304, + "flow_matching_loss": 0.012442104984074831, + "grad_norm": 0.14407888054847717, + "learning_rate": 2.221253615863519e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044592170976102356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019412483787164092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010859682643786072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6938815176486969, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22400, + "weighted_flow_matching_loss": 0.012442104984074831, + "weighted_rkd_loss": 0.007535229763016104 + }, + { + "epoch": 6.520221123072447, + "flow_matching_loss": 0.012036646064370871, + "grad_norm": 0.1174648106098175, + "learning_rate": 2.2205286163448466e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004275393905118108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017689717933535576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010319759510457516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6936401665210724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22410, + "weighted_flow_matching_loss": 0.012036646064370871, + "weighted_rkd_loss": 0.007158194063231349 + }, + { + "epoch": 6.52313063718359, + "flow_matching_loss": 0.01359306825324893, + "grad_norm": 0.14539872109889984, + "learning_rate": 2.219803397948604e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004320457624271512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018683691276237368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010509284399449825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6933987677097321, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22420, + "weighted_flow_matching_loss": 0.01359306825324893, + "weighted_rkd_loss": 0.007287113042548299 + }, + { + "epoch": 6.526040151294734, + "flow_matching_loss": 0.015047751553356647, + "grad_norm": 0.11816532909870148, + "learning_rate": 2.2190779608950953e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0047059898730367426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022994841099716725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011711463890969753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6931573152542114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22430, + "weighted_flow_matching_loss": 0.015047751553356647, + "weighted_rkd_loss": 0.008117926213890314 + }, + { + "epoch": 6.528949665405877, + "flow_matching_loss": 0.013026916421949863, + "grad_norm": 0.13718359172344208, + "learning_rate": 2.218352305404688e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005214083706960082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023773458902724086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012805513478815556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6929158091545105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22440, + "weighted_flow_matching_loss": 0.013026916421949863, + "weighted_rkd_loss": 0.008873107517138124 + }, + { + "epoch": 6.531859179517021, + "flow_matching_loss": 0.0109355210326612, + "grad_norm": 0.14337778091430664, + "learning_rate": 2.217626431697818e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004043877427466214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017208295059390367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009808584442362189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6926742553710937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22450, + "weighted_flow_matching_loss": 0.0109355210326612, + "weighted_rkd_loss": 0.006794184725731611 + }, + { + "epoch": 6.534768693628164, + "flow_matching_loss": 0.011195660894736647, + "grad_norm": 0.13880784809589386, + "learning_rate": 2.216900339994985e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00443278958555311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018224785453639925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01068805754184723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6924326360225678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22460, + "weighted_flow_matching_loss": 0.011195660894736647, + "weighted_rkd_loss": 0.007400783896446228 + }, + { + "epoch": 6.537678207739307, + "flow_matching_loss": 0.01396958064287901, + "grad_norm": 0.13987363874912262, + "learning_rate": 2.2161740305167574e-05, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004256245750002563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018098289147019386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010322320368140936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6921909749507904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22470, + "weighted_flow_matching_loss": 0.01396958064287901, + "weighted_rkd_loss": 0.007145010028034448 + }, + { + "epoch": 6.540587721850451, + "flow_matching_loss": 0.014665552647784352, + "grad_norm": 0.13520361483097076, + "learning_rate": 2.215447503483769e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004307502997107804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017928455607034266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010407851543277502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6919492602348327, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22480, + "weighted_flow_matching_loss": 0.014665552647784352, + "weighted_rkd_loss": 0.007201695535331965 + }, + { + "epoch": 6.543497235961595, + "flow_matching_loss": 0.01391679486259818, + "grad_norm": 0.1385667771100998, + "learning_rate": 2.21472075911672e-05, + "loss": 0.0221, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004785991460084915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022749989409931006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011846981756389142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6917074799537659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22490, + "weighted_flow_matching_loss": 0.01391679486259818, + "weighted_rkd_loss": 0.008194712596014142 + }, + { + "epoch": 6.546406750072737, + "flow_matching_loss": 0.013347149454057216, + "grad_norm": 0.14992380142211914, + "learning_rate": 2.2139937976363747e-05, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004643988609313965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001949701039120555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011237678211182356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6914656698703766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22500, + "weighted_flow_matching_loss": 0.013347149454057216, + "weighted_rkd_loss": 0.007770508946850896 + }, + { + "epoch": 6.549316264183881, + "flow_matching_loss": 0.01244006659835577, + "grad_norm": 0.1471625715494156, + "learning_rate": 2.2132666192635658e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00417820627335459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018187236390076577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010175136104226112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6912237942218781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22510, + "weighted_flow_matching_loss": 0.01244006659835577, + "weighted_rkd_loss": 0.007033163262531161 + }, + { + "epoch": 6.552225778295025, + "flow_matching_loss": 0.013185142539441586, + "grad_norm": 0.13755078613758087, + "learning_rate": 2.2125392242191905e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004603917570784688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020326135912910105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01124044880270958, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6909818649291992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22520, + "weighted_flow_matching_loss": 0.013185142539441586, + "weighted_rkd_loss": 0.007766908360645175 + }, + { + "epoch": 6.5551352924061685, + "flow_matching_loss": 0.013453018944710493, + "grad_norm": 0.16168271005153656, + "learning_rate": 2.2118116127242128e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004343330697156489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018429346033371984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010529596032574773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6907398700714111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22530, + "weighted_flow_matching_loss": 0.013453018944710493, + "weighted_rkd_loss": 0.007273072982206941 + }, + { + "epoch": 6.558044806517311, + "flow_matching_loss": 0.013269550632685423, + "grad_norm": 0.15151099860668182, + "learning_rate": 2.2110837849996614e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045536930672824385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019033170072361826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011010703071951866, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6904978573322296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22540, + "weighted_flow_matching_loss": 0.013269550632685423, + "weighted_rkd_loss": 0.007602884760126471 + }, + { + "epoch": 6.560954320628455, + "flow_matching_loss": 0.012376632541418076, + "grad_norm": 0.1436288058757782, + "learning_rate": 2.210355741266631e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004825872415676713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002307289803866297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011959034577012062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6902557730674743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22550, + "weighted_flow_matching_loss": 0.012376632541418076, + "weighted_rkd_loss": 0.008254776149988175 + }, + { + "epoch": 6.563863834739599, + "flow_matching_loss": 0.010473891999572516, + "grad_norm": 0.15166012942790985, + "learning_rate": 2.2096274817462823e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004719914053566754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002003271959256381, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011443100031465292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6900136411190033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22560, + "weighted_flow_matching_loss": 0.010473891999572516, + "weighted_rkd_loss": 0.007895893277600407 + }, + { + "epoch": 6.566773348850742, + "flow_matching_loss": 0.014210864156484603, + "grad_norm": 0.1563129872083664, + "learning_rate": 2.208899006659841e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004944912414066494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022262959275394676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012116120662540197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6897714495658874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22570, + "weighted_flow_matching_loss": 0.014210864156484603, + "weighted_rkd_loss": 0.008357224753126502 + }, + { + "epoch": 6.569682862961885, + "flow_matching_loss": 0.012268634140491485, + "grad_norm": 0.1568802148103714, + "learning_rate": 2.2081703162285994e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004706736002117395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020646689576096834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011478140950202942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6895292043685913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22580, + "weighted_flow_matching_loss": 0.012268634140491485, + "weighted_rkd_loss": 0.007914468785747886 + }, + { + "epoch": 6.572592377073029, + "flow_matching_loss": 0.011978870816528797, + "grad_norm": 0.147550567984581, + "learning_rate": 2.2074414106739133e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039331805892288685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016714847995899617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009537845710292458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6892869174480438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22590, + "weighted_flow_matching_loss": 0.011978870816528797, + "weighted_rkd_loss": 0.006574239674955606 + }, + { + "epoch": 6.575501891184173, + "flow_matching_loss": 0.01440724078565836, + "grad_norm": 0.18025685846805573, + "learning_rate": 2.2067122902172062e-05, + "loss": 0.0225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004815507330931723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002051241439767182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01168225621804595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6890445709228515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22600, + "weighted_flow_matching_loss": 0.01440724078565836, + "weighted_rkd_loss": 0.008049622643738985 + }, + { + "epoch": 6.5784114052953155, + "flow_matching_loss": 0.013899084646254778, + "grad_norm": 0.14709638059139252, + "learning_rate": 2.2059829550799645e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004801457189023495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020263557671569288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01162927020341158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6888021767139435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22610, + "weighted_flow_matching_loss": 0.013899084646254778, + "weighted_rkd_loss": 0.008010203903540968 + }, + { + "epoch": 6.581320919406459, + "flow_matching_loss": 0.01389847295358777, + "grad_norm": 0.1365823596715927, + "learning_rate": 2.2052534054837417e-05, + "loss": 0.0223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004910797788761556, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023802908835932612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012201886577531695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6885597288608551, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22620, + "weighted_flow_matching_loss": 0.01389847295358777, + "weighted_rkd_loss": 0.008401886047795414 + }, + { + "epoch": 6.584230433517602, + "flow_matching_loss": 0.013079369813203812, + "grad_norm": 0.12947148084640503, + "learning_rate": 2.2045236416501552e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004565249825827777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00204549478366971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011175994528457522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6883172333240509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22630, + "weighted_flow_matching_loss": 0.013079369813203812, + "weighted_rkd_loss": 0.007692625001072884 + }, + { + "epoch": 6.587139947628746, + "flow_matching_loss": 0.012623524200171233, + "grad_norm": 0.13764137029647827, + "learning_rate": 2.2037936638008885e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044409525347873565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018789932946674526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010760898422449827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6880746781826019, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22640, + "weighted_flow_matching_loss": 0.012623524200171233, + "weighted_rkd_loss": 0.007404266949743032 + }, + { + "epoch": 6.5900494617398895, + "flow_matching_loss": 0.010803352342918515, + "grad_norm": 0.11410126090049744, + "learning_rate": 2.203063472157689e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004344725585542619, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016999130952171982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01038936423137784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6878320813179016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22650, + "weighted_flow_matching_loss": 0.010803352342918515, + "weighted_rkd_loss": 0.0071462135761976246 + }, + { + "epoch": 6.592958975851033, + "flow_matching_loss": 0.012239995691925288, + "grad_norm": 0.14902366697788239, + "learning_rate": 2.2023330669423703e-05, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004813524428755045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022975624073296786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011924611125141382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.687589418888092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22660, + "weighted_flow_matching_loss": 0.012239995691925288, + "weighted_rkd_loss": 0.008199035981670022 + }, + { + "epoch": 6.595868489962176, + "flow_matching_loss": 0.011695636063814163, + "grad_norm": 0.14992953836917877, + "learning_rate": 2.20160244837681e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004694042890332639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022430334123782814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011631118971854449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6873467206954956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22670, + "weighted_flow_matching_loss": 0.011695636063814163, + "weighted_rkd_loss": 0.007994709070771933 + }, + { + "epoch": 6.59877800407332, + "flow_matching_loss": 0.013125290814787149, + "grad_norm": 0.12627021968364716, + "learning_rate": 2.20087161668295e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004452737979590893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019279899657703935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010833465773612262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6871039628982544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22680, + "weighted_flow_matching_loss": 0.013125290814787149, + "weighted_rkd_loss": 0.007443775329738856 + }, + { + "epoch": 6.601687518184463, + "flow_matching_loss": 0.012216917425394058, + "grad_norm": 0.1368207484483719, + "learning_rate": 2.200140572082799e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0047577204648405315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001994037185795605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011509478278458118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6868611514568329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22690, + "weighted_flow_matching_loss": 0.012216917425394058, + "weighted_rkd_loss": 0.007905445387586951 + }, + { + "epoch": 6.604597032295606, + "flow_matching_loss": 0.015074983797967434, + "grad_norm": 0.143743634223938, + "learning_rate": 2.1994093147984283e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004607490310445428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018379291985183954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011052909679710865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.68661829829216, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22700, + "weighted_flow_matching_loss": 0.015074983797967434, + "weighted_rkd_loss": 0.007589113758876919 + }, + { + "epoch": 6.60750654640675, + "flow_matching_loss": 0.012793526332825423, + "grad_norm": 0.12180587649345398, + "learning_rate": 2.1986778450519742e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004910104279406369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002122081385459751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01194228995591402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6863753855228424, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22710, + "weighted_flow_matching_loss": 0.012793526332825423, + "weighted_rkd_loss": 0.008196942321956158 + }, + { + "epoch": 6.610416060517894, + "flow_matching_loss": 0.011907009780406952, + "grad_norm": 0.11620453745126724, + "learning_rate": 2.1979461630656384e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0047735445201396946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019344515749253333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011481540510430932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6861324191093445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22720, + "weighted_flow_matching_loss": 0.011907009780406952, + "weighted_rkd_loss": 0.007877914514392615 + }, + { + "epoch": 6.613325574629037, + "flow_matching_loss": 0.014020474068820477, + "grad_norm": 0.13024739921092987, + "learning_rate": 2.1972142690616866e-05, + "loss": 0.0222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0049351993016898636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021017981111072005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011972196586430072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6858894109725953, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22730, + "weighted_flow_matching_loss": 0.014020474068820477, + "weighted_rkd_loss": 0.008211615215986966 + }, + { + "epoch": 6.61623508874018, + "flow_matching_loss": 0.012560355849564076, + "grad_norm": 0.13424845039844513, + "learning_rate": 2.1964821632624484e-05, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005173976975493133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023097142460756005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012657668348401785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6856463432312012, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22740, + "weighted_flow_matching_loss": 0.012560355849564076, + "weighted_rkd_loss": 0.008678658725693822 + }, + { + "epoch": 6.619144602851324, + "flow_matching_loss": 0.01350459661334753, + "grad_norm": 0.1487029492855072, + "learning_rate": 2.1957498458903188e-05, + "loss": 0.0218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004974049632437527, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021457177237607537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012093816883862019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6854032278060913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22750, + "weighted_flow_matching_loss": 0.01350459661334753, + "weighted_rkd_loss": 0.008289148891344666 + }, + { + "epoch": 6.622054116962468, + "flow_matching_loss": 0.013412454631179571, + "grad_norm": 0.13621477782726288, + "learning_rate": 2.195017317167756e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004390327632427216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019625284359790383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010743183735758065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6851600408554077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22760, + "weighted_flow_matching_loss": 0.013412454631179571, + "weighted_rkd_loss": 0.007360816048458218 + }, + { + "epoch": 6.62496363107361, + "flow_matching_loss": 0.012515323609113694, + "grad_norm": 0.1255587935447693, + "learning_rate": 2.1942845773172836e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004853925760835409, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021020571934059263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011809908598661423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6849168479442597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22770, + "weighted_flow_matching_loss": 0.012515323609113694, + "weighted_rkd_loss": 0.00808875383809209 + }, + { + "epoch": 6.627873145184754, + "flow_matching_loss": 0.011970837134867907, + "grad_norm": 0.15015272796154022, + "learning_rate": 2.193551626561488e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00507125339936465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020335470791906117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012176054064184427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6846735835075378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22780, + "weighted_flow_matching_loss": 0.011970837134867907, + "weighted_rkd_loss": 0.008336671767756343 + }, + { + "epoch": 6.630782659295898, + "flow_matching_loss": 0.01360397906973958, + "grad_norm": 0.15563759207725525, + "learning_rate": 2.19281846512302e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0047323778737336395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020765810273587703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011541336867958307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6844302654266358, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22790, + "weighted_flow_matching_loss": 0.01360397906973958, + "weighted_rkd_loss": 0.0078991930000484 + }, + { + "epoch": 6.633692173407041, + "flow_matching_loss": 0.012911694031208754, + "grad_norm": 0.15981414914131165, + "learning_rate": 2.1920850932245955e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004413503920659423, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018050599144771695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01063206777907908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6841869056224823, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22800, + "weighted_flow_matching_loss": 0.012911694031208754, + "weighted_rkd_loss": 0.007274301163852215 + }, + { + "epoch": 6.636601687518184, + "flow_matching_loss": 0.012227182649075985, + "grad_norm": 0.14427681267261505, + "learning_rate": 2.1913515110889932e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005079310573637486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022245446220040322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01238316586241126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6839434802532196, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22810, + "weighted_flow_matching_loss": 0.012227182649075985, + "weighted_rkd_loss": 0.008469336852431297 + }, + { + "epoch": 6.639511201629328, + "flow_matching_loss": 0.01276585115119815, + "grad_norm": 0.14286963641643524, + "learning_rate": 2.1906177189390557e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046421029604971405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019751357031054795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011259341612458229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6837000131607056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22820, + "weighted_flow_matching_loss": 0.01276585115119815, + "weighted_rkd_loss": 0.007698003947734833 + }, + { + "epoch": 6.642420715740471, + "flow_matching_loss": 0.013477745186537504, + "grad_norm": 0.15149709582328796, + "learning_rate": 2.18988371699769e-05, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00491869926918298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021581549546681343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011995553690940141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6834564924240112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22830, + "weighted_flow_matching_loss": 0.013477745186537504, + "weighted_rkd_loss": 0.008198565524071454 + }, + { + "epoch": 6.645330229851615, + "flow_matching_loss": 0.01176582146435976, + "grad_norm": 0.12118668109178543, + "learning_rate": 2.189149505487866e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004834962426684797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020819760975427926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011751900799572468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6832129180431366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22840, + "weighted_flow_matching_loss": 0.01176582146435976, + "weighted_rkd_loss": 0.008029179135337471 + }, + { + "epoch": 6.648239743962758, + "flow_matching_loss": 0.013882305752485991, + "grad_norm": 0.15415611863136292, + "learning_rate": 2.1884150846326186e-05, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005056939856149256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002307261456735432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012421141285449267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6829693019390106, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22850, + "weighted_flow_matching_loss": 0.013882305752485991, + "weighted_rkd_loss": 0.008483143243938684 + }, + { + "epoch": 6.651149258073902, + "flow_matching_loss": 0.012763682845979929, + "grad_norm": 0.14182938635349274, + "learning_rate": 2.187680454655045e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005152288498356939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022256776108406486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012530254572629929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6827256321907044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22860, + "weighted_flow_matching_loss": 0.012763682845979929, + "weighted_rkd_loss": 0.008554600737988948 + }, + { + "epoch": 6.654058772185045, + "flow_matching_loss": 0.011930536385625601, + "grad_norm": 0.13326197862625122, + "learning_rate": 2.186945615778306e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004864735226146877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020464615314267575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011775931995362043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6824819147586823, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22870, + "weighted_flow_matching_loss": 0.011930536385625601, + "weighted_rkd_loss": 0.008036859985440969 + }, + { + "epoch": 6.6569682862961885, + "flow_matching_loss": 0.013320185430347919, + "grad_norm": 0.125636488199234, + "learning_rate": 2.186210568225627e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004590096999891102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020506718079559504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011230865959078074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6822381317615509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22880, + "weighted_flow_matching_loss": 0.013320185430347919, + "weighted_rkd_loss": 0.00766206681728363 + }, + { + "epoch": 6.659877800407332, + "flow_matching_loss": 0.011237704753875732, + "grad_norm": 0.1340685337781906, + "learning_rate": 2.1854753122202955e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004536569630727172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00194850517436862, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011021644389256836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6819943249225616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22890, + "weighted_flow_matching_loss": 0.011237704753875732, + "weighted_rkd_loss": 0.0075167506001889706 + }, + { + "epoch": 6.662787314518475, + "flow_matching_loss": 0.01173941306769848, + "grad_norm": 0.123332180082798, + "learning_rate": 2.1847398479856638e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004214404593221843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017426289734430612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010171438101679087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6817504525184631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22900, + "weighted_flow_matching_loss": 0.01173941306769848, + "weighted_rkd_loss": 0.0069344222079962495 + }, + { + "epoch": 6.665696828629619, + "flow_matching_loss": 0.012727292347699403, + "grad_norm": 0.1312863975763321, + "learning_rate": 2.184004175745145e-05, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004146359954029321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018244818435050546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010117201693356037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6815065383911133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22910, + "weighted_flow_matching_loss": 0.012727292347699403, + "weighted_rkd_loss": 0.006894962210208177 + }, + { + "epoch": 6.6686063427407625, + "flow_matching_loss": 0.012650790251791477, + "grad_norm": 0.13679887354373932, + "learning_rate": 2.1832682957222182e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004868071456439793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022465357440523805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01198267862200737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6812625646591186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22920, + "weighted_flow_matching_loss": 0.012650790251791477, + "weighted_rkd_loss": 0.00816332814283669 + }, + { + "epoch": 6.671515856851906, + "flow_matching_loss": 0.014006641693413258, + "grad_norm": 0.1509387493133545, + "learning_rate": 2.1825322081404236e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004766130819916725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020239079953171314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011556169670075177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6810185492038727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22930, + "weighted_flow_matching_loss": 0.014006641693413258, + "weighted_rkd_loss": 0.007869970705360173 + }, + { + "epoch": 6.674425370963049, + "flow_matching_loss": 0.012972721923142671, + "grad_norm": 0.1549464911222458, + "learning_rate": 2.1817959132233655e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004834105633199215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020767520647495983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011744963377714157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6807744801044464, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22940, + "weighted_flow_matching_loss": 0.012972721923142671, + "weighted_rkd_loss": 0.007995676994323731 + }, + { + "epoch": 6.677334885074193, + "flow_matching_loss": 0.011550508067011833, + "grad_norm": 0.15575946867465973, + "learning_rate": 2.1810594111947104e-05, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044656720943748954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002070065983571112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011001410242170096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6805303573608399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22950, + "weighted_flow_matching_loss": 0.011550508067011833, + "weighted_rkd_loss": 0.007486777612939477 + }, + { + "epoch": 6.680244399185336, + "flow_matching_loss": 0.013885282818228007, + "grad_norm": 0.13290612399578094, + "learning_rate": 2.1803227022781892e-05, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004406106215901673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00187814652454108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010690359026193618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.680286192893982, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22960, + "weighted_flow_matching_loss": 0.013885282818228007, + "weighted_rkd_loss": 0.007272556284442544 + }, + { + "epoch": 6.683153913296479, + "flow_matching_loss": 0.012458581849932671, + "grad_norm": 0.12492220103740692, + "learning_rate": 2.1795857866975933e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004643260594457388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020065006567165254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011293021775782109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6800419747829437, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22970, + "weighted_flow_matching_loss": 0.012458581849932671, + "weighted_rkd_loss": 0.007679636031389237 + }, + { + "epoch": 6.686063427407623, + "flow_matching_loss": 0.013483868725597859, + "grad_norm": 0.22673377394676208, + "learning_rate": 2.178848664676779e-05, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00492708720266819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022454861900769174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012099660839885474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6797977030277252, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22980, + "weighted_flow_matching_loss": 0.013483868725597859, + "weighted_rkd_loss": 0.008225320931524038 + }, + { + "epoch": 6.688972941518767, + "flow_matching_loss": 0.013697500713169575, + "grad_norm": 0.1163158044219017, + "learning_rate": 2.1781113364396643e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004214579262770713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019176760804839432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010346834640949964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6795533895492554, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22990, + "weighted_flow_matching_loss": 0.013697500713169575, + "weighted_rkd_loss": 0.007031220756471157 + }, + { + "epoch": 6.6918824556299095, + "flow_matching_loss": 0.013404125906527042, + "grad_norm": 0.15091755986213684, + "learning_rate": 2.17737380221023e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004910875461064279, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002255067741498351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012076818570494651, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6793090105056763, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23000, + "weighted_flow_matching_loss": 0.013404125906527042, + "weighted_rkd_loss": 0.008204043563455343 + }, + { + "epoch": 6.694791969741053, + "flow_matching_loss": 0.012054427526891232, + "grad_norm": 0.1465545892715454, + "learning_rate": 2.1766360622125198e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004672285774722696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019987821811810135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011343353893607855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6790646135807037, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23010, + "weighted_flow_matching_loss": 0.012054427526891232, + "weighted_rkd_loss": 0.007702968781813979 + }, + { + "epoch": 6.697701483852197, + "flow_matching_loss": 0.011804526858031749, + "grad_norm": 0.10761777311563492, + "learning_rate": 2.175898116670639e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004673065198585391, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019285231363028287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01127465358003974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.678820151090622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23020, + "weighted_flow_matching_loss": 0.011804526858031749, + "weighted_rkd_loss": 0.0076535546220839025 + }, + { + "epoch": 6.70061099796334, + "flow_matching_loss": 0.012114769965410232, + "grad_norm": 0.13899736106395721, + "learning_rate": 2.175159965808757e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004391712765209377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001957352925091982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01074077826924622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6785756349563599, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23030, + "weighted_flow_matching_loss": 0.012114769965410232, + "weighted_rkd_loss": 0.007288502203300596 + }, + { + "epoch": 6.703520512074483, + "flow_matching_loss": 0.012685156334191561, + "grad_norm": 0.12517879903316498, + "learning_rate": 2.1744216098511038e-05, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005088874534703791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024426989490166306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012620448041707278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6783310651779175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23040, + "weighted_flow_matching_loss": 0.012685156334191561, + "weighted_rkd_loss": 0.008560838224366308 + }, + { + "epoch": 6.706430026185627, + "flow_matching_loss": 0.013013995438814163, + "grad_norm": 0.10252757370471954, + "learning_rate": 2.1736830490219722e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004261341691017151, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017879460006952285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010310629243031144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6780864655971527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23050, + "weighted_flow_matching_loss": 0.013013995438814163, + "weighted_rkd_loss": 0.006991599919274449 + }, + { + "epoch": 6.709339540296771, + "flow_matching_loss": 0.013688299944624305, + "grad_norm": 0.09753812104463577, + "learning_rate": 2.1729442835457182e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004647937673144043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019173300242982805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011213205521926284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6778418123722076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23060, + "weighted_flow_matching_loss": 0.013688299944624305, + "weighted_rkd_loss": 0.0076007994823157786 + }, + { + "epoch": 6.712249054407914, + "flow_matching_loss": 0.012334648612886667, + "grad_norm": 0.12599311769008636, + "learning_rate": 2.1722053136467594e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004983792407438159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002340484724845737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012308069504797459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6775970935821534, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23070, + "weighted_flow_matching_loss": 0.012334648612886667, + "weighted_rkd_loss": 0.008339839102700353 + }, + { + "epoch": 6.715158568519057, + "flow_matching_loss": 0.011927607655525207, + "grad_norm": 0.1384623646736145, + "learning_rate": 2.1714661395495743e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004497828776948154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002056698256637901, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01105235582217574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6773523449897766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23080, + "weighted_flow_matching_loss": 0.011927607655525207, + "weighted_rkd_loss": 0.007486251136288047 + }, + { + "epoch": 6.718068082630201, + "flow_matching_loss": 0.012775517720729112, + "grad_norm": 0.13782602548599243, + "learning_rate": 2.170726761478705e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004442841582931578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016988058807328344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010584488976746798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6771075367927551, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23090, + "weighted_flow_matching_loss": 0.012775517720729112, + "weighted_rkd_loss": 0.007166797295212746 + }, + { + "epoch": 6.720977596741344, + "flow_matching_loss": 0.011688196752220392, + "grad_norm": 0.1286454051733017, + "learning_rate": 2.1699871796587554e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004549560369923711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019205184304155408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011019639112055302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6768626809120178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23100, + "weighted_flow_matching_loss": 0.011688196752220392, + "weighted_rkd_loss": 0.007458779867738486 + }, + { + "epoch": 6.723887110852488, + "flow_matching_loss": 0.012606282718479633, + "grad_norm": 0.12080489099025726, + "learning_rate": 2.1692473943143904e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004370812885463237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001963523402810097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010705149080604315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6766177713871002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23110, + "weighted_flow_matching_loss": 0.012606282718479633, + "weighted_rkd_loss": 0.007243306469172239 + }, + { + "epoch": 6.726796624963631, + "flow_matching_loss": 0.012465517176315189, + "grad_norm": 0.10811636596918106, + "learning_rate": 2.1685074056703373e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004434056160971522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00198871212778613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010856824275106192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6763728439807892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23120, + "weighted_flow_matching_loss": 0.012465517176315189, + "weighted_rkd_loss": 0.007343282550573349 + }, + { + "epoch": 6.729706139074775, + "flow_matching_loss": 0.012213713955134153, + "grad_norm": 0.14942291378974915, + "learning_rate": 2.167767213951386e-05, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004107766319066286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001803762000054121, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010019294591620565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6761278212070465, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23130, + "weighted_flow_matching_loss": 0.012213713955134153, + "weighted_rkd_loss": 0.006774347461760044 + }, + { + "epoch": 6.732615653185918, + "flow_matching_loss": 0.013026780914515257, + "grad_norm": 0.1472884565591812, + "learning_rate": 2.1670268193823853e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004861634108237922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021370883332565427, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011860356526449322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6758827805519104, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23140, + "weighted_flow_matching_loss": 0.013026780914515257, + "weighted_rkd_loss": 0.00801615947857499 + }, + { + "epoch": 6.7355251672970615, + "flow_matching_loss": 0.011308938125148415, + "grad_norm": 0.1167597845196724, + "learning_rate": 2.1662862221882495e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004324851580895484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001809184334706515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010458887415006756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6756376922130585, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23150, + "weighted_flow_matching_loss": 0.011308938125148415, + "weighted_rkd_loss": 0.007066410360857844 + }, + { + "epoch": 6.738434681408204, + "flow_matching_loss": 0.012845029402524234, + "grad_norm": 0.12139103561639786, + "learning_rate": 2.1655454225939514e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048338879831135275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00196905427146703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011636830400675535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6753925442695617, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23160, + "weighted_flow_matching_loss": 0.012845029402524234, + "weighted_rkd_loss": 0.007859439961612225 + }, + { + "epoch": 6.741344195519348, + "flow_matching_loss": 0.013387046847492456, + "grad_norm": 0.1425110250711441, + "learning_rate": 2.1648044208245266e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0050377133302390575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021334077464416623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012208834383636713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6751473486423493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23170, + "weighted_flow_matching_loss": 0.013387046847492456, + "weighted_rkd_loss": 0.008242773963138461 + }, + { + "epoch": 6.744253709630492, + "flow_matching_loss": 0.012812396697700023, + "grad_norm": 0.1509045958518982, + "learning_rate": 2.164063217105072e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004799442715011537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001977679890114814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011576565168797969, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6749021112918854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23180, + "weighted_flow_matching_loss": 0.012812396697700023, + "weighted_rkd_loss": 0.00781305180862546 + }, + { + "epoch": 6.7471632237416355, + "flow_matching_loss": 0.012417224887758493, + "grad_norm": 0.13080169260501862, + "learning_rate": 2.163321811660746e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004641551035456359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002062807441689074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011345909349620343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6746568262577057, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23190, + "weighted_flow_matching_loss": 0.012417224887758493, + "weighted_rkd_loss": 0.007654656749218702 + }, + { + "epoch": 6.750072737852778, + "flow_matching_loss": 0.01291595697402954, + "grad_norm": 0.13014808297157288, + "learning_rate": 2.162580204716767e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046814276138320565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018581333220936357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011220988538116216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6744114875793457, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23200, + "weighted_flow_matching_loss": 0.01291595697402954, + "weighted_rkd_loss": 0.007567556668072939 + }, + { + "epoch": 6.752982251963922, + "flow_matching_loss": 0.016226459108293057, + "grad_norm": 0.15077149868011475, + "learning_rate": 2.1618383964984162e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004742459114640951, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019644374726340173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011449355632066727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6741661071777344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23210, + "weighted_flow_matching_loss": 0.016226459108293057, + "weighted_rkd_loss": 0.007718866318464279 + }, + { + "epoch": 6.755891766075066, + "flow_matching_loss": 0.010861055133864284, + "grad_norm": 0.13729605078697205, + "learning_rate": 2.1610963872310358e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042607702780514956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001725519052706659, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010247059725224971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6739206731319427, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23220, + "weighted_flow_matching_loss": 0.010861055133864284, + "weighted_rkd_loss": 0.006905667996034026 + }, + { + "epoch": 6.7588012801862085, + "flow_matching_loss": 0.011942410841584206, + "grad_norm": 0.1363958716392517, + "learning_rate": 2.160354177140028e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004650607984513044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019761394476518037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011277355393394828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6736751973628998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23230, + "weighted_flow_matching_loss": 0.011942410841584206, + "weighted_rkd_loss": 0.007597262645140291 + }, + { + "epoch": 6.761710794297352, + "flow_matching_loss": 0.013843004778027534, + "grad_norm": 0.14075294137001038, + "learning_rate": 2.159611766450857e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004307136847637593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019278683117590844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010542141925543546, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6734296679496765, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23240, + "weighted_flow_matching_loss": 0.013843004778027534, + "weighted_rkd_loss": 0.007099400414153934 + }, + { + "epoch": 6.764620308408496, + "flow_matching_loss": 0.012831865344196558, + "grad_norm": 0.1370101273059845, + "learning_rate": 2.1588691553890475e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004470740375109017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019363473751582206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010877828020602465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6731840968132019, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23250, + "weighted_flow_matching_loss": 0.012831865344196558, + "weighted_rkd_loss": 0.007322749029845 + }, + { + "epoch": 6.76752982251964, + "flow_matching_loss": 0.011503935512155294, + "grad_norm": 0.14983820915222168, + "learning_rate": 2.1581263441801846e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004776570410467684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021542360191233456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011707376968115569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6729384779930114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23260, + "weighted_flow_matching_loss": 0.011503935512155294, + "weighted_rkd_loss": 0.007878406252712012 + }, + { + "epoch": 6.7704393366307825, + "flow_matching_loss": 0.012626976147294045, + "grad_norm": 0.11888887733221054, + "learning_rate": 2.157383333049915e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004126515355892479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018695030361413957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010122533747926354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6726928055286407, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23270, + "weighted_flow_matching_loss": 0.012626976147294045, + "weighted_rkd_loss": 0.006809410499408841 + }, + { + "epoch": 6.773348850741926, + "flow_matching_loss": 0.012486377637833357, + "grad_norm": 0.12494946271181107, + "learning_rate": 2.156640122223947e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004730872390791774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021955245174467564, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01165726939216256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6724470853805542, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23280, + "weighted_flow_matching_loss": 0.012486377637833357, + "weighted_rkd_loss": 0.00783880865201354 + }, + { + "epoch": 6.77625836485307, + "flow_matching_loss": 0.012442020652815699, + "grad_norm": 0.12033551931381226, + "learning_rate": 2.1558967119280468e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004948924039490521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002073744090739638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011971592158079147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6722013235092164, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23290, + "weighted_flow_matching_loss": 0.012442020652815699, + "weighted_rkd_loss": 0.008047275571152568 + }, + { + "epoch": 6.779167878964213, + "flow_matching_loss": 0.013928080443292856, + "grad_norm": 0.1461736261844635, + "learning_rate": 2.155153102388043e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005278299935162068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022983483155258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012854948174208403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6719555139541626, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23300, + "weighted_flow_matching_loss": 0.013928080443292856, + "weighted_rkd_loss": 0.008637992944568396 + }, + { + "epoch": 6.782077393075356, + "flow_matching_loss": 0.013850710634142161, + "grad_norm": 0.1590280383825302, + "learning_rate": 2.1544092938298258e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004598528705537319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020175045472569765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01121456203982234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6717096567153931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23310, + "weighted_flow_matching_loss": 0.013850710634142161, + "weighted_rkd_loss": 0.007532928232103586 + }, + { + "epoch": 6.7849869071865, + "flow_matching_loss": 0.012094697915017605, + "grad_norm": 0.1454770267009735, + "learning_rate": 2.153665286479343e-05, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051572171971201895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021223452873528003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012436779774725437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6714637517929077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23320, + "weighted_flow_matching_loss": 0.012094697915017605, + "weighted_rkd_loss": 0.008350811339914799 + }, + { + "epoch": 6.787896421297643, + "flow_matching_loss": 0.012214331980794668, + "grad_norm": 0.13432371616363525, + "learning_rate": 2.1529210805626058e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004101667716167867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00173419815255329, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009937533736228943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6712177991867065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23330, + "weighted_flow_matching_loss": 0.012214331980794668, + "weighted_rkd_loss": 0.0066702440846711395 + }, + { + "epoch": 6.790805935408787, + "flow_matching_loss": 0.012598138814792038, + "grad_norm": 0.17210210859775543, + "learning_rate": 2.1521766763056835e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00439310169313103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019094369490630925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01069564032368362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.670971792936325, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23340, + "weighted_flow_matching_loss": 0.012598138814792038, + "weighted_rkd_loss": 0.007176463771611452 + }, + { + "epoch": 6.79371544951993, + "flow_matching_loss": 0.012500707106664777, + "grad_norm": 0.12316688150167465, + "learning_rate": 2.151432073934707e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039861964294686915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016096412902697921, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009582034032791852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6707257568836212, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23350, + "weighted_flow_matching_loss": 0.012500707106664777, + "weighted_rkd_loss": 0.006426902953535318 + }, + { + "epoch": 6.796624963631073, + "flow_matching_loss": 0.013598914444446563, + "grad_norm": 0.14808782935142517, + "learning_rate": 2.1506872736758665e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004469920205883682, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018641195259988309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010803959798067808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.670479667186737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23360, + "weighted_flow_matching_loss": 0.013598914444446563, + "weighted_rkd_loss": 0.007243824098259211 + }, + { + "epoch": 6.799534477742217, + "flow_matching_loss": 0.01227481635287404, + "grad_norm": 0.12477267533540726, + "learning_rate": 2.149942275755412e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004842533054761589, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021449045045301316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011829970683902503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6702335178852081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23370, + "weighted_flow_matching_loss": 0.01227481635287404, + "weighted_rkd_loss": 0.007928980840370059 + }, + { + "epoch": 6.802443991853361, + "flow_matching_loss": 0.013191675581037999, + "grad_norm": 0.15644246339797974, + "learning_rate": 2.1491970803996557e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044111447175964715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001818757120054215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01064104661345482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6699873387813569, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23380, + "weighted_flow_matching_loss": 0.013191675581037999, + "weighted_rkd_loss": 0.007129250606521964 + }, + { + "epoch": 6.805353505964504, + "flow_matching_loss": 0.012136599514633417, + "grad_norm": 0.17278721928596497, + "learning_rate": 2.1484516878349667e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004511840408667922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018316709669306875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010855351854115725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6697411000728607, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23390, + "weighted_flow_matching_loss": 0.012136599514633417, + "weighted_rkd_loss": 0.007270258478820324 + }, + { + "epoch": 6.808263020075647, + "flow_matching_loss": 0.01157629475928843, + "grad_norm": 0.12456289678812027, + "learning_rate": 2.147706098287777e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004410153324715793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018579205963760615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010678227245807647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6694948256015778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23400, + "weighted_flow_matching_loss": 0.01157629475928843, + "weighted_rkd_loss": 0.007149024959653616 + }, + { + "epoch": 6.811172534186791, + "flow_matching_loss": 0.012392295710742473, + "grad_norm": 0.11924134194850922, + "learning_rate": 2.146960311984576e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005051995185203851, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002347199304495007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012451189756393432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6692484974861145, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23410, + "weighted_flow_matching_loss": 0.012392295710742473, + "weighted_rkd_loss": 0.008332910807803273 + }, + { + "epoch": 6.8140820482979345, + "flow_matching_loss": 0.012886633910238744, + "grad_norm": 0.15409688651561737, + "learning_rate": 2.146214329151914e-05, + "loss": 0.0211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004938940098509192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023930247290991245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01227090498432517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6690021276473999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23420, + "weighted_flow_matching_loss": 0.012886633910238744, + "weighted_rkd_loss": 0.008209185488522053 + }, + { + "epoch": 6.816991562409077, + "flow_matching_loss": 0.014207271672785283, + "grad_norm": 0.13219532370567322, + "learning_rate": 2.1454681500164012e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004383891448378563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001966653682757169, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01073443666100502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6687557101249695, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23430, + "weighted_flow_matching_loss": 0.014207271672785283, + "weighted_rkd_loss": 0.007178620528429746 + }, + { + "epoch": 6.819901076520221, + "flow_matching_loss": 0.013026878610253334, + "grad_norm": 0.13776995241641998, + "learning_rate": 2.144721774804707e-05, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004619255801662803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018324985750950873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011071010213345289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6685092449188232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23440, + "weighted_flow_matching_loss": 0.013026878610253334, + "weighted_rkd_loss": 0.007401088159531355 + }, + { + "epoch": 6.822810590631365, + "flow_matching_loss": 0.012956425826996566, + "grad_norm": 0.1535203605890274, + "learning_rate": 2.1439752037435607e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003997130575589836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017264781752601265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009720739210024476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6682627320289611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23450, + "weighted_flow_matching_loss": 0.012956425826996566, + "weighted_rkd_loss": 0.00649603889323771 + }, + { + "epoch": 6.8257201047425085, + "flow_matching_loss": 0.011810831632465124, + "grad_norm": 0.12597551941871643, + "learning_rate": 2.1432284370597505e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004116900311782956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018383108545094728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010072111431509257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6680161774158477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23460, + "weighted_flow_matching_loss": 0.011810831632465124, + "weighted_rkd_loss": 0.006728311767801642 + }, + { + "epoch": 6.828629618853651, + "flow_matching_loss": 0.014155333582311868, + "grad_norm": 0.15359580516815186, + "learning_rate": 2.1424814749801246e-05, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004366061161272227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018091452540829778, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010541267506778241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6677695751190186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23470, + "weighted_flow_matching_loss": 0.014155333582311868, + "weighted_rkd_loss": 0.007039241306483745 + }, + { + "epoch": 6.831539132964795, + "flow_matching_loss": 0.013075851369649172, + "grad_norm": 0.1415531188249588, + "learning_rate": 2.1417343177315905e-05, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004441497079096734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020179699873551725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01090096402913332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.667522931098938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23480, + "weighted_flow_matching_loss": 0.013075851369649172, + "weighted_rkd_loss": 0.007276543648913503 + }, + { + "epoch": 6.834448647075939, + "flow_matching_loss": 0.010960824135690928, + "grad_norm": 0.1566021740436554, + "learning_rate": 2.1409869655411145e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004740453232079745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002311006188392639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011791912745684386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6672762334346771, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23490, + "weighted_flow_matching_loss": 0.010960824135690928, + "weighted_rkd_loss": 0.007868385594338178 + }, + { + "epoch": 6.8373581611870815, + "flow_matching_loss": 0.012206106586381793, + "grad_norm": 0.1317168027162552, + "learning_rate": 2.140239418635723e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00469049671664834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020217664656229316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011402759980410338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6670295000076294, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23500, + "weighted_flow_matching_loss": 0.012206106586381793, + "weighted_rkd_loss": 0.007605900056660175 + }, + { + "epoch": 6.840267675298225, + "flow_matching_loss": 0.013629496470093727, + "grad_norm": 0.13780204951763153, + "learning_rate": 2.1394916772425002e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00440001767128706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018087121890857816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010608747461810708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6667827010154724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23510, + "weighted_flow_matching_loss": 0.013629496470093727, + "weighted_rkd_loss": 0.007073783734813333 + }, + { + "epoch": 6.843177189409369, + "flow_matching_loss": 0.01124408869072795, + "grad_norm": 0.14238178730010986, + "learning_rate": 2.138743741588591e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00495026281569153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020234668627381325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01192399263381958, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6665358781814575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23520, + "weighted_flow_matching_loss": 0.01124408869072795, + "weighted_rkd_loss": 0.007947664940729737 + }, + { + "epoch": 6.846086703520512, + "flow_matching_loss": 0.013422740809619426, + "grad_norm": 0.12789759039878845, + "learning_rate": 2.137995611901198e-05, + "loss": 0.0219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005176485562697053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023113642586395145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012664335407316684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6662889897823334, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23530, + "weighted_flow_matching_loss": 0.013422740809619426, + "weighted_rkd_loss": 0.008437997568398714 + }, + { + "epoch": 6.8489962176316554, + "flow_matching_loss": 0.01189859826117754, + "grad_norm": 0.12940716743469238, + "learning_rate": 2.1372472884075835e-05, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004727568570524454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020471583702601492, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011502295546233655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6660420715808868, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23540, + "weighted_flow_matching_loss": 0.01189859826117754, + "weighted_rkd_loss": 0.007661013398319483 + }, + { + "epoch": 6.851905731742799, + "flow_matching_loss": 0.012374084442853928, + "grad_norm": 0.11806973069906235, + "learning_rate": 2.136498771335068e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051245162263512615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002322845614980906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012571878172457219, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.66579509973526, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23550, + "weighted_flow_matching_loss": 0.012374084442853928, + "weighted_rkd_loss": 0.008370410697534681 + }, + { + "epoch": 6.854815245853942, + "flow_matching_loss": 0.013955555111169814, + "grad_norm": 0.11360188573598862, + "learning_rate": 2.1357500609110324e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004335883795283735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019050230155698955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010576790664345026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6655480861663818, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23560, + "weighted_flow_matching_loss": 0.013955555111169814, + "weighted_rkd_loss": 0.007039426825940609 + }, + { + "epoch": 6.857724759965086, + "flow_matching_loss": 0.012227218132466078, + "grad_norm": 0.14034144580364227, + "learning_rate": 2.1350011573629134e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004766182065941394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018208747846074403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011353238951414824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6653010249137878, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23570, + "weighted_flow_matching_loss": 0.012227218132466078, + "weighted_rkd_loss": 0.007553353486582637 + }, + { + "epoch": 6.860634274076229, + "flow_matching_loss": 0.01221288638189435, + "grad_norm": 0.18920034170150757, + "learning_rate": 2.1342520609182093e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004215571261011064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018563592224381865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010287501802667975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.665053915977478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23580, + "weighted_flow_matching_loss": 0.01221288638189435, + "weighted_rkd_loss": 0.00684173391200602 + }, + { + "epoch": 6.863543788187373, + "flow_matching_loss": 0.012598925083875657, + "grad_norm": 0.1533042937517166, + "learning_rate": 2.1335027718044753e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004964105552062392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020882192417047917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01201643031090498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6648067653179168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23590, + "weighted_flow_matching_loss": 0.012598925083875657, + "weighted_rkd_loss": 0.007988665206357836 + }, + { + "epoch": 6.866453302298516, + "flow_matching_loss": 0.013048904947936535, + "grad_norm": 0.1588541865348816, + "learning_rate": 2.1327532902493263e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004424059973098337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017659458797425031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010614065639674663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6645595729351044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23600, + "weighted_flow_matching_loss": 0.013048904947936535, + "weighted_rkd_loss": 0.007053619343787431 + }, + { + "epoch": 6.86936281640966, + "flow_matching_loss": 0.013279880676418544, + "grad_norm": 0.12157098203897476, + "learning_rate": 2.1320036164804345e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004968174383975565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002134535543154925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012070884369313717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.664312332868576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23610, + "weighted_flow_matching_loss": 0.013279880676418544, + "weighted_rkd_loss": 0.008018898125737906 + }, + { + "epoch": 6.872272330520803, + "flow_matching_loss": 0.011874888930469752, + "grad_norm": 0.13813412189483643, + "learning_rate": 2.131253750725531e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00461362509522587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022559457924216987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01148319598287344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6640650451183319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23620, + "weighted_flow_matching_loss": 0.011874888930469752, + "weighted_rkd_loss": 0.007625584909692407 + }, + { + "epoch": 6.875181844631946, + "flow_matching_loss": 0.01272389367222786, + "grad_norm": 0.25496208667755127, + "learning_rate": 2.1305036932124053e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046803486766293645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002024905104190111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011385602597147226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6638177335262299, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23630, + "weighted_flow_matching_loss": 0.01272389367222786, + "weighted_rkd_loss": 0.007557821366935969 + }, + { + "epoch": 6.87809135874309, + "flow_matching_loss": 0.012035996839404105, + "grad_norm": 0.12632997334003448, + "learning_rate": 2.1297534441689054e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003919106884859502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016662897658534348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009504503384232522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6635703384876251, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23640, + "weighted_flow_matching_loss": 0.012035996839404105, + "weighted_rkd_loss": 0.006306894170120358 + }, + { + "epoch": 6.881000872854234, + "flow_matching_loss": 0.013343307469040156, + "grad_norm": 0.12503965198993683, + "learning_rate": 2.1290030038229365e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004751244233921171, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001948036381509155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011450524860993027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.663322925567627, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23650, + "weighted_flow_matching_loss": 0.013343307469040156, + "weighted_rkd_loss": 0.007595474133267999 + }, + { + "epoch": 6.883910386965377, + "flow_matching_loss": 0.012964849546551704, + "grad_norm": 0.1347222775220871, + "learning_rate": 2.1282523724024632e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004684974625706673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002013895462732762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011383844818919898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6630754590034484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23660, + "weighted_flow_matching_loss": 0.012964849546551704, + "weighted_rkd_loss": 0.007548376359045506 + }, + { + "epoch": 6.88681990107652, + "flow_matching_loss": 0.012139649130403996, + "grad_norm": 0.17097108066082, + "learning_rate": 2.127501550135507e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005176847381517291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022285961313173175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012582290824502707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6628279447555542, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23670, + "weighted_flow_matching_loss": 0.012139649130403996, + "weighted_rkd_loss": 0.008339824993163348 + }, + { + "epoch": 6.889729415187664, + "flow_matching_loss": 0.012953888904303313, + "grad_norm": 0.15985414385795593, + "learning_rate": 2.126750537250148e-05, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005105897225439549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022941293776966632, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012505923956632614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6625803887844086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23680, + "weighted_flow_matching_loss": 0.012953888904303313, + "weighted_rkd_loss": 0.008286192920058966 + }, + { + "epoch": 6.892638929298807, + "flow_matching_loss": 0.013220309605821967, + "grad_norm": 0.14334280788898468, + "learning_rate": 2.1259993339745244e-05, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004465805017389357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018777903402224182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010809400351718068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6623327910900116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23690, + "weighted_flow_matching_loss": 0.013220309605821967, + "weighted_rkd_loss": 0.007159355934709311 + }, + { + "epoch": 6.89554844340995, + "flow_matching_loss": 0.012742285057902336, + "grad_norm": 0.1657908707857132, + "learning_rate": 2.1252479405368314e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00504552295897156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022921775700524448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012383223790675402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6620851457118988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23700, + "weighted_flow_matching_loss": 0.012742285057902336, + "weighted_rkd_loss": 0.008198696607723831 + }, + { + "epoch": 6.898457957521094, + "flow_matching_loss": 0.011999879498034716, + "grad_norm": 0.15080571174621582, + "learning_rate": 2.1244963571653232e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004878347739577293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022480053827166556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0120047009550035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6618374586105347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23710, + "weighted_flow_matching_loss": 0.011999879498034716, + "weighted_rkd_loss": 0.007945101847872138 + }, + { + "epoch": 6.901367471632238, + "flow_matching_loss": 0.013558409921824933, + "grad_norm": 0.12298759073019028, + "learning_rate": 2.1237445840883103e-05, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004983838531188667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023951211012899877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012362798070535063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6615897297859192, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23720, + "weighted_flow_matching_loss": 0.013558409921824933, + "weighted_rkd_loss": 0.008179122768342494 + }, + { + "epoch": 6.904276985743381, + "flow_matching_loss": 0.01273064138367772, + "grad_norm": 0.14399833977222443, + "learning_rate": 2.1229926215341616e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004930346249602735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021236578235402704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011984350346028804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6613419532775879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23730, + "weighted_flow_matching_loss": 0.01273064138367772, + "weighted_rkd_loss": 0.007925773691385984 + }, + { + "epoch": 6.907186499854524, + "flow_matching_loss": 0.013635682733729481, + "grad_norm": 0.16701143980026245, + "learning_rate": 2.1222404697313036e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004576506279408932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001971228013280779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011124240513890982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6610941290855408, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23740, + "weighted_flow_matching_loss": 0.013635682733729481, + "weighted_rkd_loss": 0.007354145357385278 + }, + { + "epoch": 6.910096013965668, + "flow_matching_loss": 0.013761371839791536, + "grad_norm": 0.1293705701828003, + "learning_rate": 2.1214881289082205e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043058812385424975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019798788940533995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010591641534119844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6608462631702423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23750, + "weighted_flow_matching_loss": 0.013761371839791536, + "weighted_rkd_loss": 0.006999485567212105 + }, + { + "epoch": 6.913005528076811, + "flow_matching_loss": 0.01411342117935419, + "grad_norm": 0.13483871519565582, + "learning_rate": 2.1207355992934535e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004541397118009627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001926651073154062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011009445413947105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6605983376502991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23760, + "weighted_flow_matching_loss": 0.01411342117935419, + "weighted_rkd_loss": 0.007272895984351635 + }, + { + "epoch": 6.9159150421879545, + "flow_matching_loss": 0.012950681056827307, + "grad_norm": 0.1229194924235344, + "learning_rate": 2.119982881115601e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004955803463235498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021663315827026965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012077938579022885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6603503823280334, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23770, + "weighted_flow_matching_loss": 0.012950681056827307, + "weighted_rkd_loss": 0.007975663291290402 + }, + { + "epoch": 6.918824556299098, + "flow_matching_loss": 0.013407660834491252, + "grad_norm": 0.14425702393054962, + "learning_rate": 2.1192299746033187e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004108316940255463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001818637375254184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010035271337255836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6601024031639099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23780, + "weighted_flow_matching_loss": 0.013407660834491252, + "weighted_rkd_loss": 0.00662422995083034 + }, + { + "epoch": 6.921734070410242, + "flow_matching_loss": 0.014530217088758946, + "grad_norm": 0.1289755403995514, + "learning_rate": 2.1184768799853195e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005151831568218768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002370280958712101, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012673944141715765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6598543584346771, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23790, + "weighted_flow_matching_loss": 0.014530217088758946, + "weighted_rkd_loss": 0.008362848358228803 + }, + { + "epoch": 6.924643584521385, + "flow_matching_loss": 0.012335775420069695, + "grad_norm": 0.11845104396343231, + "learning_rate": 2.1177235974903748e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004845583299174905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002086795156355947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01177796171978116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6596062779426575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23800, + "weighted_flow_matching_loss": 0.012335775420069695, + "weighted_rkd_loss": 0.007768819946795702 + }, + { + "epoch": 6.927553098632528, + "flow_matching_loss": 0.01258522281423211, + "grad_norm": 0.11572683602571487, + "learning_rate": 2.116970127347311e-05, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004264985769987106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020353381405584514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01056530987843871, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.659358149766922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23810, + "weighted_flow_matching_loss": 0.01258522281423211, + "weighted_rkd_loss": 0.006966379703953862 + }, + { + "epoch": 6.930462612743672, + "flow_matching_loss": 0.012675085477530957, + "grad_norm": 0.12003514170646667, + "learning_rate": 2.1162164697850127e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040781148709356785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001673148013651371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009829377848654985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6591099739074707, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23820, + "weighted_flow_matching_loss": 0.012675085477530957, + "weighted_rkd_loss": 0.00647866390645504 + }, + { + "epoch": 6.933372126854815, + "flow_matching_loss": 0.012298698723316192, + "grad_norm": 0.15121355652809143, + "learning_rate": 2.1154626250324206e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004671820905059576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019763103337027133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011319952178746462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6588617563247681, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23830, + "weighted_flow_matching_loss": 0.012298698723316192, + "weighted_rkd_loss": 0.007458228059113026 + }, + { + "epoch": 6.936281640965959, + "flow_matching_loss": 0.013982191309332847, + "grad_norm": 0.14940303564071655, + "learning_rate": 2.1147085933185338e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004584892885759473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020645006094127893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011234286427497863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6586134970188141, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23840, + "weighted_flow_matching_loss": 0.013982191309332847, + "weighted_rkd_loss": 0.0073989722412079574 + }, + { + "epoch": 6.939191155077102, + "flow_matching_loss": 0.011489089159294964, + "grad_norm": 0.12237085402011871, + "learning_rate": 2.113954374872407e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00447579228784889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018384466995485127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010790031217038632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6583651959896087, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23850, + "weighted_flow_matching_loss": 0.011489089159294964, + "weighted_rkd_loss": 0.00710377017967403 + }, + { + "epoch": 6.942100669188245, + "flow_matching_loss": 0.013358617946505547, + "grad_norm": 0.13758200407028198, + "learning_rate": 2.1131999699231513e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004546237410977483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019372693030163645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011029744055122137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6581168532371521, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23860, + "weighted_flow_matching_loss": 0.013358617946505547, + "weighted_rkd_loss": 0.007258792174980044 + }, + { + "epoch": 6.945010183299389, + "flow_matching_loss": 0.013008522242307663, + "grad_norm": 0.13205260038375854, + "learning_rate": 2.1124453786999354e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004901093221269548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021938606631010773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011996047105640173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6578684628009797, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23870, + "weighted_flow_matching_loss": 0.013008522242307663, + "weighted_rkd_loss": 0.007891899067908525 + }, + { + "epoch": 6.947919697410533, + "flow_matching_loss": 0.012661202065646648, + "grad_norm": 0.11852356046438217, + "learning_rate": 2.111690601431984e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004716882249340415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019501916598528624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011383956205099821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6576200187206268, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23880, + "weighted_flow_matching_loss": 0.012661202065646648, + "weighted_rkd_loss": 0.007486299332231283 + }, + { + "epoch": 6.950829211521675, + "flow_matching_loss": 0.011789107136428357, + "grad_norm": 0.14521323144435883, + "learning_rate": 2.110935638348579e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00427739720325917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019218645757064223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010476658865809441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6573715507984161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23890, + "weighted_flow_matching_loss": 0.011789107136428357, + "weighted_rkd_loss": 0.006886983150616288 + }, + { + "epoch": 6.953738725632819, + "flow_matching_loss": 0.012944830302149057, + "grad_norm": 0.1309138983488083, + "learning_rate": 2.1101804896790576e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004433823539875448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001898398296907544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010766045469790698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6571230351924896, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23900, + "weighted_flow_matching_loss": 0.012944830302149057, + "weighted_rkd_loss": 0.007074676966294646 + }, + { + "epoch": 6.956648239743963, + "flow_matching_loss": 0.012712010927498341, + "grad_norm": 0.15062040090560913, + "learning_rate": 2.1094251556528143e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004338750708848238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020502962288446724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010727797541767359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6568744540214538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23910, + "weighted_flow_matching_loss": 0.012712010927498341, + "weighted_rkd_loss": 0.007046834984794259 + }, + { + "epoch": 6.9595577538551066, + "flow_matching_loss": 0.014224014896899461, + "grad_norm": 0.13336944580078125, + "learning_rate": 2.1086696364993e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00425059103872627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018825028790161013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010383685026317835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6566258609294892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23920, + "weighted_flow_matching_loss": 0.014224014896899461, + "weighted_rkd_loss": 0.006818252149969339 + }, + { + "epoch": 6.962467267966249, + "flow_matching_loss": 0.012655187491327524, + "grad_norm": 0.1349928230047226, + "learning_rate": 2.107913932448021e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004625609889626503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019053336349315942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011156553449109197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6563772082328796, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23930, + "weighted_flow_matching_loss": 0.012655187491327524, + "weighted_rkd_loss": 0.0073228659573942425 + }, + { + "epoch": 6.965376782077393, + "flow_matching_loss": 0.013172418903559446, + "grad_norm": 0.13037239015102386, + "learning_rate": 2.10715804372854e-05, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003993964940309525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001871511631179601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009859441546723247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6561285197734833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23940, + "weighted_flow_matching_loss": 0.013172418903559446, + "weighted_rkd_loss": 0.006469128606840968 + }, + { + "epoch": 6.968286296188537, + "flow_matching_loss": 0.012562229577451945, + "grad_norm": 0.14475591480731964, + "learning_rate": 2.1064019705704767e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004043734283186495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016506798448972404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009738148376345634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6558797895908356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23950, + "weighted_flow_matching_loss": 0.012562229577451945, + "weighted_rkd_loss": 0.006387038575485349 + }, + { + "epoch": 6.97119581029968, + "flow_matching_loss": 0.012068021111190319, + "grad_norm": 0.129886656999588, + "learning_rate": 2.1056457132035062e-05, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004706942313350737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020550113287754355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011468895990401507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.655631011724472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23960, + "weighted_flow_matching_loss": 0.012068021111190319, + "weighted_rkd_loss": 0.007519293250516057 + }, + { + "epoch": 6.974105324410823, + "flow_matching_loss": 0.012392820930108428, + "grad_norm": 0.1138048842549324, + "learning_rate": 2.1048892718573586e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042328029870986935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001790224970318377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01025583096779883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6553821861743927, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23970, + "weighted_flow_matching_loss": 0.012392820930108428, + "weighted_rkd_loss": 0.006721475254744291 + }, + { + "epoch": 6.977014838521967, + "flow_matching_loss": 0.01155055481940508, + "grad_norm": 0.13975201547145844, + "learning_rate": 2.1041326467618215e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004118883982300758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018025281373411418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010040296241641045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6551333248615265, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23980, + "weighted_flow_matching_loss": 0.01155055481940508, + "weighted_rkd_loss": 0.006577759515494108 + }, + { + "epoch": 6.979924352633111, + "flow_matching_loss": 0.012331664096564054, + "grad_norm": 0.13280123472213745, + "learning_rate": 2.103375838146738e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004019649885594845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017749812104739249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009814281016588211, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.654884421825409, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23990, + "weighted_flow_matching_loss": 0.012331664096564054, + "weighted_rkd_loss": 0.006427213130518794 + }, + { + "epoch": 6.9828338667442535, + "flow_matching_loss": 0.012852819822728633, + "grad_norm": 0.14816507697105408, + "learning_rate": 2.1026188462420052e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004549634899012745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020385968149639667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011137866647914052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6546354770660401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24000, + "weighted_flow_matching_loss": 0.012852819822728633, + "weighted_rkd_loss": 0.007291206857189536 + }, + { + "epoch": 6.985743380855397, + "flow_matching_loss": 0.01231338046491146, + "grad_norm": 0.12771748006343842, + "learning_rate": 2.1018616712775792e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004329971759580076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018517748452723027, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010511718271300197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6543864846229553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24010, + "weighted_flow_matching_loss": 0.01231338046491146, + "weighted_rkd_loss": 0.006878726929426193 + }, + { + "epoch": 6.988652894966541, + "flow_matching_loss": 0.01171292974613607, + "grad_norm": 0.13741940259933472, + "learning_rate": 2.1011043134834676e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004247016296721995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017538660089485348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010247898567467929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6541374564170838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24020, + "weighted_flow_matching_loss": 0.01171292974613607, + "weighted_rkd_loss": 0.0067034672480076555 + }, + { + "epoch": 6.991562409077684, + "flow_matching_loss": 0.0121675212867558, + "grad_norm": 0.15215876698493958, + "learning_rate": 2.1003467730897374e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004299096716567874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002045732596889138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010643926076591015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6538883805274963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24030, + "weighted_flow_matching_loss": 0.0121675212867558, + "weighted_rkd_loss": 0.00695989029482007 + }, + { + "epoch": 6.9944719231888275, + "flow_matching_loss": 0.011957756010815501, + "grad_norm": 0.11079666018486023, + "learning_rate": 2.0995890503265086e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004273786302655935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001888587512075901, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010436160210520028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6536392569541931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24040, + "weighted_flow_matching_loss": 0.011957756010815501, + "weighted_rkd_loss": 0.0068215086124837395 + }, + { + "epoch": 6.997381437299971, + "flow_matching_loss": 0.012537654861807824, + "grad_norm": 0.1253570318222046, + "learning_rate": 2.0988311454239575e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004455043142661453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020647135796025397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010974799888208509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.653390109539032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24050, + "weighted_flow_matching_loss": 0.012537654861807824, + "weighted_rkd_loss": 0.007170855952426791 + }, + { + "epoch": 7.000290951411114, + "flow_matching_loss": 0.013783885724842548, + "grad_norm": 0.13175234198570251, + "learning_rate": 2.098073058612316e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004467217973433435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018545098952017724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010788945900276303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6531409025192261, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24060, + "weighted_flow_matching_loss": 0.013783885724842548, + "weighted_rkd_loss": 0.007046758010983467 + }, + { + "epoch": 7.003200465522258, + "flow_matching_loss": 0.011817315965890885, + "grad_norm": 0.13240914046764374, + "learning_rate": 2.0973147901218705e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004568842216394841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020776954130269586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011215379927307367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6528916597366333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24070, + "weighted_flow_matching_loss": 0.011817315965890885, + "weighted_rkd_loss": 0.007322369143366814 + }, + { + "epoch": 7.006109979633401, + "flow_matching_loss": 0.012055878713726997, + "grad_norm": 0.15373508632183075, + "learning_rate": 2.0965563401829625e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004334119101986289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018338977359235286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010502135986462236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6526423752307892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24080, + "weighted_flow_matching_loss": 0.012055878713726997, + "weighted_rkd_loss": 0.0068541104439646 + }, + { + "epoch": 7.009019493744544, + "flow_matching_loss": 0.01337758218869567, + "grad_norm": 0.14300431311130524, + "learning_rate": 2.0957977090259897e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004493409884162247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019022417603991927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010889061447232961, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6523930549621582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24090, + "weighted_flow_matching_loss": 0.01337758218869567, + "weighted_rkd_loss": 0.007104012090712786 + }, + { + "epoch": 7.011929007855688, + "flow_matching_loss": 0.012753681186586619, + "grad_norm": 0.1416812241077423, + "learning_rate": 2.095038896881404e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004022487392649055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017140115029178559, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009758986253291368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.652143681049347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24100, + "weighted_flow_matching_loss": 0.012753681186586619, + "weighted_rkd_loss": 0.0063642487861216065 + }, + { + "epoch": 7.014838521966832, + "flow_matching_loss": 0.012173536792397499, + "grad_norm": 0.11986871063709259, + "learning_rate": 2.0942799039797128e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004529509879648686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001957578305155039, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011016598157584667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6518942713737488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24110, + "weighted_flow_matching_loss": 0.012173536792397499, + "weighted_rkd_loss": 0.007181753637269139 + }, + { + "epoch": 7.017748036077975, + "flow_matching_loss": 0.013559770304709673, + "grad_norm": 0.1498069167137146, + "learning_rate": 2.0935207305514777e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00503855615388602, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021502184099517763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012227330636233092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6516448199748993, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24120, + "weighted_flow_matching_loss": 0.013559770304709673, + "weighted_rkd_loss": 0.007967777131125332 + }, + { + "epoch": 7.020657550189118, + "flow_matching_loss": 0.012142159882932902, + "grad_norm": 0.131413534283638, + "learning_rate": 2.0927613768273156e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038816101383417844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016699644387699665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009433184843510389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6513953268527984, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24130, + "weighted_flow_matching_loss": 0.012142159882932902, + "weighted_rkd_loss": 0.006144705880433321 + }, + { + "epoch": 7.023567064300262, + "flow_matching_loss": 0.011302322708070278, + "grad_norm": 0.11054199188947678, + "learning_rate": 2.0920018430378984e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004233100265264511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017666127299889923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010232813330367208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6511457920074463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24140, + "weighted_flow_matching_loss": 0.011302322708070278, + "weighted_rkd_loss": 0.006663103681057691 + }, + { + "epoch": 7.026476578411406, + "flow_matching_loss": 0.0127297667786479, + "grad_norm": 0.15849217772483826, + "learning_rate": 2.0912421294139517e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004632636252790689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001970378018449992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011235650535672903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6508962213993073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24150, + "weighted_flow_matching_loss": 0.0127297667786479, + "weighted_rkd_loss": 0.007313140342012048 + }, + { + "epoch": 7.029386092522548, + "flow_matching_loss": 0.01237767217680812, + "grad_norm": 0.12385335564613342, + "learning_rate": 2.0904822361862572e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004457256384193897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018427790957503021, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0107572918292135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6506465971469879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24160, + "weighted_flow_matching_loss": 0.01237767217680812, + "weighted_rkd_loss": 0.006999215530231595 + }, + { + "epoch": 7.032295606633692, + "flow_matching_loss": 0.011296105198562145, + "grad_norm": 0.10378610342741013, + "learning_rate": 2.0897221635856498e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004646607651375234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019333928707055748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011226608045399189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6503969430923462, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24170, + "weighted_flow_matching_loss": 0.011296105198562145, + "weighted_rkd_loss": 0.0073018176946789025 + }, + { + "epoch": 7.035205120744836, + "flow_matching_loss": 0.01246905755251646, + "grad_norm": 0.14394716918468475, + "learning_rate": 2.08896191184302e-05, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004547226754948497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019458225928246976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011040276009589433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6501472294330597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24180, + "weighted_flow_matching_loss": 0.01246905755251646, + "weighted_rkd_loss": 0.007177743036299944 + }, + { + "epoch": 7.038114634855979, + "flow_matching_loss": 0.012865625787526369, + "grad_norm": 0.12868019938468933, + "learning_rate": 2.0882014811893124e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004466957692056894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020085938624106346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010942509211599828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6498974859714508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24190, + "weighted_flow_matching_loss": 0.012865625787526369, + "weighted_rkd_loss": 0.007111507328227162 + }, + { + "epoch": 7.041024148967122, + "flow_matching_loss": 0.012353547848761082, + "grad_norm": 0.14419740438461304, + "learning_rate": 2.0874408718555254e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004084739880636334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017447966267354787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009914276422932744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.649647718667984, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24200, + "weighted_flow_matching_loss": 0.012353547848761082, + "weighted_rkd_loss": 0.006440809462219477 + }, + { + "epoch": 7.043933663078266, + "flow_matching_loss": 0.012823763862252235, + "grad_norm": 0.11941105872392654, + "learning_rate": 2.0866800840727114e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004357770341448486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018630404025316238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010578581038862467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6493978917598724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24210, + "weighted_flow_matching_loss": 0.012823763862252235, + "weighted_rkd_loss": 0.006869703577831387 + }, + { + "epoch": 7.04684317718941, + "flow_matching_loss": 0.012734199501574039, + "grad_norm": 0.16790325939655304, + "learning_rate": 2.0859191180719786e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004679199657402933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021039549726992847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01146235428750515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.649148029088974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24220, + "weighted_flow_matching_loss": 0.012734199501574039, + "weighted_rkd_loss": 0.007440678682178259 + }, + { + "epoch": 7.049752691300553, + "flow_matching_loss": 0.013664416130632162, + "grad_norm": 0.13219986855983734, + "learning_rate": 2.085157974084488e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004551652329973876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018669043784029782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01097020898014307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6488981246948242, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24230, + "weighted_flow_matching_loss": 0.013664416130632162, + "weighted_rkd_loss": 0.007118511805310845 + }, + { + "epoch": 7.052662205411696, + "flow_matching_loss": 0.011362730897963047, + "grad_norm": 0.15075038373470306, + "learning_rate": 2.0843966523414553e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003821378503926098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016156430006958543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00925839999690652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6486481785774231, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24240, + "weighted_flow_matching_loss": 0.011362730897963047, + "weighted_rkd_loss": 0.006005340674892068 + }, + { + "epoch": 7.05557171952284, + "flow_matching_loss": 0.014257279317826032, + "grad_norm": 0.11016899347305298, + "learning_rate": 2.0836351530741492e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00396047302056104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017480931128375232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009669039165601134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6483981907367706, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24250, + "weighted_flow_matching_loss": 0.014257279317826032, + "weighted_rkd_loss": 0.006269413093104958 + }, + { + "epoch": 7.058481233633983, + "flow_matching_loss": 0.012665334343910217, + "grad_norm": 0.11575847119092941, + "learning_rate": 2.082873476513894e-05, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004216497670859099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018157769227400422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010248772241175175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6481481611728668, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24260, + "weighted_flow_matching_loss": 0.012665334343910217, + "weighted_rkd_loss": 0.006642688903957606 + }, + { + "epoch": 7.0613907477451265, + "flow_matching_loss": 0.013576273247599602, + "grad_norm": 0.11032005399465561, + "learning_rate": 2.0821116228920663e-05, + "loss": 0.021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004712990578263998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020229530637152493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011448934022337199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6478980958461762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24270, + "weighted_flow_matching_loss": 0.013576273247599602, + "weighted_rkd_loss": 0.007417799811810255 + }, + { + "epoch": 7.06430026185627, + "flow_matching_loss": 0.014544065389782191, + "grad_norm": 0.16302251815795898, + "learning_rate": 2.0813495924400977e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005118085187859833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00231562975095585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012551800068467855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6476479887962341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24280, + "weighted_flow_matching_loss": 0.014544065389782191, + "weighted_rkd_loss": 0.008129053516313434 + }, + { + "epoch": 7.067209775967413, + "flow_matching_loss": 0.012260599993169308, + "grad_norm": 0.12260858714580536, + "learning_rate": 2.0805873853894723e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00434989077039063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018089013989083468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010508682951331139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6473978400230408, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24290, + "weighted_flow_matching_loss": 0.012260599993169308, + "weighted_rkd_loss": 0.006803231686353684 + }, + { + "epoch": 7.070119290078557, + "flow_matching_loss": 0.012737061083316802, + "grad_norm": 0.15792743861675262, + "learning_rate": 2.0798250019717284e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00442453830037266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00187211234588176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010721188876777888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6471476495265961, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24300, + "weighted_flow_matching_loss": 0.012737061083316802, + "weighted_rkd_loss": 0.006938111456111074 + }, + { + "epoch": 7.0730288041897005, + "flow_matching_loss": 0.011688003875315189, + "grad_norm": 0.12465894967317581, + "learning_rate": 2.079062442418459e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004483636585064232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019527477328665555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010920021031051875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6468974173069, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24310, + "weighted_flow_matching_loss": 0.011688003875315189, + "weighted_rkd_loss": 0.0070641755126416685 + }, + { + "epoch": 7.075938318300844, + "flow_matching_loss": 0.012510093580931424, + "grad_norm": 0.15580175817012787, + "learning_rate": 2.078299706961308e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004759070579893887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002195713296532631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01171385427005589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6466471493244171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24320, + "weighted_flow_matching_loss": 0.012510093580931424, + "weighted_rkd_loss": 0.007574644964188337 + }, + { + "epoch": 7.078847832411987, + "flow_matching_loss": 0.01322571923956275, + "grad_norm": 0.11904685944318771, + "learning_rate": 2.077536795831976e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004443219001404941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001822482654824853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010708920657634735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.646396815776825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24330, + "weighted_flow_matching_loss": 0.01322571923956275, + "weighted_rkd_loss": 0.0069221238605678085 + }, + { + "epoch": 7.081757346523131, + "flow_matching_loss": 0.011992171313613654, + "grad_norm": 0.13932958245277405, + "learning_rate": 2.0767737092622137e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038919424870982764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016155070392414928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009399391990154982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6461464762687683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24340, + "weighted_flow_matching_loss": 0.011992171313613654, + "weighted_rkd_loss": 0.006073422683402896 + }, + { + "epoch": 7.084666860634274, + "flow_matching_loss": 0.012527660839259625, + "grad_norm": 0.13295358419418335, + "learning_rate": 2.076010447483828e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004947081930004061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020121029810979962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011906266771256923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6458961009979248, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24350, + "weighted_flow_matching_loss": 0.012527660839259625, + "weighted_rkd_loss": 0.007690362492576241 + }, + { + "epoch": 7.087576374745417, + "flow_matching_loss": 0.013594894437119365, + "grad_norm": 0.13515937328338623, + "learning_rate": 2.0752470107286765e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004070219560526311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001690576144028455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009831015253439546, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6456456661224366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24360, + "weighted_flow_matching_loss": 0.013594894437119365, + "weighted_rkd_loss": 0.006347354268655181 + }, + { + "epoch": 7.090485888856561, + "flow_matching_loss": 0.014861862733960151, + "grad_norm": 0.11916367709636688, + "learning_rate": 2.074483399228672e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005025401408784091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021680235397070645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012218826171010732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6453951954841614, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24370, + "weighted_flow_matching_loss": 0.014861862733960151, + "weighted_rkd_loss": 0.007885970594361424 + }, + { + "epoch": 7.093395402967705, + "flow_matching_loss": 0.011827607778832317, + "grad_norm": 0.12105803191661835, + "learning_rate": 2.0737196132157788e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041504046646878125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001809973013587296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010110782366245985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6451446890830994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24380, + "weighted_flow_matching_loss": 0.011827607778832317, + "weighted_rkd_loss": 0.006522972229868173 + }, + { + "epoch": 7.0963049170788475, + "flow_matching_loss": 0.012722467631101608, + "grad_norm": 0.11690031737089157, + "learning_rate": 2.0729556529220156e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004445821815170348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019587570219300686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010850400757044553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6448941349983215, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24390, + "weighted_flow_matching_loss": 0.012722467631101608, + "weighted_rkd_loss": 0.006997337052598596 + }, + { + "epoch": 7.099214431189991, + "flow_matching_loss": 0.01232232442125678, + "grad_norm": 0.1161559447646141, + "learning_rate": 2.0721915185794532e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037891795160248875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001509634853573516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00908799390308559, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6446435451507568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24400, + "weighted_flow_matching_loss": 0.01232232442125678, + "weighted_rkd_loss": 0.0058584923157468435 + }, + { + "epoch": 7.102123945301135, + "flow_matching_loss": 0.012188238278031348, + "grad_norm": 0.12463315576314926, + "learning_rate": 2.071427210420215e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004272335674613714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017861531232483685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010330824553966523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6443929076194763, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24410, + "weighted_flow_matching_loss": 0.012188238278031348, + "weighted_rkd_loss": 0.006657067500054836 + }, + { + "epoch": 7.105033459412278, + "flow_matching_loss": 0.012238301150500775, + "grad_norm": 0.1372300535440445, + "learning_rate": 2.070662728676478e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003940438758581877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017245961353182793, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009605473559349775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6441422522068023, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24420, + "weighted_flow_matching_loss": 0.012238301150500775, + "weighted_rkd_loss": 0.0061872407328337434 + }, + { + "epoch": 7.107942973523421, + "flow_matching_loss": 0.011832608329132199, + "grad_norm": 0.17964357137680054, + "learning_rate": 2.0698980735804712e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004149283701553941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018072257866151632, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01010579327121377, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6438915312290192, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24430, + "weighted_flow_matching_loss": 0.011832608329132199, + "weighted_rkd_loss": 0.006506938580423593 + }, + { + "epoch": 7.110852487634565, + "flow_matching_loss": 0.012715901993215085, + "grad_norm": 0.1629556566476822, + "learning_rate": 2.0691332453644775e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004520129295997322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019587975926697253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01099905613809824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6436407923698425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24440, + "weighted_flow_matching_loss": 0.012715901993215085, + "weighted_rkd_loss": 0.0070795046631246805 + }, + { + "epoch": 7.113762001745709, + "flow_matching_loss": 0.012688753008842469, + "grad_norm": 0.13535751402378082, + "learning_rate": 2.0683682442608306e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005072878883220255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021225708653219042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012268328573554754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6433899998664856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24450, + "weighted_flow_matching_loss": 0.012688753008842469, + "weighted_rkd_loss": 0.0078934240154922 + }, + { + "epoch": 7.116671515856852, + "flow_matching_loss": 0.011948985140770673, + "grad_norm": 0.10972319543361664, + "learning_rate": 2.0676030705019182e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004068633425049484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016947925905697048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00983205959200859, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6431391775608063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24460, + "weighted_flow_matching_loss": 0.011948985140770673, + "weighted_rkd_loss": 0.006323441304266453 + }, + { + "epoch": 7.119581029967995, + "flow_matching_loss": 0.01411473285406828, + "grad_norm": 0.14287251234054565, + "learning_rate": 2.066837724320179e-05, + "loss": 0.0224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005166928842663765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00247889383463189, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012812751531600951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6428883135318756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24470, + "weighted_flow_matching_loss": 0.01411473285406828, + "weighted_rkd_loss": 0.008237110450863838 + }, + { + "epoch": 7.122490544079139, + "flow_matching_loss": 0.015019430872052908, + "grad_norm": 0.1565748006105423, + "learning_rate": 2.0660722059481066e-05, + "loss": 0.0226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004881430277600884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020477780839428306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011810638755559922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6426374018192291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24480, + "weighted_flow_matching_loss": 0.015019430872052908, + "weighted_rkd_loss": 0.007589898817241192 + }, + { + "epoch": 7.125400058190282, + "flow_matching_loss": 0.01163440877571702, + "grad_norm": 0.12791456282138824, + "learning_rate": 2.065306515618244e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043937461217865344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018854203633964061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010672912560403347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6423864662647247, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24490, + "weighted_flow_matching_loss": 0.01163440877571702, + "weighted_rkd_loss": 0.006856202054768801 + }, + { + "epoch": 7.128309572301426, + "flow_matching_loss": 0.013033071346580983, + "grad_norm": 0.12569467723369598, + "learning_rate": 2.064540653563188e-05, + "loss": 0.0215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005320399370975793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002487133676186204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013127932511270046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.64213547706604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24500, + "weighted_flow_matching_loss": 0.013033071346580983, + "weighted_rkd_loss": 0.008429908193647861 + }, + { + "epoch": 7.131219086412569, + "flow_matching_loss": 0.012813015282154084, + "grad_norm": 0.1183777004480362, + "learning_rate": 2.063774620015587e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004990015760995448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020997923333197833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012079823855310678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6418844521045685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24510, + "weighted_flow_matching_loss": 0.012813015282154084, + "weighted_rkd_loss": 0.007753808470442891 + }, + { + "epoch": 7.134128600523712, + "flow_matching_loss": 0.012437147554010153, + "grad_norm": 0.13237538933753967, + "learning_rate": 2.0630084152081422e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004104531463235616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018527269014157355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010061789862811565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6416333913803101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24520, + "weighted_flow_matching_loss": 0.012437147554010153, + "weighted_rkd_loss": 0.006456020288169384 + }, + { + "epoch": 7.137038114634856, + "flow_matching_loss": 0.01287862341850996, + "grad_norm": 0.14770740270614624, + "learning_rate": 2.0622420393736065e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004425787832587957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018263520440086722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010677927546203137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6413822948932648, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24530, + "weighted_flow_matching_loss": 0.01287862341850996, + "weighted_rkd_loss": 0.006848586723208427 + }, + { + "epoch": 7.1399476287459995, + "flow_matching_loss": 0.01247053206898272, + "grad_norm": 0.1546698659658432, + "learning_rate": 2.061475492744784e-05, + "loss": 0.0209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005410195654258132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022688340046443044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013089225580915808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6411311566829682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24540, + "weighted_flow_matching_loss": 0.01247053206898272, + "weighted_rkd_loss": 0.008392017195001244 + }, + { + "epoch": 7.142857142857143, + "flow_matching_loss": 0.013208959810435772, + "grad_norm": 0.11945304274559021, + "learning_rate": 2.060708775554532e-05, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038480335380882023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017563072382472456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009452374186366796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6408799767494202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24550, + "weighted_flow_matching_loss": 0.013208959810435772, + "weighted_rkd_loss": 0.0060578694101423025 + }, + { + "epoch": 7.145766656968286, + "flow_matching_loss": 0.012803385686129331, + "grad_norm": 0.14514900743961334, + "learning_rate": 2.0599418880357586e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044744965387508275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021627456415444614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011111738672479988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6406287550926208, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24560, + "weighted_flow_matching_loss": 0.012803385686129331, + "weighted_rkd_loss": 0.007118606800213456 + }, + { + "epoch": 7.14867617107943, + "flow_matching_loss": 0.013335394114255905, + "grad_norm": 0.1297364979982376, + "learning_rate": 2.059174830421424e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004150403221137822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017726359190419316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01007344238460064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6403775095939637, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24570, + "weighted_flow_matching_loss": 0.013335394114255905, + "weighted_rkd_loss": 0.0064508512150496244 + }, + { + "epoch": 7.1515856851905735, + "flow_matching_loss": 0.012211885582655669, + "grad_norm": 0.11363917589187622, + "learning_rate": 2.05840760294454e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004628091189078987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002055594779085368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01131177693605423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.640126210451126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24580, + "weighted_flow_matching_loss": 0.012211885582655669, + "weighted_rkd_loss": 0.007241013552993536 + }, + { + "epoch": 7.154495199301716, + "flow_matching_loss": 0.013304139021784067, + "grad_norm": 0.15190523862838745, + "learning_rate": 2.05764020583817e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004361595306545496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019410060485824943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010664196778088808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6398748815059662, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24590, + "weighted_flow_matching_loss": 0.013304139021784067, + "weighted_rkd_loss": 0.006823670724406839 + }, + { + "epoch": 7.15740471341286, + "flow_matching_loss": 0.012498363107442855, + "grad_norm": 0.1428905427455902, + "learning_rate": 2.056872639335429e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004306670511141419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018075674888677895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010420908406376838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6396235167980194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24600, + "weighted_flow_matching_loss": 0.012498363107442855, + "weighted_rkd_loss": 0.006665504351258278 + }, + { + "epoch": 7.160314227524004, + "flow_matching_loss": 0.013369633164256812, + "grad_norm": 0.14917925000190735, + "learning_rate": 2.056104903669484e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004026993503794074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017698636278510095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009823850682005286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6393721044063568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24610, + "weighted_flow_matching_loss": 0.013369633164256812, + "weighted_rkd_loss": 0.006281106546521187 + }, + { + "epoch": 7.1632237416351465, + "flow_matching_loss": 0.012764558475464583, + "grad_norm": 0.12256036698818207, + "learning_rate": 2.055336999073552e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035364709328860044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015355851268395782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008608526922762394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6391206622123718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24620, + "weighted_flow_matching_loss": 0.012764558475464583, + "weighted_rkd_loss": 0.00550191467627883 + }, + { + "epoch": 7.16613325574629, + "flow_matching_loss": 0.011174832982942463, + "grad_norm": 0.1595601886510849, + "learning_rate": 2.0545689257809028e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043309335596859455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019687250955030323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010630592331290245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6388691663742065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24630, + "weighted_flow_matching_loss": 0.011174832982942463, + "weighted_rkd_loss": 0.006791571155190468 + }, + { + "epoch": 7.169042769857434, + "flow_matching_loss": 0.01365718776360154, + "grad_norm": 0.13472753763198853, + "learning_rate": 2.053800684024856e-05, + "loss": 0.0212, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004879263555631041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020706930547021328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011829220270738006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6386176347732544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24640, + "weighted_flow_matching_loss": 0.01365718776360154, + "weighted_rkd_loss": 0.007554414356127382 + }, + { + "epoch": 7.171952283968578, + "flow_matching_loss": 0.012337521556764841, + "grad_norm": 0.12863843142986298, + "learning_rate": 2.0530322740387842e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004744809190742672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002038304565940052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011527922935783863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6383660972118378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24650, + "weighted_flow_matching_loss": 0.012337521556764841, + "weighted_rkd_loss": 0.007358975498937071 + }, + { + "epoch": 7.1748617980797205, + "flow_matching_loss": 0.012256862316280603, + "grad_norm": 0.11108668148517609, + "learning_rate": 2.0522636960561096e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004470772040076554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019727800972759724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010914324223995209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6381144881248474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24660, + "weighted_flow_matching_loss": 0.012256862316280603, + "weighted_rkd_loss": 0.006964644929394126 + }, + { + "epoch": 7.177771312190864, + "flow_matching_loss": 0.012520497944206, + "grad_norm": 0.14078620076179504, + "learning_rate": 2.0514949503103056e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004457718716003001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001822619209997356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010738056618720293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6378628551959992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24670, + "weighted_flow_matching_loss": 0.012520497944206, + "weighted_rkd_loss": 0.0068493586033582686 + }, + { + "epoch": 7.180680826302008, + "flow_matching_loss": 0.013235676288604736, + "grad_norm": 0.16751891374588013, + "learning_rate": 2.0507260370348968e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004622546583414077, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021490964689292015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01139418976381421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6376111805438995, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24680, + "weighted_flow_matching_loss": 0.013235676288604736, + "weighted_rkd_loss": 0.0072648950386792425 + }, + { + "epoch": 7.183590340413151, + "flow_matching_loss": 0.011913091875612736, + "grad_norm": 0.0917496383190155, + "learning_rate": 2.049956956463459e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037379555171355603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015792430727742611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009055154072120786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6373594701290131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24690, + "weighted_flow_matching_loss": 0.011913091875612736, + "weighted_rkd_loss": 0.005771382478997111 + }, + { + "epoch": 7.186499854524294, + "flow_matching_loss": 0.011498002568259835, + "grad_norm": 0.1304822415113449, + "learning_rate": 2.049187708829619e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004026153939776123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017830410157330333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009835348883643746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6371077179908753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24700, + "weighted_flow_matching_loss": 0.011498002568259835, + "weighted_rkd_loss": 0.0062661631498485805 + }, + { + "epoch": 7.189409368635438, + "flow_matching_loss": 0.01289404509589076, + "grad_norm": 0.1509341448545456, + "learning_rate": 2.0484182943670537e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004773681238293648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002055753709282726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011603116244077682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6368559300899506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24710, + "weighted_flow_matching_loss": 0.01289404509589076, + "weighted_rkd_loss": 0.007389472937211395 + }, + { + "epoch": 7.192318882746581, + "flow_matching_loss": 0.012486349418759346, + "grad_norm": 0.14913292229175568, + "learning_rate": 2.04764871330949e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003997290623374283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019301048596389592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009924686141312122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6366041004657745, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24720, + "weighted_flow_matching_loss": 0.012486349418759346, + "weighted_rkd_loss": 0.006318058026954532 + }, + { + "epoch": 7.195228396857725, + "flow_matching_loss": 0.01156665598973632, + "grad_norm": 0.12003912031650543, + "learning_rate": 2.0468789658907075e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004425288550555706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019575826125219466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010808159783482551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6363522469997406, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24730, + "weighted_flow_matching_loss": 0.01156665598973632, + "weighted_rkd_loss": 0.006877798633649945 + }, + { + "epoch": 7.198137910968868, + "flow_matching_loss": 0.012474592123180629, + "grad_norm": 0.1951625943183899, + "learning_rate": 2.0461090523445344e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045444602379575375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019428594969213009, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011031780019402504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6361003518104553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24740, + "weighted_flow_matching_loss": 0.012474592123180629, + "weighted_rkd_loss": 0.007017275039106608 + }, + { + "epoch": 7.201047425080012, + "flow_matching_loss": 0.013394192839041353, + "grad_norm": 0.1333392858505249, + "learning_rate": 2.04533897290485e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042660852894186975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018283357261680067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010360506363213062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6358484148979187, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24750, + "weighted_flow_matching_loss": 0.013394192839041353, + "weighted_rkd_loss": 0.006587812164798379 + }, + { + "epoch": 7.203956939191155, + "flow_matching_loss": 0.011564562842249871, + "grad_norm": 0.15560321509838104, + "learning_rate": 2.0445687278055852e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042130057932809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001806507131550461, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010232518706470729, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6355964362621307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24760, + "weighted_flow_matching_loss": 0.011564562842249871, + "weighted_rkd_loss": 0.006503784796223044 + }, + { + "epoch": 7.206866453302299, + "flow_matching_loss": 0.012585264630615711, + "grad_norm": 0.1263088434934616, + "learning_rate": 2.0437983172807196e-05, + "loss": 0.0208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00535839288495481, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022863846854306756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01300317021086812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6353444278240203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24770, + "weighted_flow_matching_loss": 0.012585264630615711, + "weighted_rkd_loss": 0.008261559577658773 + }, + { + "epoch": 7.209775967413442, + "flow_matching_loss": 0.011784426122903823, + "grad_norm": 0.12560565769672394, + "learning_rate": 2.043027741564283e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004695344809442758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002300289459526539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011690978892147541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6350923776626587, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24780, + "weighted_flow_matching_loss": 0.011784426122903823, + "weighted_rkd_loss": 0.0074249012395739555 + }, + { + "epoch": 7.212685481524585, + "flow_matching_loss": 0.013244502898305654, + "grad_norm": 0.13373087346553802, + "learning_rate": 2.042257000890356e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003881104663014412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016491468995809555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00941135622560978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6348402798175812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24790, + "weighted_flow_matching_loss": 0.013244502898305654, + "weighted_rkd_loss": 0.005974683770909905 + }, + { + "epoch": 7.215594995635729, + "flow_matching_loss": 0.011422712728381157, + "grad_norm": 0.1396796554327011, + "learning_rate": 2.0414860954930698e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004570008395239711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019341651699505747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0110741819255054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6345881521701813, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24800, + "weighted_flow_matching_loss": 0.011422712728381157, + "weighted_rkd_loss": 0.007027528109028935 + }, + { + "epoch": 7.2185045097468725, + "flow_matching_loss": 0.01185210756957531, + "grad_norm": 0.13467274606227875, + "learning_rate": 2.0407150256066052e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004077037749812007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017012782278470694, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009855353645980358, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6343360126018525, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24810, + "weighted_flow_matching_loss": 0.01185210756957531, + "weighted_rkd_loss": 0.006251574447378516 + }, + { + "epoch": 7.221414023858015, + "flow_matching_loss": 0.011962781380861997, + "grad_norm": 0.13151919841766357, + "learning_rate": 2.0399437914651927e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003897765930742025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001732085517141968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009527617320418357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6340838134288788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24820, + "weighted_flow_matching_loss": 0.011962781380861997, + "weighted_rkd_loss": 0.006041294848546386 + }, + { + "epoch": 7.224323537969159, + "flow_matching_loss": 0.011887172143906355, + "grad_norm": 0.10549068450927734, + "learning_rate": 2.039172393303112e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004309557541273534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001865605521015823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01048472048714757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6338315844535828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24830, + "weighted_flow_matching_loss": 0.011887172143906355, + "weighted_rkd_loss": 0.006645545549690724 + }, + { + "epoch": 7.227233052080303, + "flow_matching_loss": 0.011977085750550032, + "grad_norm": 0.12925763428211212, + "learning_rate": 2.0384008313546947e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004061683127656579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017348141875118018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009858180489391088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6335793137550354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24840, + "weighted_flow_matching_loss": 0.011977085750550032, + "weighted_rkd_loss": 0.0062458519358187916 + }, + { + "epoch": 7.2301425661914465, + "flow_matching_loss": 0.013273516017943621, + "grad_norm": 0.11318987607955933, + "learning_rate": 2.0376291058543206e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004353553405962884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018398978281766176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010547004640102386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6333270072937012, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24850, + "weighted_flow_matching_loss": 0.013273516017943621, + "weighted_rkd_loss": 0.0066797020845115185 + }, + { + "epoch": 7.233052080302589, + "flow_matching_loss": 0.012023564195260406, + "grad_norm": 0.12369142472743988, + "learning_rate": 2.036857217036419e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004414695943705737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018122387817129492, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010641630785539747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6330746591091156, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24860, + "weighted_flow_matching_loss": 0.012023564195260406, + "weighted_rkd_loss": 0.006737079657614231 + }, + { + "epoch": 7.235961594413733, + "flow_matching_loss": 0.012353829387575389, + "grad_norm": 0.122984878718853, + "learning_rate": 2.0360851651354702e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00413334087934345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001778403588104993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01004508538171649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6328222870826721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24870, + "weighted_flow_matching_loss": 0.012353829387575389, + "weighted_rkd_loss": 0.006356750521808863 + }, + { + "epoch": 7.238871108524877, + "flow_matching_loss": 0.013944001682102681, + "grad_norm": 0.12757553160190582, + "learning_rate": 2.0353129503860025e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004198811971582473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019089499954134225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010306573938578367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6325698733329773, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24880, + "weighted_flow_matching_loss": 0.013944001682102681, + "weighted_rkd_loss": 0.006519638607278467 + }, + { + "epoch": 7.2417806226360195, + "flow_matching_loss": 0.012505447026342154, + "grad_norm": 0.12135352194309235, + "learning_rate": 2.034540573022595e-05, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004439963982440532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019380717421881855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010817999579012394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6323174178600312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24890, + "weighted_flow_matching_loss": 0.012505447026342154, + "weighted_rkd_loss": 0.0068404823541641235 + }, + { + "epoch": 7.244690136747163, + "flow_matching_loss": 0.01289018765091896, + "grad_norm": 0.11153358966112137, + "learning_rate": 2.0337680332798742e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004321746248751879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018351940205320715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010478686727583408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6320649266242981, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24900, + "weighted_flow_matching_loss": 0.01289018765091896, + "weighted_rkd_loss": 0.00662324926815927 + }, + { + "epoch": 7.247599650858307, + "flow_matching_loss": 0.012773636355996132, + "grad_norm": 0.1331503838300705, + "learning_rate": 2.0329953313925188e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004450144735164941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018167717149481176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010717061208561062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6318123996257782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24910, + "weighted_flow_matching_loss": 0.012773636355996132, + "weighted_rkd_loss": 0.0067712010350078344 + }, + { + "epoch": 7.25050916496945, + "flow_matching_loss": 0.011804584600031375, + "grad_norm": 0.14113394916057587, + "learning_rate": 2.032222467595255e-05, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004813791741617024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021942552528344096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011821838747709989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6315598487854004, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24920, + "weighted_flow_matching_loss": 0.011804584600031375, + "weighted_rkd_loss": 0.007466208562254906 + }, + { + "epoch": 7.2534186790805935, + "flow_matching_loss": 0.01344399256631732, + "grad_norm": 0.13746392726898193, + "learning_rate": 2.0314494421228577e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004196759150363505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001785870303865522, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010179388709366322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6313072443008423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24930, + "weighted_flow_matching_loss": 0.01344399256631732, + "weighted_rkd_loss": 0.006426302948966622 + }, + { + "epoch": 7.256328193191737, + "flow_matching_loss": 0.012900721374899149, + "grad_norm": 0.1379203349351883, + "learning_rate": 2.0306762552101516e-05, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004251863830722869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016942971968092024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010198024846613408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6310546159744262, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24940, + "weighted_flow_matching_loss": 0.012900721374899149, + "weighted_rkd_loss": 0.006435567792505026 + }, + { + "epoch": 7.25923770730288, + "flow_matching_loss": 0.011795972567051649, + "grad_norm": 0.12625926733016968, + "learning_rate": 2.0299029070920113e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004353955644182861, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019966715131886303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01070458274334669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6308019459247589, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24950, + "weighted_flow_matching_loss": 0.011795972567051649, + "weighted_rkd_loss": 0.006752544222399593 + }, + { + "epoch": 7.262147221414024, + "flow_matching_loss": 0.010946775274351239, + "grad_norm": 0.13679619133472443, + "learning_rate": 2.0291293980033594e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004724359209649265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019792283535934986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011427946854382753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6305492520332336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24960, + "weighted_flow_matching_loss": 0.010946775274351239, + "weighted_rkd_loss": 0.007205787161365151 + }, + { + "epoch": 7.265056735525167, + "flow_matching_loss": 0.012263369467109441, + "grad_norm": 0.16112668812274933, + "learning_rate": 2.028355728179167e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004140440630726517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017196258995682001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010000507114455104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.630296504497528, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24970, + "weighted_flow_matching_loss": 0.012263369467109441, + "weighted_rkd_loss": 0.0063032802194356915 + }, + { + "epoch": 7.267966249636311, + "flow_matching_loss": 0.013098040781915187, + "grad_norm": 0.12579625844955444, + "learning_rate": 2.0275818978544546e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043127616634592416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019166674348525703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010542190726846456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6300437211990356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24980, + "weighted_flow_matching_loss": 0.013098040781915187, + "weighted_rkd_loss": 0.006642088154330849 + }, + { + "epoch": 7.270875763747454, + "flow_matching_loss": 0.01189156761392951, + "grad_norm": 0.14638811349868774, + "learning_rate": 2.026807907264292e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004792471695691347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002045190008357167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011630133260041475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6297909140586853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24990, + "weighted_flow_matching_loss": 0.01189156761392951, + "weighted_rkd_loss": 0.007324494328349829 + }, + { + "epoch": 7.273785277858598, + "flow_matching_loss": 0.013116703927516937, + "grad_norm": 0.1525396704673767, + "learning_rate": 2.026033756643796e-05, + "loss": 0.0206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004896569903939963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021157053648494183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011908845324069262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6295380651950836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25000, + "weighted_flow_matching_loss": 0.013116703927516937, + "weighted_rkd_loss": 0.007497105421498418 + }, + { + "epoch": 7.276694791969741, + "flow_matching_loss": 0.012903978489339352, + "grad_norm": 0.13228505849838257, + "learning_rate": 2.0252594462281347e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0047146814409643415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002169351396150887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011598714347928762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6292851805686951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25010, + "weighted_flow_matching_loss": 0.012903978489339352, + "weighted_rkd_loss": 0.007298825774341822 + }, + { + "epoch": 7.279604306080884, + "flow_matching_loss": 0.012696878239512443, + "grad_norm": 0.13368913531303406, + "learning_rate": 2.024484976252522e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004055695421993732, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017060076002962886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00981739852577448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6290322661399841, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25020, + "weighted_flow_matching_loss": 0.012696878239512443, + "weighted_rkd_loss": 0.006175444554537534 + }, + { + "epoch": 7.282513820192028, + "flow_matching_loss": 0.012464439310133458, + "grad_norm": 0.10613605380058289, + "learning_rate": 2.0237103469522212e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004838092508725822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021723796264268456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0118485645391047, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6287793099880219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25030, + "weighted_flow_matching_loss": 0.012464439310133458, + "weighted_rkd_loss": 0.0074501476716250185 + }, + { + "epoch": 7.285423334303172, + "flow_matching_loss": 0.013818078581243754, + "grad_norm": 0.16021524369716644, + "learning_rate": 2.022935558562545e-05, + "loss": 0.0207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004474724736064672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019992832327261566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010948732728138566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6285263299942017, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25040, + "weighted_flow_matching_loss": 0.013818078581243754, + "weighted_rkd_loss": 0.006881581572815776 + }, + { + "epoch": 7.288332848414315, + "flow_matching_loss": 0.011934074759483337, + "grad_norm": 0.1278667449951172, + "learning_rate": 2.0221606113188538e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040452485671266915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017433986766263842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00983389588072896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6282733023166657, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25050, + "weighted_flow_matching_loss": 0.011934074759483337, + "weighted_rkd_loss": 0.006178375612944364 + }, + { + "epoch": 7.291242362525458, + "flow_matching_loss": 0.012678363732993603, + "grad_norm": 0.1075155958533287, + "learning_rate": 2.0213855054565554e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004480705852620304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019147112150676548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010876122862100601, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6280202388763427, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25060, + "weighted_flow_matching_loss": 0.012678363732993603, + "weighted_rkd_loss": 0.006830561626702547 + }, + { + "epoch": 7.294151876636602, + "flow_matching_loss": 0.01364561663940549, + "grad_norm": 0.11400751769542694, + "learning_rate": 2.020610241211107e-05, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004399938858114183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019319977378472685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0107318754773587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6277671456336975, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25070, + "weighted_flow_matching_loss": 0.01364561663940549, + "weighted_rkd_loss": 0.0067371685523539785 + }, + { + "epoch": 7.2970613907477455, + "flow_matching_loss": 0.01254284530878067, + "grad_norm": 0.12455922365188599, + "learning_rate": 2.0198348188180137e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004945708438754082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022646156954579054, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012156032398343087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6275140166282653, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25080, + "weighted_flow_matching_loss": 0.01254284530878067, + "weighted_rkd_loss": 0.007628194475546479 + }, + { + "epoch": 7.299970904858888, + "flow_matching_loss": 0.01211564764380455, + "grad_norm": 0.11560799181461334, + "learning_rate": 2.0190592385128274e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004110135091468692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017822613823227583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010002531623467804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6272608458995819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25090, + "weighted_flow_matching_loss": 0.01211564764380455, + "weighted_rkd_loss": 0.006274264818057418 + }, + { + "epoch": 7.302880418970032, + "flow_matching_loss": 0.011003835406154395, + "grad_norm": 0.12274054437875748, + "learning_rate": 2.018283500531149e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005149370385333895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023679287522099914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012666669394820928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6270076513290406, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25100, + "weighted_flow_matching_loss": 0.011003835406154395, + "weighted_rkd_loss": 0.007942140521481632 + }, + { + "epoch": 7.305789933081176, + "flow_matching_loss": 0.012749510118737817, + "grad_norm": 0.13993729650974274, + "learning_rate": 2.0175076051086282e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004131048475392163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016187974368222057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009880894515663385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6267544150352478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25110, + "weighted_flow_matching_loss": 0.012749510118737817, + "weighted_rkd_loss": 0.006192873977124691 + }, + { + "epoch": 7.308699447192319, + "flow_matching_loss": 0.012532578548416495, + "grad_norm": 0.13552287220954895, + "learning_rate": 2.016731552480961e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004180055367760361, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015596414916217328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009919752180576325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6265011548995971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25120, + "weighted_flow_matching_loss": 0.012532578548416495, + "weighted_rkd_loss": 0.006214783573523164 + }, + { + "epoch": 7.311608961303462, + "flow_matching_loss": 0.011457303538918495, + "grad_norm": 0.13412249088287354, + "learning_rate": 2.0159553428838914e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003913477086462081, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016829525120556355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009509906731545925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6262478530406952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25130, + "weighted_flow_matching_loss": 0.011457303538918495, + "weighted_rkd_loss": 0.005955516127869487 + }, + { + "epoch": 7.314518475414606, + "flow_matching_loss": 0.011998730711638927, + "grad_norm": 0.1194634959101677, + "learning_rate": 2.0151789765532117e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004234484536573291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001738027366809547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01020699655637145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6259945094585418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25140, + "weighted_flow_matching_loss": 0.011998730711638927, + "weighted_rkd_loss": 0.006389533448964357 + }, + { + "epoch": 7.317427989525749, + "flow_matching_loss": 0.0115048598498106, + "grad_norm": 0.12015850841999054, + "learning_rate": 2.0144024537247612e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004208382987417281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001873003551736474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010289769619703293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6257411360740661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25150, + "weighted_flow_matching_loss": 0.0115048598498106, + "weighted_rkd_loss": 0.006438733730465174 + }, + { + "epoch": 7.3203375036368925, + "flow_matching_loss": 0.012409427994862198, + "grad_norm": 0.13193096220493317, + "learning_rate": 2.013625774634428e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004117266298271716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018328010686673225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010067333560436964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6254877269268035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25160, + "weighted_flow_matching_loss": 0.012409427994862198, + "weighted_rkd_loss": 0.006296870112419128 + }, + { + "epoch": 7.323247017748036, + "flow_matching_loss": 0.012124317139387131, + "grad_norm": 0.12991076707839966, + "learning_rate": 2.0128489395181455e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003847072320058942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016140162362717092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009308160981163383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6252342879772186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25170, + "weighted_flow_matching_loss": 0.012124317139387131, + "weighted_rkd_loss": 0.005819775955751538 + }, + { + "epoch": 7.32615653185918, + "flow_matching_loss": 0.013420004025101661, + "grad_norm": 0.12710779905319214, + "learning_rate": 2.0120719486118958e-05, + "loss": 0.0199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004244408919475973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001800684630870819, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010289502469822765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6249808132648468, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25180, + "weighted_flow_matching_loss": 0.013420004025101661, + "weighted_rkd_loss": 0.00643063304014504 + }, + { + "epoch": 7.329066045970323, + "flow_matching_loss": 0.01156319072470069, + "grad_norm": 0.1063341572880745, + "learning_rate": 2.0112948021517093e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004066893155686558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019398258067667485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010073612071573734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6247273027896881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25190, + "weighted_flow_matching_loss": 0.01156319072470069, + "weighted_rkd_loss": 0.0062933462671935555 + }, + { + "epoch": 7.3319755600814664, + "flow_matching_loss": 0.014608677383512259, + "grad_norm": 0.13526342809200287, + "learning_rate": 2.0105175003736615e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004399842885322869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018472871743142605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010646972898393869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6244737565517425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25200, + "weighted_flow_matching_loss": 0.014608677383512259, + "weighted_rkd_loss": 0.006648867996409536 + }, + { + "epoch": 7.33488507419261, + "flow_matching_loss": 0.01353730782866478, + "grad_norm": 0.14206601679325104, + "learning_rate": 2.0097400435138764e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004053551098331809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001733650139067322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009840752370655537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6242201805114747, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25210, + "weighted_flow_matching_loss": 0.01353730782866478, + "weighted_rkd_loss": 0.0061428412329405544 + }, + { + "epoch": 7.337794588303753, + "flow_matching_loss": 0.010858991090208292, + "grad_norm": 0.10622250288724899, + "learning_rate": 2.0089624318085253e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004131660936400295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017014011973515153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00996472304686904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6239665746688843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25220, + "weighted_flow_matching_loss": 0.010858991090208292, + "weighted_rkd_loss": 0.0062176947947591545 + }, + { + "epoch": 7.340704102414897, + "flow_matching_loss": 0.01194399930536747, + "grad_norm": 0.11764594167470932, + "learning_rate": 2.0081846654938256e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043694269843399525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018722448963671922, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010611098911613225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6237129151821137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25230, + "weighted_flow_matching_loss": 0.01194399930536747, + "weighted_rkd_loss": 0.006618302222341299 + }, + { + "epoch": 7.34361361652604, + "flow_matching_loss": 0.012075778562575579, + "grad_norm": 0.12148793041706085, + "learning_rate": 2.0074067448060423e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038743879878893493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001748334290459752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009497110266238451, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6234592437744141, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25240, + "weighted_flow_matching_loss": 0.012075778562575579, + "weighted_rkd_loss": 0.005921079497784376 + }, + { + "epoch": 7.346523130637183, + "flow_matching_loss": 0.012170008523389697, + "grad_norm": 0.12929072976112366, + "learning_rate": 2.0066286699814873e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003990759537555277, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015926268650218844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009574145870283246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6232055246829986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25250, + "weighted_flow_matching_loss": 0.012170008523389697, + "weighted_rkd_loss": 0.005966653116047382 + }, + { + "epoch": 7.349432644748327, + "flow_matching_loss": 0.01102059306576848, + "grad_norm": 0.13095709681510925, + "learning_rate": 2.005850441256519e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003969582566060126, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016113253659568727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009550490509718656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6229517757892609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25260, + "weighted_flow_matching_loss": 0.01102059306576848, + "weighted_rkd_loss": 0.005949503695592284 + }, + { + "epoch": 7.352342158859471, + "flow_matching_loss": 0.01274725878611207, + "grad_norm": 0.11510150879621506, + "learning_rate": 2.0050720588675432e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040471687680110335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016770403133705258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009771377872675657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6226979911327362, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25270, + "weighted_flow_matching_loss": 0.01274725878611207, + "weighted_rkd_loss": 0.006084616808220744 + }, + { + "epoch": 7.355251672970613, + "flow_matching_loss": 0.014059070125222206, + "grad_norm": 0.12235251814126968, + "learning_rate": 2.0042935230510113e-05, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004159449064172804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018851090921089054, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010204007010906934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6224441826343536, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25280, + "weighted_flow_matching_loss": 0.014059070125222206, + "weighted_rkd_loss": 0.006351461121812463 + }, + { + "epoch": 7.358161187081757, + "flow_matching_loss": 0.01276638898998499, + "grad_norm": 0.12384837865829468, + "learning_rate": 2.0035148340434222e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037219737423583865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016513357986696064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009095283271744848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6221903324127197, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25290, + "weighted_flow_matching_loss": 0.01276638898998499, + "weighted_rkd_loss": 0.00565898441709578 + }, + { + "epoch": 7.361070701192901, + "flow_matching_loss": 0.011980704125016927, + "grad_norm": 0.10947676748037338, + "learning_rate": 2.0027359920813212e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003964462364092469, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001716549228876829, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009645474050194025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6219364404678345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25300, + "weighted_flow_matching_loss": 0.011980704125016927, + "weighted_rkd_loss": 0.0059988399036228655 + }, + { + "epoch": 7.363980215304045, + "flow_matching_loss": 0.012314752489328385, + "grad_norm": 0.13002242147922516, + "learning_rate": 2.0019569974012994e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038307287031784654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015648563159629703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009226313699036837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6216825246810913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25310, + "weighted_flow_matching_loss": 0.012314752489328385, + "weighted_rkd_loss": 0.005735727632418275 + }, + { + "epoch": 7.366889729415187, + "flow_matching_loss": 0.011402997840195894, + "grad_norm": 0.13207340240478516, + "learning_rate": 2.0011778502399954e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004184739687480033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001726443972438574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010095923487097026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6214285850524902, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25320, + "weighted_flow_matching_loss": 0.011402997840195894, + "weighted_rkd_loss": 0.006273924745619297 + }, + { + "epoch": 7.369799243526331, + "flow_matching_loss": 0.013347298838198185, + "grad_norm": 0.11791286617517471, + "learning_rate": 2.0003985508340937e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004047956154681742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001675972668454051, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009771884931251407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6211746037006378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25330, + "weighted_flow_matching_loss": 0.013347298838198185, + "weighted_rkd_loss": 0.006070023262873292 + }, + { + "epoch": 7.372708757637475, + "flow_matching_loss": 0.012048515398055315, + "grad_norm": 0.16235846281051636, + "learning_rate": 1.9996190994203244e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00412556582596153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001722561230417341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009973692893981933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.620920580625534, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25340, + "weighted_flow_matching_loss": 0.012048515398055315, + "weighted_rkd_loss": 0.006192933768033981 + }, + { + "epoch": 7.375618271748618, + "flow_matching_loss": 0.012938545644283294, + "grad_norm": 0.13894417881965637, + "learning_rate": 1.998839496235464e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004561077384278178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019263477646745742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011048502521589398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6206665337085724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25350, + "weighted_flow_matching_loss": 0.012938545644283294, + "weighted_rkd_loss": 0.0068574146367609504 + }, + { + "epoch": 7.378527785859761, + "flow_matching_loss": 0.014671104587614536, + "grad_norm": 0.1349349319934845, + "learning_rate": 1.9980597415163358e-05, + "loss": 0.0213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004384175734594464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018638519919477402, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01063220351934433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6204124569892884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25360, + "weighted_flow_matching_loss": 0.014671104587614536, + "weighted_rkd_loss": 0.006596311833709479 + }, + { + "epoch": 7.381437299970905, + "flow_matching_loss": 0.014051208831369877, + "grad_norm": 0.10721634328365326, + "learning_rate": 1.997279835499809e-05, + "loss": 0.0205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004311315203085542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018226968590170146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010445327078923584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.620158338546753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25370, + "weighted_flow_matching_loss": 0.014051208831369877, + "weighted_rkd_loss": 0.0064778659958392385 + }, + { + "epoch": 7.384346814082049, + "flow_matching_loss": 0.011987074557691813, + "grad_norm": 0.11645661294460297, + "learning_rate": 1.9964997784227976e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038982369005680084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017095305607654153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009506004210561515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6199041903018951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25380, + "weighted_flow_matching_loss": 0.011987074557691813, + "weighted_rkd_loss": 0.005892796348780394 + }, + { + "epoch": 7.387256328193192, + "flow_matching_loss": 0.012210395699366928, + "grad_norm": 0.1632527858018875, + "learning_rate": 1.9957195705222633e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004433365049771964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001867633662186563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010734363785013556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6196500062942505, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25390, + "weighted_flow_matching_loss": 0.012210395699366928, + "weighted_rkd_loss": 0.0066514815203845504 + }, + { + "epoch": 7.390165842304335, + "flow_matching_loss": 0.01291259452700615, + "grad_norm": 0.12865467369556427, + "learning_rate": 1.994939212035212e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004133668728172779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018941626301966607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010161499958485365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6193957924842834, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25400, + "weighted_flow_matching_loss": 0.01291259452700615, + "weighted_rkd_loss": 0.006293912651017308 + }, + { + "epoch": 7.393075356415479, + "flow_matching_loss": 0.011955923028290271, + "grad_norm": 0.10850270092487335, + "learning_rate": 1.9941587031986963e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00411038629245013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017589690862223506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009979741694405676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.619141548871994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25410, + "weighted_flow_matching_loss": 0.011955923028290271, + "weighted_rkd_loss": 0.006178881274536252 + }, + { + "epoch": 7.395984870526622, + "flow_matching_loss": 0.011814377596601843, + "grad_norm": 0.12478096038103104, + "learning_rate": 1.9933780442498137e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004108548606745898, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017518050968647002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009968902356922627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6188872694969177, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25420, + "weighted_flow_matching_loss": 0.011814377596601843, + "weighted_rkd_loss": 0.006169617781415582 + }, + { + "epoch": 7.3988943846377655, + "flow_matching_loss": 0.014155354490503669, + "grad_norm": 0.1404275894165039, + "learning_rate": 1.9925972354257085e-05, + "loss": 0.0214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004807522054761648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020902484538964927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01170529257506132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6186329603195191, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25430, + "weighted_flow_matching_loss": 0.014155354490503669, + "weighted_rkd_loss": 0.007241201540455222 + }, + { + "epoch": 7.401803898748909, + "flow_matching_loss": 0.012638483382761478, + "grad_norm": 0.12927581369876862, + "learning_rate": 1.9918162769635696e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0047403828939422965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002008602360729128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011489368136972189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.618378609418869, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25440, + "weighted_flow_matching_loss": 0.012638483382761478, + "weighted_rkd_loss": 0.007104897731915116 + }, + { + "epoch": 7.404713412860052, + "flow_matching_loss": 0.01188212214037776, + "grad_norm": 0.12840090692043304, + "learning_rate": 1.9910351691006308e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004901204514317215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020830130204558374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011885422188788653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6181242346763611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25450, + "weighted_flow_matching_loss": 0.01188212214037776, + "weighted_rkd_loss": 0.0073465330991894 + }, + { + "epoch": 7.407622926971196, + "flow_matching_loss": 0.01277283551171422, + "grad_norm": 0.12994714081287384, + "learning_rate": 1.990253912074173e-05, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00449911595787853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020235198666341604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01102175167761743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6178698301315307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25460, + "weighted_flow_matching_loss": 0.01277283551171422, + "weighted_rkd_loss": 0.0068100470583885905 + }, + { + "epoch": 7.410532441082339, + "flow_matching_loss": 0.012799451034516096, + "grad_norm": 0.10389427095651627, + "learning_rate": 1.989472506121521e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004068577685393393, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016524705337360502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009789625834673642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6176153898239136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25470, + "weighted_flow_matching_loss": 0.012799451034516096, + "weighted_rkd_loss": 0.006046225223690272 + }, + { + "epoch": 7.413441955193482, + "flow_matching_loss": 0.010345002170652151, + "grad_norm": 0.14965742826461792, + "learning_rate": 1.9886909514800454e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004120247368700802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018198623904027045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010060357209295035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.617360919713974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25480, + "weighted_flow_matching_loss": 0.010345002170652151, + "weighted_rkd_loss": 0.006210936466231942 + }, + { + "epoch": 7.416351469304626, + "flow_matching_loss": 0.01429867846891284, + "grad_norm": 0.12396255880594254, + "learning_rate": 1.9879092483871622e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040061477106064555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017920246464200318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009804320149123668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6171064078807831, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25490, + "weighted_flow_matching_loss": 0.01429867846891284, + "weighted_rkd_loss": 0.0060503648594021795 + }, + { + "epoch": 7.41926098341577, + "flow_matching_loss": 0.013327144552022219, + "grad_norm": 0.14321163296699524, + "learning_rate": 1.987127397080332e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004403481027111411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019507393124513328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01075770128518343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6168518722057342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25500, + "weighted_flow_matching_loss": 0.013327144552022219, + "weighted_rkd_loss": 0.006635928573086858 + }, + { + "epoch": 7.422170497526913, + "flow_matching_loss": 0.010859923157840966, + "grad_norm": 0.12468979507684708, + "learning_rate": 1.98634539779706e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004068269580602646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019924948108382525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01012903400696814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6165973007678985, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25510, + "weighted_flow_matching_loss": 0.010859923157840966, + "weighted_rkd_loss": 0.006245499476790428 + }, + { + "epoch": 7.425080011638056, + "flow_matching_loss": 0.01311224652454257, + "grad_norm": 0.1308349221944809, + "learning_rate": 1.985563250774897e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00457368758507073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00199275683844462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011140132043510675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6163426995277405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25520, + "weighted_flow_matching_loss": 0.01311224652454257, + "weighted_rkd_loss": 0.006866243621334434 + }, + { + "epoch": 7.4279895257492, + "flow_matching_loss": 0.01225520707666874, + "grad_norm": 0.1455673724412918, + "learning_rate": 1.98478095625144e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004752193740569055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021946058725006877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011698993295431137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6160880625247955, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25530, + "weighted_flow_matching_loss": 0.01225520707666874, + "weighted_rkd_loss": 0.007207493437454105 + }, + { + "epoch": 7.430899039860344, + "flow_matching_loss": 0.011515861982479692, + "grad_norm": 0.156762033700943, + "learning_rate": 1.9839985144643275e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004267633799463511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017520583351142704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010287325922399759, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6158333957195282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25540, + "weighted_flow_matching_loss": 0.011515861982479692, + "weighted_rkd_loss": 0.006335188169032335 + }, + { + "epoch": 7.433808553971486, + "flow_matching_loss": 0.013203081116080284, + "grad_norm": 0.11972909420728683, + "learning_rate": 1.9832159256512463e-05, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004046509205363691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001749620062764734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009842638485133648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6155787110328674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25550, + "weighted_flow_matching_loss": 0.013203081116080284, + "weighted_rkd_loss": 0.006058867927640677 + }, + { + "epoch": 7.43671806808263, + "flow_matching_loss": 0.013050747383385897, + "grad_norm": 0.1268158257007599, + "learning_rate": 1.982433190049925e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004273301898501813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017414703383110464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010288074146956205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6153239727020263, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25560, + "weighted_flow_matching_loss": 0.013050747383385897, + "weighted_rkd_loss": 0.006330525269731879 + }, + { + "epoch": 7.439627582193774, + "flow_matching_loss": 0.011679868306964636, + "grad_norm": 0.1158350259065628, + "learning_rate": 1.981650307898138e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004302282864227891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018136578495614232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010418223636224865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6150692164897918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25570, + "weighted_flow_matching_loss": 0.011679868306964636, + "weighted_rkd_loss": 0.006408021133393049 + }, + { + "epoch": 7.442537096304917, + "flow_matching_loss": 0.013141502812504768, + "grad_norm": 0.1314978450536728, + "learning_rate": 1.980867279433705e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004177470505237579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018062513321638108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010161192342638969, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.614814430475235, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25580, + "weighted_flow_matching_loss": 0.013141502812504768, + "weighted_rkd_loss": 0.006247251573950052 + }, + { + "epoch": 7.44544661041606, + "flow_matching_loss": 0.012421852722764016, + "grad_norm": 0.14690427482128143, + "learning_rate": 1.98008410489449e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004301457665860653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002000446547754109, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01060336190275848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6145595967769623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25590, + "weighted_flow_matching_loss": 0.012421852722764016, + "weighted_rkd_loss": 0.006516284355893731 + }, + { + "epoch": 7.448356124527204, + "flow_matching_loss": 0.012214601598680019, + "grad_norm": 0.12144285440444946, + "learning_rate": 1.9793007845183987e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004381079645827413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018032798310741783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010565439006313682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6143047451972962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25600, + "weighted_flow_matching_loss": 0.012214601598680019, + "weighted_rkd_loss": 0.006490393378771842 + }, + { + "epoch": 7.451265638638348, + "flow_matching_loss": 0.011678053066134453, + "grad_norm": 0.17254957556724548, + "learning_rate": 1.9785173185433846e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003776481305249035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001666370127350092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00921933283098042, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6140498578548431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25610, + "weighted_flow_matching_loss": 0.011678053066134453, + "weighted_rkd_loss": 0.005661090859211982 + }, + { + "epoch": 7.454175152749491, + "flow_matching_loss": 0.012241774331778288, + "grad_norm": 0.15087978541851044, + "learning_rate": 1.977733707207444e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004237328446470201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018272928660735487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010301949642598628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6137949407100678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25620, + "weighted_flow_matching_loss": 0.012241774331778288, + "weighted_rkd_loss": 0.006323245819658041 + }, + { + "epoch": 7.457084666860634, + "flow_matching_loss": 0.01112211742438376, + "grad_norm": 0.12760162353515625, + "learning_rate": 1.976949950748616e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004450448742136359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018928326782770455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010793730150908232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.61353999376297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25630, + "weighted_flow_matching_loss": 0.01112211742438376, + "weighted_rkd_loss": 0.00662240544334054 + }, + { + "epoch": 7.459994180971778, + "flow_matching_loss": 0.011838898621499538, + "grad_norm": 0.14156727492809296, + "learning_rate": 1.976166049404987e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004776518396101892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020352026214823125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01158823948353529, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6132850110530853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25640, + "weighted_flow_matching_loss": 0.011838898621499538, + "weighted_rkd_loss": 0.007106807082891464 + }, + { + "epoch": 7.462903695082921, + "flow_matching_loss": 0.012106885574758053, + "grad_norm": 0.11147069931030273, + "learning_rate": 1.9753820034146848e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004415613552555442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001796351233497262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010627578385174275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6130299985408783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25650, + "weighted_flow_matching_loss": 0.012106885574758053, + "weighted_rkd_loss": 0.0065151276532560585 + }, + { + "epoch": 7.4658132091940645, + "flow_matching_loss": 0.013054182846099138, + "grad_norm": 0.12691034376621246, + "learning_rate": 1.9745978130158813e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004333094344474375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018711040087509901, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010537292761728167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6127749562263489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25660, + "weighted_flow_matching_loss": 0.013054182846099138, + "weighted_rkd_loss": 0.0064570027170702815 + }, + { + "epoch": 7.468722723305208, + "flow_matching_loss": 0.011158103588968515, + "grad_norm": 0.09989948570728302, + "learning_rate": 1.9738134784467932e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004101259890012443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017514254082925617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009953945176675916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6125198662281036, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25670, + "weighted_flow_matching_loss": 0.011158103588968515, + "weighted_rkd_loss": 0.006097024912014604 + }, + { + "epoch": 7.471632237416351, + "flow_matching_loss": 0.011773331835865975, + "grad_norm": 0.12771977484226227, + "learning_rate": 1.9730289999456804e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0046426855027675625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019524073402862995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011237778374925256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6122647762298584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25680, + "weighted_flow_matching_loss": 0.011773331835865975, + "weighted_rkd_loss": 0.006880676187574863 + }, + { + "epoch": 7.474541751527495, + "flow_matching_loss": 0.012809293810278177, + "grad_norm": 0.12169993668794632, + "learning_rate": 1.9722443777508473e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003939306433312595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016377395193558186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009516352508217096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6120096445083618, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25690, + "weighted_flow_matching_loss": 0.012809293810278177, + "weighted_rkd_loss": 0.005824107048101723 + }, + { + "epoch": 7.4774512656386385, + "flow_matching_loss": 0.012541188485920429, + "grad_norm": 0.12693850696086884, + "learning_rate": 1.9714596121006414e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004451711219735444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017601174302399158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010663539823144674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6117544770240784, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25700, + "weighted_flow_matching_loss": 0.012541188485920429, + "weighted_rkd_loss": 0.0065234954934567215 + }, + { + "epoch": 7.480360779749782, + "flow_matching_loss": 0.012713660299777985, + "grad_norm": 0.12999288737773895, + "learning_rate": 1.9706747032334535e-05, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004150988138280809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001899822347331792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01020179851911962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.611499285697937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25710, + "weighted_flow_matching_loss": 0.012713660299777985, + "weighted_rkd_loss": 0.006238393764942884 + }, + { + "epoch": 7.483270293860925, + "flow_matching_loss": 0.011418477073311806, + "grad_norm": 0.11387404054403305, + "learning_rate": 1.9698896513877177e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038126079831272363, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015206063049845397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009145822422578932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6112440586090088, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25720, + "weighted_flow_matching_loss": 0.011418477073311806, + "weighted_rkd_loss": 0.005590356886386871 + }, + { + "epoch": 7.486179807972069, + "flow_matching_loss": 0.011805381346493959, + "grad_norm": 0.12336082756519318, + "learning_rate": 1.9691044568019128e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004433202696964145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019211153499782085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010787520837038756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6109888017177582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25730, + "weighted_flow_matching_loss": 0.011805381346493959, + "weighted_rkd_loss": 0.006591033935546875 + }, + { + "epoch": 7.489089322083212, + "flow_matching_loss": 0.01297516506165266, + "grad_norm": 0.1235591396689415, + "learning_rate": 1.96831911971456e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036530581302940844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017213197308592499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009027435909956694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6107335209846496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25740, + "weighted_flow_matching_loss": 0.01297516506165266, + "weighted_rkd_loss": 0.0055132446810603145 + }, + { + "epoch": 7.491998836194355, + "flow_matching_loss": 0.011839564424008131, + "grad_norm": 0.12821897864341736, + "learning_rate": 1.9675336403642234e-05, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004826562036760152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002015119080897421, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011668243259191514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6104781925678253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25750, + "weighted_flow_matching_loss": 0.011839564424008131, + "weighted_rkd_loss": 0.007123119849711657 + }, + { + "epoch": 7.494908350305499, + "flow_matching_loss": 0.012889466434717178, + "grad_norm": 0.12218111753463745, + "learning_rate": 1.966748018989512e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003709906106814742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014999826671555638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008919794764369725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6102228462696075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25760, + "weighted_flow_matching_loss": 0.012889466434717178, + "weighted_rkd_loss": 0.005443069851025939 + }, + { + "epoch": 7.497817864416643, + "flow_matching_loss": 0.011895548924803734, + "grad_norm": 0.13057585060596466, + "learning_rate": 1.965962255829076e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004767196881584823, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020281167118810116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01156251048669219, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6099674701690674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25770, + "weighted_flow_matching_loss": 0.011895548924803734, + "weighted_rkd_loss": 0.007052765972912311 + }, + { + "epoch": 7.5007273785277855, + "flow_matching_loss": 0.014103512559086085, + "grad_norm": 0.14419198036193848, + "learning_rate": 1.9651763511216096e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003991372650489211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016983620240353047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009681107383221388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6097120642662048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25780, + "weighted_flow_matching_loss": 0.014103512559086085, + "weighted_rkd_loss": 0.005902692256495356 + }, + { + "epoch": 7.503636892638929, + "flow_matching_loss": 0.013795596174895763, + "grad_norm": 0.13825224339962006, + "learning_rate": 1.96439030510585e-05, + "loss": 0.0202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004411092260852456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016686723451130093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010490856738761068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6094566226005554, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25790, + "weighted_flow_matching_loss": 0.013795596174895763, + "weighted_rkd_loss": 0.006393747637048364 + }, + { + "epoch": 7.506546406750073, + "flow_matching_loss": 0.011872144974768161, + "grad_norm": 0.1311977505683899, + "learning_rate": 1.9636041180205774e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003941091382876039, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017230347380973398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009605217352509498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6092011511325837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25800, + "weighted_flow_matching_loss": 0.011872144974768161, + "weighted_rkd_loss": 0.00585152255371213 + }, + { + "epoch": 7.509455920861216, + "flow_matching_loss": 0.013347654882818461, + "grad_norm": 0.13263988494873047, + "learning_rate": 1.9628177901046144e-05, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004246946563944221, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017916218377649785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01028551496565342, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6089456498622894, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25810, + "weighted_flow_matching_loss": 0.013347654882818461, + "weighted_rkd_loss": 0.006263391301035881 + }, + { + "epoch": 7.512365434972359, + "flow_matching_loss": 0.011784555297344923, + "grad_norm": 0.11632808297872543, + "learning_rate": 1.962031321596827e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004152690316550434, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016117450082674622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009917125571519136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6086901247501373, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25820, + "weighted_flow_matching_loss": 0.011784555297344923, + "weighted_rkd_loss": 0.006036472180858255 + }, + { + "epoch": 7.515274949083503, + "flow_matching_loss": 0.013278154097497464, + "grad_norm": 0.12970243394374847, + "learning_rate": 1.961244712736123e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004586693411692977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018069796264171601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010980366403236986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6084345698356628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25830, + "weighted_flow_matching_loss": 0.013278154097497464, + "weighted_rkd_loss": 0.0066809059586375955 + }, + { + "epoch": 7.518184463194647, + "flow_matching_loss": 0.012758163223043083, + "grad_norm": 0.16747158765792847, + "learning_rate": 1.9604579637614535e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040628042770549655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001648450770881027, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00977405933663249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6081789791584015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25840, + "weighted_flow_matching_loss": 0.012758163223043083, + "weighted_rkd_loss": 0.005944331549108029 + }, + { + "epoch": 7.52109397730579, + "flow_matching_loss": 0.011580291576683522, + "grad_norm": 0.12091898173093796, + "learning_rate": 1.9596710749118126e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004192075598984957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018159904051572085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010200141463428736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6079233646392822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25850, + "weighted_flow_matching_loss": 0.011580291576683522, + "weighted_rkd_loss": 0.006200894434005022 + }, + { + "epoch": 7.524003491416933, + "flow_matching_loss": 0.01157343965023756, + "grad_norm": 0.16091348230838776, + "learning_rate": 1.9588840464262364e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004809062276035547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002194516360759735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01181264091283083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6076677143573761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25860, + "weighted_flow_matching_loss": 0.01157343965023756, + "weighted_rkd_loss": 0.007178152585402131 + }, + { + "epoch": 7.526913005528077, + "flow_matching_loss": 0.011655752360820771, + "grad_norm": 0.11241810023784637, + "learning_rate": 1.9580968785438026e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004310107510536909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020362700801342727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010656485054641961, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6074120402336121, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25870, + "weighted_flow_matching_loss": 0.011655752360820771, + "weighted_rkd_loss": 0.006472875783219934 + }, + { + "epoch": 7.52982251963922, + "flow_matching_loss": 0.011424438189715147, + "grad_norm": 0.13989409804344177, + "learning_rate": 1.9573095715036324e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004362581996247172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001870331261307001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010595495253801346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6071563303470612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25880, + "weighted_flow_matching_loss": 0.011424438189715147, + "weighted_rkd_loss": 0.006433050846680999 + }, + { + "epoch": 7.532732033750364, + "flow_matching_loss": 0.012440904323011637, + "grad_norm": 0.12548218667507172, + "learning_rate": 1.9565221255448892e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004161796835251153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018015978450421244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010125191509723663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6069006025791168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25890, + "weighted_flow_matching_loss": 0.012440904323011637, + "weighted_rkd_loss": 0.006144962832331657 + }, + { + "epoch": 7.535641547861507, + "flow_matching_loss": 0.01166869206354022, + "grad_norm": 0.16296978294849396, + "learning_rate": 1.9557345409067775e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004437617515213788, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001848608464933932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010723843425512313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6066448271274567, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25900, + "weighted_flow_matching_loss": 0.01166869206354022, + "weighted_rkd_loss": 0.006505501084029674 + }, + { + "epoch": 7.538551061972651, + "flow_matching_loss": 0.011944688996300102, + "grad_norm": 0.16997624933719635, + "learning_rate": 1.954946817828546e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004452446685172617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020805692882277073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010985462740063668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6063890397548676, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25910, + "weighted_flow_matching_loss": 0.011944688996300102, + "weighted_rkd_loss": 0.006661385623738169 + }, + { + "epoch": 7.541460576083794, + "flow_matching_loss": 0.011818921566009522, + "grad_norm": 0.14037524163722992, + "learning_rate": 1.954158956549483e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004239403642714024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018973605008795857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01037616766989231, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6061332166194916, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25920, + "weighted_flow_matching_loss": 0.011818921566009522, + "weighted_rkd_loss": 0.0062892683781683445 + }, + { + "epoch": 7.5443700901949375, + "flow_matching_loss": 0.010600908007472754, + "grad_norm": 0.10671139508485794, + "learning_rate": 1.9533709573089204e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004709545732475817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021091228700242937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011528214300051331, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6058773696422577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25930, + "weighted_flow_matching_loss": 0.010600908007472754, + "weighted_rkd_loss": 0.0069847228471189736 + }, + { + "epoch": 7.547279604306081, + "flow_matching_loss": 0.011632854677736759, + "grad_norm": 0.11344154924154282, + "learning_rate": 1.9525828203462313e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003910985821858048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015621861326508223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00938415778800845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.605621474981308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25940, + "weighted_flow_matching_loss": 0.011632854677736759, + "weighted_rkd_loss": 0.005683212867006659 + }, + { + "epoch": 7.550189118417224, + "flow_matching_loss": 0.012937019858509303, + "grad_norm": 0.12301462888717651, + "learning_rate": 1.9517945459008318e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003785122465342283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015976787311956287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00916792368516326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6053655743598938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25950, + "weighted_flow_matching_loss": 0.012937019858509303, + "weighted_rkd_loss": 0.005549963563680649 + }, + { + "epoch": 7.553098632528368, + "flow_matching_loss": 0.01201171539723873, + "grad_norm": 0.12670746445655823, + "learning_rate": 1.9510061342121776e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004128128057345748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017520609777420758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010008317232131959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6051096320152283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25960, + "weighted_flow_matching_loss": 0.01201171539723873, + "weighted_rkd_loss": 0.00605616308748722 + }, + { + "epoch": 7.5560081466395115, + "flow_matching_loss": 0.013222566340118647, + "grad_norm": 0.13023515045642853, + "learning_rate": 1.950217585519768e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004283207468688488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018185932654887437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010385008249431848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6048536598682404, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25970, + "weighted_flow_matching_loss": 0.013222566340118647, + "weighted_rkd_loss": 0.006281416630372405 + }, + { + "epoch": 7.558917660750654, + "flow_matching_loss": 0.010552478116005659, + "grad_norm": 0.1223161593079567, + "learning_rate": 1.9494289000631437e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004635600303299725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019999779760837555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011271178629249335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.60459765791893, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25980, + "weighted_flow_matching_loss": 0.010552478116005659, + "weighted_rkd_loss": 0.0068145109806209804 + }, + { + "epoch": 7.561827174861798, + "flow_matching_loss": 0.013557506119832396, + "grad_norm": 0.11089671403169632, + "learning_rate": 1.9486400780818857e-05, + "loss": 0.0201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004439099761657417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001879186904989183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010757386405020952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6043416440486908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25990, + "weighted_flow_matching_loss": 0.013557506119832396, + "weighted_rkd_loss": 0.00650116903707385 + }, + { + "epoch": 7.564736688972942, + "flow_matching_loss": 0.012871523946523666, + "grad_norm": 0.15057186782360077, + "learning_rate": 1.9478511198156175e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004506615805439651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019786106422543526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010991842206567525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6040855884552002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26000, + "weighted_flow_matching_loss": 0.012871523946523666, + "weighted_rkd_loss": 0.0066400018986314535 + }, + { + "epoch": 7.5676462030840845, + "flow_matching_loss": 0.010896171908825636, + "grad_norm": 0.14345407485961914, + "learning_rate": 1.9470620255040043e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004631979577243328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018489855108782649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011112944735214115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6038295090198517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26010, + "weighted_flow_matching_loss": 0.010896171908825636, + "weighted_rkd_loss": 0.006710283039137721 + }, + { + "epoch": 7.570555717195228, + "flow_matching_loss": 0.01243220092728734, + "grad_norm": 0.14096756279468536, + "learning_rate": 1.9462727953867515e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003659291844815016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015183537849225103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008836937556043267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6035733997821808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26020, + "weighted_flow_matching_loss": 0.01243220092728734, + "weighted_rkd_loss": 0.005333764012902975 + }, + { + "epoch": 7.573465231306372, + "flow_matching_loss": 0.01174339335411787, + "grad_norm": 0.11597073823213577, + "learning_rate": 1.9454834297036066e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003407653700560331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013268950511701404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008142202533781529, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6033172607421875, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26030, + "weighted_flow_matching_loss": 0.01174339335411787, + "weighted_rkd_loss": 0.0049123736564069985 + }, + { + "epoch": 7.576374745417516, + "flow_matching_loss": 0.011989746289327741, + "grad_norm": 0.10621149092912674, + "learning_rate": 1.944693928694358e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045416712993755935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020124253816902636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011095767980441451, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6030610859394073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26040, + "weighted_flow_matching_loss": 0.011989746289327741, + "weighted_rkd_loss": 0.006691411230713129 + }, + { + "epoch": 7.5792842595286585, + "flow_matching_loss": 0.010213910322636365, + "grad_norm": 0.13375096023082733, + "learning_rate": 1.943904292598835e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004201734578236938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018183480366133153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010221817158162593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6028048932552338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26050, + "weighted_flow_matching_loss": 0.010213910322636365, + "weighted_rkd_loss": 0.006161660421639681 + }, + { + "epoch": 7.582193773639802, + "flow_matching_loss": 0.012386936135590077, + "grad_norm": 0.14105917513370514, + "learning_rate": 1.9431145216569093e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00419421293772757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017882379819639028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010176663752645254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6025486707687377, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26060, + "weighted_flow_matching_loss": 0.012386936135590077, + "weighted_rkd_loss": 0.006132001569494605 + }, + { + "epoch": 7.585103287750946, + "flow_matching_loss": 0.011977864243090153, + "grad_norm": 0.13976900279521942, + "learning_rate": 1.942324616108491e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00447736238129437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019079549587331713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010862679686397314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6022924184799194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26070, + "weighted_flow_matching_loss": 0.011977864243090153, + "weighted_rkd_loss": 0.006542527722194791 + }, + { + "epoch": 7.588012801862089, + "flow_matching_loss": 0.012211561016738415, + "grad_norm": 0.13224512338638306, + "learning_rate": 1.9415345761935342e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037609033053740857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015275611192919313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0090493677649647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6020361185073853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26080, + "weighted_flow_matching_loss": 0.012211561016738415, + "weighted_rkd_loss": 0.005448068515397609 + }, + { + "epoch": 7.590922315973232, + "flow_matching_loss": 0.012056117597967386, + "grad_norm": 0.11101607978343964, + "learning_rate": 1.9407444021520302e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004084047884680331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016908515943214297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009858947480097413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6017798185348511, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26090, + "weighted_flow_matching_loss": 0.012056117597967386, + "weighted_rkd_loss": 0.005932876048609615 + }, + { + "epoch": 7.593831830084376, + "flow_matching_loss": 0.011202598921954632, + "grad_norm": 0.11673720926046371, + "learning_rate": 1.9399540942240144e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00400730581022799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016006635618396104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009615275217220188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.601523494720459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26100, + "weighted_flow_matching_loss": 0.011202598921954632, + "weighted_rkd_loss": 0.0057838076492771505 + }, + { + "epoch": 7.59674134419552, + "flow_matching_loss": 0.011323820101097226, + "grad_norm": 0.09567104279994965, + "learning_rate": 1.9391636526495613e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003922170703299343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016997047583572566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009544046223163604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6012671291828156, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26110, + "weighted_flow_matching_loss": 0.011323820101097226, + "weighted_rkd_loss": 0.005738490913063288 + }, + { + "epoch": 7.599650858306663, + "flow_matching_loss": 0.012326455116271973, + "grad_norm": 0.11800482869148254, + "learning_rate": 1.9383730776687858e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035864981822669507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015622961218468845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008735292544588446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6010107338428498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26120, + "weighted_flow_matching_loss": 0.012326455116271973, + "weighted_rkd_loss": 0.005249928310513496 + }, + { + "epoch": 7.602560372417806, + "flow_matching_loss": 0.011072390386834741, + "grad_norm": 0.1668427437543869, + "learning_rate": 1.9375823695218437e-05, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041714109480381015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018482192303054036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010191041184589267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.600754314661026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26130, + "weighted_flow_matching_loss": 0.011072390386834741, + "weighted_rkd_loss": 0.00612225141376257 + }, + { + "epoch": 7.605469886528949, + "flow_matching_loss": 0.011782808322459459, + "grad_norm": 0.11586008965969086, + "learning_rate": 1.9367915284489316e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004173444071784616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001786336931400001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010133225144818426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6004978716373444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26140, + "weighted_flow_matching_loss": 0.011782808322459459, + "weighted_rkd_loss": 0.006084924237802625 + }, + { + "epoch": 7.608379400640093, + "flow_matching_loss": 0.011937901750206948, + "grad_norm": 0.10593578219413757, + "learning_rate": 1.9360005546902856e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004195230454206467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017850746517069638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01017553573474288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6002413928508759, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26150, + "weighted_flow_matching_loss": 0.011937901750206948, + "weighted_rkd_loss": 0.006107764225453139 + }, + { + "epoch": 7.611288914751237, + "flow_matching_loss": 0.013308283034712076, + "grad_norm": 0.12077146768569946, + "learning_rate": 1.9352094484861836e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035881235264241694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014689276576973498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008645174698904157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5999848902225494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26160, + "weighted_flow_matching_loss": 0.013308283034712076, + "weighted_rkd_loss": 0.005186952371150255 + }, + { + "epoch": 7.61419842886238, + "flow_matching_loss": 0.012710040993988514, + "grad_norm": 0.11626885086297989, + "learning_rate": 1.934418210076942e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033407495124265552, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013858210644684733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008067320194095373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5997283637523652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26170, + "weighted_flow_matching_loss": 0.012710040993988514, + "weighted_rkd_loss": 0.004838271462358534 + }, + { + "epoch": 7.617107942973523, + "flow_matching_loss": 0.01278913589194417, + "grad_norm": 0.11930225789546967, + "learning_rate": 1.9336268397029176e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004349888442084194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019113054731860757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010611082427203655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5994718074798584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26180, + "weighted_flow_matching_loss": 0.01278913589194417, + "weighted_rkd_loss": 0.006361165922135115 + }, + { + "epoch": 7.620017457084667, + "flow_matching_loss": 0.01169254695996642, + "grad_norm": 0.14038436114788055, + "learning_rate": 1.9328353376045096e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004154773359186947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017067390377633273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010016285767778754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5992152094841003, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26190, + "weighted_flow_matching_loss": 0.01169254695996642, + "weighted_rkd_loss": 0.006001925934106112 + }, + { + "epoch": 7.6229269711958105, + "flow_matching_loss": 0.013188999611884356, + "grad_norm": 0.13528130948543549, + "learning_rate": 1.932043704022154e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037414243444800375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016487481072545052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009131596842780709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5989586055278778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26200, + "weighted_flow_matching_loss": 0.013188999611884356, + "weighted_rkd_loss": 0.005469455779530108 + }, + { + "epoch": 7.625836485306953, + "flow_matching_loss": 0.011064124386757612, + "grad_norm": 0.11021123826503754, + "learning_rate": 1.9312519391963284e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038542904425412417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016808013431727886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009389382228255272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5987019717693329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26210, + "weighted_flow_matching_loss": 0.011064124386757612, + "weighted_rkd_loss": 0.0056214996613562105 + }, + { + "epoch": 7.628745999418097, + "flow_matching_loss": 0.012504037842154503, + "grad_norm": 0.1326000690460205, + "learning_rate": 1.9304600433675504e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039046348072588445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016234185080975295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009432688076049089, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5984453082084655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26220, + "weighted_flow_matching_loss": 0.012504037842154503, + "weighted_rkd_loss": 0.00564490226097405 + }, + { + "epoch": 7.631655513529241, + "flow_matching_loss": 0.01186357494443655, + "grad_norm": 0.15269702672958374, + "learning_rate": 1.929668016776377e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004487331095151603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017854070756584406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010760069079697131, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5981886148452759, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26230, + "weighted_flow_matching_loss": 0.01186357494443655, + "weighted_rkd_loss": 0.006436533061787486 + }, + { + "epoch": 7.6345650276403845, + "flow_matching_loss": 0.01160776657052338, + "grad_norm": 0.12830476462841034, + "learning_rate": 1.9288758596634046e-05, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003863825066946447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015826145885512234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009310264652594925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5979318916797638, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26240, + "weighted_flow_matching_loss": 0.01160776657052338, + "weighted_rkd_loss": 0.0055669717025011774 + }, + { + "epoch": 7.637474541751527, + "flow_matching_loss": 0.011886334279552102, + "grad_norm": 0.12636300921440125, + "learning_rate": 1.9280835722692698e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004112155712209642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017900820588693024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010014393692836165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5976751446723938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26250, + "weighted_flow_matching_loss": 0.011886334279552102, + "weighted_rkd_loss": 0.005985434912145138 + }, + { + "epoch": 7.640384055862671, + "flow_matching_loss": 0.013007785473018885, + "grad_norm": 0.10836988687515259, + "learning_rate": 1.9272911548346487e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004483339795842767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001961747056338936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010928426682949067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5974183738231659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26260, + "weighted_flow_matching_loss": 0.013007785473018885, + "weighted_rkd_loss": 0.0065288869198411705 + }, + { + "epoch": 7.643293569973815, + "flow_matching_loss": 0.012780715059489012, + "grad_norm": 0.13425740599632263, + "learning_rate": 1.926498607600257e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003704146924428642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017548620468005537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009163155965507031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5971615791320801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26270, + "weighted_flow_matching_loss": 0.012780715059489012, + "weighted_rkd_loss": 0.00547187770716846 + }, + { + "epoch": 7.6462030840849575, + "flow_matching_loss": 0.012845335248857736, + "grad_norm": 0.12872569262981415, + "learning_rate": 1.9257059308068492e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00440977057442069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018436894752085209, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010663230624049901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5969047486782074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26280, + "weighted_flow_matching_loss": 0.012845335248857736, + "weighted_rkd_loss": 0.006364959804341197 + }, + { + "epoch": 7.649112598196101, + "flow_matching_loss": 0.01214103978127241, + "grad_norm": 0.1546725481748581, + "learning_rate": 1.9249131246952195e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004378724680282176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018161159590817989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010573565308004618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5966478884220123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26290, + "weighted_flow_matching_loss": 0.01214103978127241, + "weighted_rkd_loss": 0.006308629876002669 + }, + { + "epoch": 7.652022112307245, + "flow_matching_loss": 0.013154549337923527, + "grad_norm": 0.1368815153837204, + "learning_rate": 1.924120189506202e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003546208655461669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015692061220761388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008661623392254115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5963909924030304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26300, + "weighted_flow_matching_loss": 0.013154549337923527, + "weighted_rkd_loss": 0.005165717494674027 + }, + { + "epoch": 7.654931626418388, + "flow_matching_loss": 0.013517090305685996, + "grad_norm": 0.13665737211704254, + "learning_rate": 1.9233271254806686e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004376195278018713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017207027296535671, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010473093297332526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5961340963840485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26310, + "weighted_flow_matching_loss": 0.013517090305685996, + "weighted_rkd_loss": 0.006243254989385605 + }, + { + "epoch": 7.6578411405295315, + "flow_matching_loss": 0.011549506522715091, + "grad_norm": 0.12073030322790146, + "learning_rate": 1.9225339328595323e-05, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003935798560269177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001623110892251134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009494708152487874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.595877182483673, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26320, + "weighted_flow_matching_loss": 0.011549506522715091, + "weighted_rkd_loss": 0.005657529062591493 + }, + { + "epoch": 7.660750654640675, + "flow_matching_loss": 0.01255979984998703, + "grad_norm": 0.12877337634563446, + "learning_rate": 1.9217406118837435e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004150865855626762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00161938559031114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009921117359772325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5956202208995819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26330, + "weighted_flow_matching_loss": 0.01255979984998703, + "weighted_rkd_loss": 0.005909298057667911 + }, + { + "epoch": 7.663660168751818, + "flow_matching_loss": 0.011890244111418724, + "grad_norm": 0.11998844891786575, + "learning_rate": 1.9209471627942922e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004554038750939071, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019891280913725494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011097205709666014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5953632354736328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26340, + "weighted_flow_matching_loss": 0.011890244111418724, + "weighted_rkd_loss": 0.006606803368777036 + }, + { + "epoch": 7.666569682862962, + "flow_matching_loss": 0.011869540717452765, + "grad_norm": 0.13010036945343018, + "learning_rate": 1.9201535858322072e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004665802558884024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023084538173861802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01164005920290947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5951062202453613, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26350, + "weighted_flow_matching_loss": 0.011869540717452765, + "weighted_rkd_loss": 0.006927046505734324 + }, + { + "epoch": 7.669479196974105, + "flow_matching_loss": 0.011750167701393366, + "grad_norm": 0.1029500886797905, + "learning_rate": 1.9193598812385562e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00391787679400295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016548108658753335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009490564418956637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5948491871356965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26360, + "weighted_flow_matching_loss": 0.011750167701393366, + "weighted_rkd_loss": 0.005645474418997765 + }, + { + "epoch": 7.672388711085249, + "flow_matching_loss": 0.011950783059000969, + "grad_norm": 0.13322778046131134, + "learning_rate": 1.9185660492544464e-05, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00365510662086308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014588063582777976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0087690195068717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5945921242237091, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26370, + "weighted_flow_matching_loss": 0.011950783059000969, + "weighted_rkd_loss": 0.005213991133496165 + }, + { + "epoch": 7.675298225196392, + "flow_matching_loss": 0.012075341958552598, + "grad_norm": 0.13118457794189453, + "learning_rate": 1.917772090121022e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003808346064761281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015967288287356495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00921342084184289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5943350374698639, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26380, + "weighted_flow_matching_loss": 0.012075341958552598, + "weighted_rkd_loss": 0.005475819064304233 + }, + { + "epoch": 7.678207739307536, + "flow_matching_loss": 0.011927056312561034, + "grad_norm": 0.1173861101269722, + "learning_rate": 1.9169780040794678e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035107893869280814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015706413658335805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008592220116406679, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5940779209136963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26390, + "weighted_flow_matching_loss": 0.011927056312561034, + "weighted_rkd_loss": 0.0051044727675616745 + }, + { + "epoch": 7.681117253418679, + "flow_matching_loss": 0.01230621887370944, + "grad_norm": 0.1397821307182312, + "learning_rate": 1.916183791371005e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00449556321837008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018838761490769685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010875002481043339, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5938207745552063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26400, + "weighted_flow_matching_loss": 0.01230621887370944, + "weighted_rkd_loss": 0.006457830173894763 + }, + { + "epoch": 7.684026767529822, + "flow_matching_loss": 0.0124635249376297, + "grad_norm": 0.12675872445106506, + "learning_rate": 1.9153894522368952e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038587038405239584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016405175789259373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009357925318181515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5935636162757874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26410, + "weighted_flow_matching_loss": 0.0124635249376297, + "weighted_rkd_loss": 0.005554471304640174 + }, + { + "epoch": 7.686936281640966, + "flow_matching_loss": 0.012955979816615582, + "grad_norm": 0.10730256140232086, + "learning_rate": 1.9145949869184377e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004545612190850079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019052774761803449, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0109965018928051, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5933064222335815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26420, + "weighted_flow_matching_loss": 0.012955979816615582, + "weighted_rkd_loss": 0.006524310167878866 + }, + { + "epoch": 7.68984579575211, + "flow_matching_loss": 0.010235021077096462, + "grad_norm": 0.1155097633600235, + "learning_rate": 1.91380039565697e-05, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003993679420091212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001759294338990003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009746653167530895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5930492043495178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26430, + "weighted_flow_matching_loss": 0.010235021077096462, + "weighted_rkd_loss": 0.005780365411192178 + }, + { + "epoch": 7.692755309863253, + "flow_matching_loss": 0.012952908407896757, + "grad_norm": 0.12674769759178162, + "learning_rate": 1.9130056786938678e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037980570225045083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001686274167150259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009282388119027018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5927919626235962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26440, + "weighted_flow_matching_loss": 0.012952908407896757, + "weighted_rkd_loss": 0.005502546252682805 + }, + { + "epoch": 7.695664823974396, + "flow_matching_loss": 0.01198135893791914, + "grad_norm": 0.12705962359905243, + "learning_rate": 1.9122108362705447e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038101096171885727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016372539452277125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00925747319124639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5925346910953522, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26450, + "weighted_flow_matching_loss": 0.01198135893791914, + "weighted_rkd_loss": 0.00548532260581851 + }, + { + "epoch": 7.69857433808554, + "flow_matching_loss": 0.012128226552158594, + "grad_norm": 0.1311628669500351, + "learning_rate": 1.9114158686284538e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003949381178244948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016326370183378457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009531399328261613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5922773957252503, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26460, + "weighted_flow_matching_loss": 0.012128226552158594, + "weighted_rkd_loss": 0.005645215814001858 + }, + { + "epoch": 7.7014838521966835, + "flow_matching_loss": 0.011733073648065329, + "grad_norm": 0.12107864767313004, + "learning_rate": 1.910620776009085e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041714402381330725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001723329210653901, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010066209733486176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5920200824737549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26470, + "weighted_flow_matching_loss": 0.011733073648065329, + "weighted_rkd_loss": 0.005959318391978741 + }, + { + "epoch": 7.704393366307826, + "flow_matching_loss": 0.012299473024904728, + "grad_norm": 0.13655780255794525, + "learning_rate": 1.9098255586539653e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004506010073237121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021759562776423992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011187976598739624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5917627334594726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26480, + "weighted_flow_matching_loss": 0.012299473024904728, + "weighted_rkd_loss": 0.006620500516146421 + }, + { + "epoch": 7.70730288041897, + "flow_matching_loss": 0.01313595725223422, + "grad_norm": 0.11644618213176727, + "learning_rate": 1.909030216804663e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004202643781900406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017711823922581971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010176469944417477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.591505366563797, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26490, + "weighted_flow_matching_loss": 0.01313595725223422, + "weighted_rkd_loss": 0.006019446579739451 + }, + { + "epoch": 7.710212394530114, + "flow_matching_loss": 0.012716086255386472, + "grad_norm": 0.11966034770011902, + "learning_rate": 1.9082347507027797e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038810411468148233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016167097026482224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00937879215925932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.591247969865799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26500, + "weighted_flow_matching_loss": 0.012716086255386472, + "weighted_rkd_loss": 0.005545233143493533 + }, + { + "epoch": 7.713121908641257, + "flow_matching_loss": 0.011623035091906786, + "grad_norm": 0.10737276077270508, + "learning_rate": 1.9074391605899584e-05, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036905797896906734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015543715679086744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008935531182214618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.590990549325943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26510, + "weighted_flow_matching_loss": 0.011623035091906786, + "weighted_rkd_loss": 0.005280769569799304 + }, + { + "epoch": 7.7160314227524, + "flow_matching_loss": 0.012022676318883896, + "grad_norm": 0.11820026487112045, + "learning_rate": 1.906643446707878e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041346086887642745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017229762626811862, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00999219361692667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5907331109046936, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26520, + "weighted_flow_matching_loss": 0.012022676318883896, + "weighted_rkd_loss": 0.005902637727558613 + }, + { + "epoch": 7.718940936863544, + "flow_matching_loss": 0.01296754302456975, + "grad_norm": 0.137826070189476, + "learning_rate": 1.9058476092982552e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004136476619169116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017791268764995039, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01005208007991314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5904756188392639, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26530, + "weighted_flow_matching_loss": 0.01296754302456975, + "weighted_rkd_loss": 0.005935416650027036 + }, + { + "epoch": 7.721850450974687, + "flow_matching_loss": 0.011978735495358706, + "grad_norm": 0.1607394814491272, + "learning_rate": 1.9050516486028444e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00433529878500849, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017312735435552895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010401870962232352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5902181267738342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26540, + "weighted_flow_matching_loss": 0.011978735495358706, + "weighted_rkd_loss": 0.006139363301917911 + }, + { + "epoch": 7.7247599650858305, + "flow_matching_loss": 0.010848220391198992, + "grad_norm": 0.10894320905208588, + "learning_rate": 1.9042555648634375e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003864941792562604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001687980385031551, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009417863935232163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5899606287479401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26550, + "weighted_flow_matching_loss": 0.010848220391198992, + "weighted_rkd_loss": 0.005556142050772905 + }, + { + "epoch": 7.727669479196974, + "flow_matching_loss": 0.01172208096832037, + "grad_norm": 0.12084362655878067, + "learning_rate": 1.903459358321864e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037485143868252633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014958850806578994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008992913831025361, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5897030770778656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26560, + "weighted_flow_matching_loss": 0.01172208096832037, + "weighted_rkd_loss": 0.005303179984912277 + }, + { + "epoch": 7.730578993308118, + "flow_matching_loss": 0.011986975651234388, + "grad_norm": 0.12471292167901993, + "learning_rate": 1.9026630292199896e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00369426803663373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015965080354362726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008985044062137603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5894455075263977, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26570, + "weighted_flow_matching_loss": 0.011986975651234388, + "weighted_rkd_loss": 0.005296169640496373 + }, + { + "epoch": 7.733488507419261, + "flow_matching_loss": 0.011304520349949599, + "grad_norm": 0.1393686980009079, + "learning_rate": 1.901866577799719e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004548620083369315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018462350242771208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010943475179374218, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5891879141330719, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26580, + "weighted_flow_matching_loss": 0.011304520349949599, + "weighted_rkd_loss": 0.006447782460600138 + }, + { + "epoch": 7.7363980215304045, + "flow_matching_loss": 0.012900522816926241, + "grad_norm": 0.11295805126428604, + "learning_rate": 1.901070004302993e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004404111346229911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019464464508928358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010754669085144997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5889303028583527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26590, + "weighted_flow_matching_loss": 0.012900522816926241, + "weighted_rkd_loss": 0.006333841569721698 + }, + { + "epoch": 7.739307535641548, + "flow_matching_loss": 0.010904006473720074, + "grad_norm": 0.1412069946527481, + "learning_rate": 1.900273308971789e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003944748663343489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017016347264871001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009591131983324886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5886726617813111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26600, + "weighted_flow_matching_loss": 0.010904006473720074, + "weighted_rkd_loss": 0.005646014679223299 + }, + { + "epoch": 7.742217049752691, + "flow_matching_loss": 0.01322648860514164, + "grad_norm": 0.13810789585113525, + "learning_rate": 1.899476492048122e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003667234047316015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016790158464573325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009013483906164766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5884149968624115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26610, + "weighted_flow_matching_loss": 0.01322648860514164, + "weighted_rkd_loss": 0.005303656565956771 + }, + { + "epoch": 7.745126563863835, + "flow_matching_loss": 0.012089006416499615, + "grad_norm": 0.1492581069469452, + "learning_rate": 1.8986795537740448e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035746119683608413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017772160237655044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008926439937204123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.588157308101654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26620, + "weighted_flow_matching_loss": 0.012089006416499615, + "weighted_rkd_loss": 0.005250167893245816 + }, + { + "epoch": 7.748036077974978, + "flow_matching_loss": 0.014791769813746214, + "grad_norm": 0.11617450416088104, + "learning_rate": 1.897882494391645e-05, + "loss": 0.0204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003986549004912377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016339890309609473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009607087122276425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5878995895385742, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26630, + "weighted_flow_matching_loss": 0.014791769813746214, + "weighted_rkd_loss": 0.005648061167448759 + }, + { + "epoch": 7.750945592086122, + "flow_matching_loss": 0.011570022907108068, + "grad_norm": 0.12429366260766983, + "learning_rate": 1.8970853141430497e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039015776012092827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016627394128590823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009465894661843777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5876418471336364, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26640, + "weighted_flow_matching_loss": 0.011570022907108068, + "weighted_rkd_loss": 0.005562531435862184 + }, + { + "epoch": 7.753855106197265, + "flow_matching_loss": 0.013164197281002998, + "grad_norm": 0.13183720409870148, + "learning_rate": 1.89628801327042e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042092395713552834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001824196963571012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010242675989866256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5873840868473053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26650, + "weighted_flow_matching_loss": 0.013164197281002998, + "weighted_rkd_loss": 0.006016363296657801 + }, + { + "epoch": 7.756764620308409, + "flow_matching_loss": 0.011935422569513321, + "grad_norm": 0.13733626902103424, + "learning_rate": 1.8954905920159555e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003909952752292156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017650071182288229, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009584912750869989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5871263027191163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26660, + "weighted_flow_matching_loss": 0.011935422569513321, + "weighted_rkd_loss": 0.005627511907368899 + }, + { + "epoch": 7.7596741344195515, + "flow_matching_loss": 0.012554674968123436, + "grad_norm": 0.12845809757709503, + "learning_rate": 1.8946930506218916e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004308698186650872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019063344574533404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010523730795830489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5868684887886048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26670, + "weighted_flow_matching_loss": 0.012554674968123436, + "weighted_rkd_loss": 0.00617602034471929 + }, + { + "epoch": 7.762583648530695, + "flow_matching_loss": 0.012387987598776818, + "grad_norm": 0.11579778045415878, + "learning_rate": 1.8938953893305e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037689623655751347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014991273055784404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009037051955237985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5866106569766998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26680, + "weighted_flow_matching_loss": 0.012387987598776818, + "weighted_rkd_loss": 0.0053013238124549385 + }, + { + "epoch": 7.765493162641839, + "flow_matching_loss": 0.011449648346751929, + "grad_norm": 0.11607977002859116, + "learning_rate": 1.893097608384089e-05, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003789585130289197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015402075136080385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009119377750903369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.586352801322937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26690, + "weighted_flow_matching_loss": 0.011449648346751929, + "weighted_rkd_loss": 0.005347207793965936 + }, + { + "epoch": 7.768402676752983, + "flow_matching_loss": 0.011218276526778936, + "grad_norm": 0.1157790869474411, + "learning_rate": 1.892299708025004e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036866943119093776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015844638925045728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008957852469757199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5860949158668518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26700, + "weighted_flow_matching_loss": 0.011218276526778936, + "weighted_rkd_loss": 0.0052501819096505645 + }, + { + "epoch": 7.771312190864125, + "flow_matching_loss": 0.011027116980403661, + "grad_norm": 0.11668441444635391, + "learning_rate": 1.891501688495626e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003943890007212758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016925530624575913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009580333018675447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5858370125293731, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26710, + "weighted_flow_matching_loss": 0.011027116980403661, + "weighted_rkd_loss": 0.005612503457814455 + }, + { + "epoch": 7.774221704975269, + "flow_matching_loss": 0.011352650495246053, + "grad_norm": 0.12610357999801636, + "learning_rate": 1.8907035500383717e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003553031710907817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014318548841401935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008537918236106634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5855790793895721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26720, + "weighted_flow_matching_loss": 0.011352650495246053, + "weighted_rkd_loss": 0.004999605147168041 + }, + { + "epoch": 7.777131219086413, + "flow_matching_loss": 0.013232457265257836, + "grad_norm": 0.13087576627731323, + "learning_rate": 1.889905292895695e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003453317750245333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016338047513272612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008540440211072564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5853211283683777, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26730, + "weighted_flow_matching_loss": 0.013232457265257836, + "weighted_rkd_loss": 0.004998912615701556 + }, + { + "epoch": 7.780040733197556, + "flow_matching_loss": 0.010267983516678214, + "grad_norm": 0.11771523207426071, + "learning_rate": 1.889106917310085e-05, + "loss": 0.0155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003698089765384793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015400388976559043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008936218451708555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5850631594657898, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26740, + "weighted_flow_matching_loss": 0.010267983516678214, + "weighted_rkd_loss": 0.005228221556171775 + }, + { + "epoch": 7.782950247308699, + "flow_matching_loss": 0.013085705880075692, + "grad_norm": 0.1174994483590126, + "learning_rate": 1.8883084235240672e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039272122783586385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017124806647188962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00956690525636077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5848051488399506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26750, + "weighted_flow_matching_loss": 0.013085705880075692, + "weighted_rkd_loss": 0.005594761203974485 + }, + { + "epoch": 7.785859761419843, + "flow_matching_loss": 0.012736531533300876, + "grad_norm": 0.13102072477340698, + "learning_rate": 1.887509811780203e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004357974114827812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018827158608473837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010598664032295346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5845471322536469, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26760, + "weighted_flow_matching_loss": 0.012736531533300876, + "weighted_rkd_loss": 0.006195453368127346 + }, + { + "epoch": 7.788769275530987, + "flow_matching_loss": 0.011816261615604163, + "grad_norm": 0.11990918964147568, + "learning_rate": 1.88671108232109e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004158405633643269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019170007668435573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010233812034130096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5842890977859497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26770, + "weighted_flow_matching_loss": 0.011816261615604163, + "weighted_rkd_loss": 0.005979560734704137 + }, + { + "epoch": 7.79167878964213, + "flow_matching_loss": 0.012065039295703173, + "grad_norm": 0.13348278403282166, + "learning_rate": 1.88591223538936e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037523983977735043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016375699895434081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009142366610467434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5840310156345367, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26780, + "weighted_flow_matching_loss": 0.012065039295703173, + "weighted_rkd_loss": 0.005339372390881181 + }, + { + "epoch": 7.794588303753273, + "flow_matching_loss": 0.013002899195998906, + "grad_norm": 0.12857560813426971, + "learning_rate": 1.885113271227683e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038445705315098165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016411219257861376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0093302630353719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5837729275226593, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26790, + "weighted_flow_matching_loss": 0.013002899195998906, + "weighted_rkd_loss": 0.0054467767244204875 + }, + { + "epoch": 7.797497817864417, + "flow_matching_loss": 0.01195538891479373, + "grad_norm": 0.10320762544870377, + "learning_rate": 1.8843141900787625e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003877141326665878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00170491331955418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009459195798262954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5835148215293884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26800, + "weighted_flow_matching_loss": 0.01195538891479373, + "weighted_rkd_loss": 0.0055196164175868034 + }, + { + "epoch": 7.80040733197556, + "flow_matching_loss": 0.013080641627311707, + "grad_norm": 0.13707147538661957, + "learning_rate": 1.8835149921853383e-05, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004168910533189773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001826935331337154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010164756467565894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5832566797733307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26810, + "weighted_flow_matching_loss": 0.013080641627311707, + "weighted_rkd_loss": 0.005928657739423216 + }, + { + "epoch": 7.8033168460867035, + "flow_matching_loss": 0.012244921177625656, + "grad_norm": 0.13384795188903809, + "learning_rate": 1.8827156777901853e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004113582381978631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018535754643380642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010080740367993713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5829985201358795, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26820, + "weighted_flow_matching_loss": 0.012244921177625656, + "weighted_rkd_loss": 0.005877238884568215 + }, + { + "epoch": 7.806226360197847, + "flow_matching_loss": 0.012504712864756583, + "grad_norm": 0.1289973258972168, + "learning_rate": 1.8819162471361144e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004040220216847956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017459162161685526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009826356871053576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5827403366565704, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26830, + "weighted_flow_matching_loss": 0.012504712864756583, + "weighted_rkd_loss": 0.005726174311712384 + }, + { + "epoch": 7.80913587430899, + "flow_matching_loss": 0.01141017060726881, + "grad_norm": 0.11418264359235764, + "learning_rate": 1.8811167004659715e-05, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004032548679970205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018343367613852025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00989943416789174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5824821293354034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26840, + "weighted_flow_matching_loss": 0.01141017060726881, + "weighted_rkd_loss": 0.005766243347898126 + }, + { + "epoch": 7.812045388420134, + "flow_matching_loss": 0.010974101582542061, + "grad_norm": 0.11577781289815903, + "learning_rate": 1.8803170380226383e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00423826880287379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018682834808714688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010344821307808161, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.582223904132843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26850, + "weighted_flow_matching_loss": 0.010974101582542061, + "weighted_rkd_loss": 0.006022951891645789 + }, + { + "epoch": 7.8149549025312774, + "flow_matching_loss": 0.01171234268695116, + "grad_norm": 0.10709407925605774, + "learning_rate": 1.8795172600490302e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040897555649280545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017041030572727323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009883614117279648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5819656491279602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26860, + "weighted_flow_matching_loss": 0.01171234268695116, + "weighted_rkd_loss": 0.005751944379881024 + }, + { + "epoch": 7.81786441664242, + "flow_matching_loss": 0.012645342759788037, + "grad_norm": 0.12106430530548096, + "learning_rate": 1.878717366788099e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003584180702455342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015995172143448144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008767878729850053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.581707376241684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26870, + "weighted_flow_matching_loss": 0.012645342759788037, + "weighted_rkd_loss": 0.005100308661349117 + }, + { + "epoch": 7.820773930753564, + "flow_matching_loss": 0.012856861762702464, + "grad_norm": 0.13372474908828735, + "learning_rate": 1.8779173584828316e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004332654154859483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001778438431210816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010443746903911233, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5814490854740143, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26880, + "weighted_flow_matching_loss": 0.012856861762702464, + "weighted_rkd_loss": 0.006072452943772078 + }, + { + "epoch": 7.823683444864708, + "flow_matching_loss": 0.011439161095768213, + "grad_norm": 0.12212498486042023, + "learning_rate": 1.8771172353762488e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004244797048158944, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00179661464644596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01028620870783925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5811907589435578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26890, + "weighted_flow_matching_loss": 0.011439161095768213, + "weighted_rkd_loss": 0.005978211900219322 + }, + { + "epoch": 7.826592958975851, + "flow_matching_loss": 0.012029116181656719, + "grad_norm": 0.1066049113869667, + "learning_rate": 1.876316997711407e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004336612834595144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017358468729071319, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010409072693437337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5809324204921722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26900, + "weighted_flow_matching_loss": 0.012029116181656719, + "weighted_rkd_loss": 0.006046995893120766 + }, + { + "epoch": 7.829502473086994, + "flow_matching_loss": 0.012270040344446898, + "grad_norm": 0.21143856644630432, + "learning_rate": 1.8755166457313972e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003986713592894375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018145935493521393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009788020746782422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5806740522384644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26910, + "weighted_flow_matching_loss": 0.012270040344446898, + "weighted_rkd_loss": 0.005683583137579262 + }, + { + "epoch": 7.832411987198138, + "flow_matching_loss": 0.012105194106698035, + "grad_norm": 0.12657363712787628, + "learning_rate": 1.8747161796793457e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00397490996401757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016797004733234643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009629520354792476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.580415666103363, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26920, + "weighted_flow_matching_loss": 0.012105194106698035, + "weighted_rkd_loss": 0.005589052522554994 + }, + { + "epoch": 7.835321501309282, + "flow_matching_loss": 0.013878989592194557, + "grad_norm": 0.14473244547843933, + "learning_rate": 1.8739155997984124e-05, + "loss": 0.0197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004143970413133502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018204846302978695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010108425375074148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5801572740077973, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26930, + "weighted_flow_matching_loss": 0.013878989592194557, + "weighted_rkd_loss": 0.005864382605068386 + }, + { + "epoch": 7.838231015420424, + "flow_matching_loss": 0.01248282492160797, + "grad_norm": 0.13295653462409973, + "learning_rate": 1.8731149063317927e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004424408543854952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019660543068312107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010814871359616517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5798988461494445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26940, + "weighted_flow_matching_loss": 0.01248282492160797, + "weighted_rkd_loss": 0.0062713879626244305 + }, + { + "epoch": 7.841140529531568, + "flow_matching_loss": 0.010830728197470307, + "grad_norm": 0.12173369526863098, + "learning_rate": 1.872314099522716e-05, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038138205651193856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016032150131650269, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009230856085196137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5796403884887695, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26950, + "weighted_flow_matching_loss": 0.010830728197470307, + "weighted_rkd_loss": 0.005350645701400936 + }, + { + "epoch": 7.844050043642712, + "flow_matching_loss": 0.014411777164787054, + "grad_norm": 0.18508751690387726, + "learning_rate": 1.8715131796144466e-05, + "loss": 0.0203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004173718974925578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001886835542973131, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010234273504465818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.579381912946701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26960, + "weighted_flow_matching_loss": 0.014411777164787054, + "weighted_rkd_loss": 0.00592954340390861 + }, + { + "epoch": 7.846959557753856, + "flow_matching_loss": 0.011875903885811567, + "grad_norm": 0.129333958029747, + "learning_rate": 1.870712146850282e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038901617983356116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016984524903818966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00947877625003457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5791234254837037, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26970, + "weighted_flow_matching_loss": 0.011875903885811567, + "weighted_rkd_loss": 0.00548942843452096 + }, + { + "epoch": 7.849869071864998, + "flow_matching_loss": 0.012346784491091967, + "grad_norm": 0.11103178560733795, + "learning_rate": 1.8699110014735554e-05, + "loss": 0.0187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004484436567872762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019286268972791732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010897500067949295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5788649082183838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26980, + "weighted_flow_matching_loss": 0.012346784491091967, + "weighted_rkd_loss": 0.006308152573183179 + }, + { + "epoch": 7.852778585976142, + "flow_matching_loss": 0.012819674890488386, + "grad_norm": 0.1202368214726448, + "learning_rate": 1.8691097437276334e-05, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004332276643253863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001999624422751367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010664177685976028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.578606367111206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26990, + "weighted_flow_matching_loss": 0.012819674890488386, + "weighted_rkd_loss": 0.006170286238193512 + }, + { + "epoch": 7.855688100087286, + "flow_matching_loss": 0.011526833474636077, + "grad_norm": 0.13138623535633087, + "learning_rate": 1.868308373855917e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003466080920770764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001496719941496849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008428881829604506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5783478081226349, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27000, + "weighted_flow_matching_loss": 0.011526833474636077, + "weighted_rkd_loss": 0.004874780657701194 + }, + { + "epoch": 7.858597614198429, + "flow_matching_loss": 0.012842505425214767, + "grad_norm": 0.12774860858917236, + "learning_rate": 1.8675068921018415e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003892907127737999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015924156294204294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009378229873254896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5780892312526703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27010, + "weighted_flow_matching_loss": 0.012842505425214767, + "weighted_rkd_loss": 0.005421481421217322 + }, + { + "epoch": 7.861507128309572, + "flow_matching_loss": 0.011586054228246212, + "grad_norm": 0.11931123584508896, + "learning_rate": 1.8667052987088756e-05, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040409495821222665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016290952917188406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00971099454909563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5778306305408478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27020, + "weighted_flow_matching_loss": 0.011586054228246212, + "weighted_rkd_loss": 0.005611351714469493 + }, + { + "epoch": 7.864416642420716, + "flow_matching_loss": 0.012914030626416207, + "grad_norm": 0.13548097014427185, + "learning_rate": 1.8659035939205224e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003520392277278006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015282689244486391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008569053560495376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5775720059871674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27030, + "weighted_flow_matching_loss": 0.012914030626416207, + "weighted_rkd_loss": 0.004949191305786371 + }, + { + "epoch": 7.867326156531859, + "flow_matching_loss": 0.012606288865208625, + "grad_norm": 0.13672369718551636, + "learning_rate": 1.8651017779803178e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003744587954133749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015653250040486455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009054500935599208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5773133635520935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27040, + "weighted_flow_matching_loss": 0.012606288865208625, + "weighted_rkd_loss": 0.005227228906005621 + }, + { + "epoch": 7.870235670643003, + "flow_matching_loss": 0.011131384037435054, + "grad_norm": 0.10318375378847122, + "learning_rate": 1.8642998511318336e-05, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003704052045941353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015102957491762935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008918399829417467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5770546793937683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27050, + "weighted_flow_matching_loss": 0.011131384037435054, + "weighted_rkd_loss": 0.005146475997753442 + }, + { + "epoch": 7.873145184754146, + "flow_matching_loss": 0.011293976660817861, + "grad_norm": 0.11839981377124786, + "learning_rate": 1.863497813618673e-05, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004032139247283339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016253603505901992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009689638949930668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5767959952354431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27060, + "weighted_flow_matching_loss": 0.011293976660817861, + "weighted_rkd_loss": 0.005588917154818773 + }, + { + "epoch": 7.876054698865289, + "flow_matching_loss": 0.012125091161578894, + "grad_norm": 0.11768340319395065, + "learning_rate": 1.8626956656844746e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042366074630990624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018710969830863177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010344311874359847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5765373110771179, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27070, + "weighted_flow_matching_loss": 0.012125091161578894, + "weighted_rkd_loss": 0.0059638611972332 + }, + { + "epoch": 7.878964212976433, + "flow_matching_loss": 0.012517532426863908, + "grad_norm": 0.13800396025180817, + "learning_rate": 1.861893407572909e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004100902262143791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017394302529282867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009941234765574337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.576278567314148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27080, + "weighted_flow_matching_loss": 0.012517532426863908, + "weighted_rkd_loss": 0.005728893633931875 + }, + { + "epoch": 7.8818737270875765, + "flow_matching_loss": 0.012953375279903413, + "grad_norm": 0.12174038589000702, + "learning_rate": 1.861091039527682e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043633028632029894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019137676572427154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010640373406931757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.576019823551178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27090, + "weighted_flow_matching_loss": 0.012953375279903413, + "weighted_rkd_loss": 0.00612916499376297 + }, + { + "epoch": 7.88478324119872, + "flow_matching_loss": 0.012271346803754569, + "grad_norm": 0.1251438856124878, + "learning_rate": 1.8602885617925307e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004110548505559563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001667728158645332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009888825193047523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5757610499858856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27100, + "weighted_flow_matching_loss": 0.012271346803754569, + "weighted_rkd_loss": 0.005693589034490287 + }, + { + "epoch": 7.887692755309863, + "flow_matching_loss": 0.011525092646479607, + "grad_norm": 0.1131121888756752, + "learning_rate": 1.8594859746112273e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004386974428780377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020215722382999957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010795521223917604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5755022585391998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27110, + "weighted_flow_matching_loss": 0.011525092646479607, + "weighted_rkd_loss": 0.006212887354195118 + }, + { + "epoch": 7.890602269421007, + "flow_matching_loss": 0.011314522195607423, + "grad_norm": 0.11253797262907028, + "learning_rate": 1.8586832782275768e-05, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035935034742578865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014481773076113313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00863518426194787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.575243455171585, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27120, + "weighted_flow_matching_loss": 0.011314522195607423, + "weighted_rkd_loss": 0.0049672559136524795 + }, + { + "epoch": 7.89351178353215, + "flow_matching_loss": 0.012334603443741798, + "grad_norm": 0.11252736300230026, + "learning_rate": 1.857880472885417e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036515628919005392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015942528727464378, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008897378761321306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5749846160411834, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27130, + "weighted_flow_matching_loss": 0.012334603443741798, + "weighted_rkd_loss": 0.0051159163704141974 + }, + { + "epoch": 7.896421297643293, + "flow_matching_loss": 0.011432763747870922, + "grad_norm": 0.10143346339464188, + "learning_rate": 1.8570775588286185e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003751375386491418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014647993142716587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008967550192028283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.574725764989853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27140, + "weighted_flow_matching_loss": 0.011432763747870922, + "weighted_rkd_loss": 0.0051538754720240835 + }, + { + "epoch": 7.899330811754437, + "flow_matching_loss": 0.013269321341067553, + "grad_norm": 0.14221952855587006, + "learning_rate": 1.8562745363010854e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004130160319618881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001858716702554375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010119037237018347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5744668960571289, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27150, + "weighted_flow_matching_loss": 0.013269321341067553, + "weighted_rkd_loss": 0.005812973529100418 + }, + { + "epoch": 7.902240325865581, + "flow_matching_loss": 0.010975590813905001, + "grad_norm": 0.1005566343665123, + "learning_rate": 1.8554714055467555e-05, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038200382143259047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016976912971585988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009337767725810409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5742079973220825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27160, + "weighted_flow_matching_loss": 0.010975590813905001, + "weighted_rkd_loss": 0.005361902085132897 + }, + { + "epoch": 7.905149839976724, + "flow_matching_loss": 0.01166382534429431, + "grad_norm": 0.12470906227827072, + "learning_rate": 1.8546681668095978e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035615803906694056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014461857033893466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008569346414878965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5739490866661072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27170, + "weighted_flow_matching_loss": 0.01166382534429431, + "weighted_rkd_loss": 0.00491829055827111 + }, + { + "epoch": 7.908059354087867, + "flow_matching_loss": 0.012834502197802067, + "grad_norm": 0.14689385890960693, + "learning_rate": 1.853864820333616e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004033925314433873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017199487774632871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009787799324840307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5736901462078094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27180, + "weighted_flow_matching_loss": 0.012834502197802067, + "weighted_rkd_loss": 0.005615200661122799 + }, + { + "epoch": 7.910968868199011, + "flow_matching_loss": 0.011868636356666684, + "grad_norm": 0.12141537666320801, + "learning_rate": 1.8530613663628448e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003866402432322502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016855483525432646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009418353158980607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5734311938285828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27190, + "weighted_flow_matching_loss": 0.011868636356666684, + "weighted_rkd_loss": 0.005400771694257856 + }, + { + "epoch": 7.913878382310154, + "flow_matching_loss": 0.013034631870687008, + "grad_norm": 0.12226799875497818, + "learning_rate": 1.8522578051413525e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004370565596036613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019412295776419341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01068236087448895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5731722116470337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27200, + "weighted_flow_matching_loss": 0.013034631870687008, + "weighted_rkd_loss": 0.0061229022685438395 + }, + { + "epoch": 7.916787896421297, + "flow_matching_loss": 0.013668528478592634, + "grad_norm": 0.10852576792240143, + "learning_rate": 1.85145413691324e-05, + "loss": 0.02, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004492054833099246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020508411806076767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011034950800240039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5729132235050202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27210, + "weighted_flow_matching_loss": 0.013668528478592634, + "weighted_rkd_loss": 0.006322115031071007 + }, + { + "epoch": 7.919697410532441, + "flow_matching_loss": 0.01115085044875741, + "grad_norm": 0.10614313185214996, + "learning_rate": 1.8506503619226402e-05, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004111954546533525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017282685614190996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009952177666127682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5726542055606842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27220, + "weighted_flow_matching_loss": 0.01115085044875741, + "weighted_rkd_loss": 0.005699174711480737 + }, + { + "epoch": 7.922606924643585, + "flow_matching_loss": 0.011778089310973882, + "grad_norm": 0.10285308957099915, + "learning_rate": 1.849846480413719e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0043767375173047185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017879119142889977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010541387042030692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5723951756954193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27230, + "weighted_flow_matching_loss": 0.011778089310973882, + "weighted_rkd_loss": 0.006033922010101378 + }, + { + "epoch": 7.925516438754728, + "flow_matching_loss": 0.010616291780024767, + "grad_norm": 0.15129202604293823, + "learning_rate": 1.8490424926306745e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004175336542539298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018403538153506816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010191026888787747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.572136116027832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27240, + "weighted_flow_matching_loss": 0.010616291780024767, + "weighted_rkd_loss": 0.0058305627200752495 + }, + { + "epoch": 7.928425952865871, + "flow_matching_loss": 0.01216585300862789, + "grad_norm": 0.1246284618973732, + "learning_rate": 1.8482383988177362e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003524765931069851, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014218955766409636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008471427392214537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5718770444393158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27250, + "weighted_flow_matching_loss": 0.01216585300862789, + "weighted_rkd_loss": 0.004844625946134329 + }, + { + "epoch": 7.931335466977015, + "flow_matching_loss": 0.011160132242366672, + "grad_norm": 0.10914234071969986, + "learning_rate": 1.8474341992191666e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040152279892936345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016431317664682866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009673587884753943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5716179490089417, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27260, + "weighted_flow_matching_loss": 0.011160132242366672, + "weighted_rkd_loss": 0.005529579380527139 + }, + { + "epoch": 7.934244981088158, + "flow_matching_loss": 0.01294635171070695, + "grad_norm": 0.1315530240535736, + "learning_rate": 1.8466298940792614e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038693634094670417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016208918765187263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009359618742018938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5713588297367096, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27270, + "weighted_flow_matching_loss": 0.01294635171070695, + "weighted_rkd_loss": 0.005347675085067749 + }, + { + "epoch": 7.937154495199302, + "flow_matching_loss": 0.012276974925771355, + "grad_norm": 0.17229397594928741, + "learning_rate": 1.8458254836423467e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00435943475458771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002018528862390667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010737398406490684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5710996985435486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27280, + "weighted_flow_matching_loss": 0.012276974925771355, + "weighted_rkd_loss": 0.006132055353373289 + }, + { + "epoch": 7.940064009310445, + "flow_matching_loss": 0.011396382236853242, + "grad_norm": 0.14575150609016418, + "learning_rate": 1.8450209681527808e-05, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00399520737119019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016076080035418272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009598022792488337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5708405494689941, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27290, + "weighted_flow_matching_loss": 0.011396382236853242, + "weighted_rkd_loss": 0.0054790212772786616 + }, + { + "epoch": 7.942973523421589, + "flow_matching_loss": 0.012371825380250811, + "grad_norm": 0.10600314289331436, + "learning_rate": 1.8442163478549546e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041012242436409, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017074448987841606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009909893479198217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5705813765525818, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27300, + "weighted_flow_matching_loss": 0.012371825380250811, + "weighted_rkd_loss": 0.005654391506686807 + }, + { + "epoch": 7.945883037532732, + "flow_matching_loss": 0.012943687103688716, + "grad_norm": 0.15439000725746155, + "learning_rate": 1.8434116229932905e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004009935655631125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018156073754653334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00983547866344452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5703221797943115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27310, + "weighted_flow_matching_loss": 0.012943687103688716, + "weighted_rkd_loss": 0.005609391792677343 + }, + { + "epoch": 7.9487925516438755, + "flow_matching_loss": 0.011202472168952227, + "grad_norm": 0.10759956389665604, + "learning_rate": 1.8426067938122423e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038301757303997874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016810293658636511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009341380838304759, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5700629711151123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27320, + "weighted_flow_matching_loss": 0.011202472168952227, + "weighted_rkd_loss": 0.005325175961479545 + }, + { + "epoch": 7.951702065755019, + "flow_matching_loss": 0.011572288908064365, + "grad_norm": 0.12849488854408264, + "learning_rate": 1.841801860556296e-05, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034075153758749365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014532280969433487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008268258767202497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5698037445545197, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27330, + "weighted_flow_matching_loss": 0.011572288908064365, + "weighted_rkd_loss": 0.004711251868866384 + }, + { + "epoch": 7.954611579866162, + "flow_matching_loss": 0.012588723748922347, + "grad_norm": 0.1192120909690857, + "learning_rate": 1.8409968234699698e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036782021867111324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016369971563108265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008993401657789946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5695444941520691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27340, + "weighted_flow_matching_loss": 0.012588723748922347, + "weighted_rkd_loss": 0.005122193554416299 + }, + { + "epoch": 7.957521093977306, + "flow_matching_loss": 0.012786070443689823, + "grad_norm": 0.1185225173830986, + "learning_rate": 1.840191682797812e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004022955591790378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016858313465490937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009731742553412914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5692852437496185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27350, + "weighted_flow_matching_loss": 0.012786070443689823, + "weighted_rkd_loss": 0.005540143558755517 + }, + { + "epoch": 7.9604306080884495, + "flow_matching_loss": 0.012230818625539541, + "grad_norm": 0.1255171149969101, + "learning_rate": 1.839386438784403e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038524942006915807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018290535314008593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009534041956067085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5690259456634521, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27360, + "weighted_flow_matching_loss": 0.012230818625539541, + "weighted_rkd_loss": 0.0054250848712399605 + }, + { + "epoch": 7.963340122199592, + "flow_matching_loss": 0.01058316845446825, + "grad_norm": 0.1342909187078476, + "learning_rate": 1.838581091674355e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004273029183968902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018729967065155505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010419055074453353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5687666237354279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27370, + "weighted_flow_matching_loss": 0.01058316845446825, + "weighted_rkd_loss": 0.005926043493673205 + }, + { + "epoch": 7.966249636310736, + "flow_matching_loss": 0.011933599365875125, + "grad_norm": 0.08372550457715988, + "learning_rate": 1.8377756417123108e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041812245268374685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019064504536800086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01026889942586422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5685073018074036, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27380, + "weighted_flow_matching_loss": 0.011933599365875125, + "weighted_rkd_loss": 0.005838025314733386 + }, + { + "epoch": 7.96915915042188, + "flow_matching_loss": 0.012610406521707774, + "grad_norm": 0.11393408477306366, + "learning_rate": 1.8369700891429455e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004122452577576041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017385867075063288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009983491851016879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5682479619979859, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27390, + "weighted_flow_matching_loss": 0.012610406521707774, + "weighted_rkd_loss": 0.00567311579361558 + }, + { + "epoch": 7.9720686645330225, + "flow_matching_loss": 0.011686190124601125, + "grad_norm": 0.1091187372803688, + "learning_rate": 1.8361644342109642e-05, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036798465298488736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00158983200090006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008949525095522404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5679885983467102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27400, + "weighted_flow_matching_loss": 0.011686190124601125, + "weighted_rkd_loss": 0.00508327609859407 + }, + { + "epoch": 7.974978178644166, + "flow_matching_loss": 0.012485071457922459, + "grad_norm": 0.11702258139848709, + "learning_rate": 1.8353586771611038e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044476112583652135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00191058493219316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010805807448923588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5677292168140411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27410, + "weighted_flow_matching_loss": 0.012485071457922459, + "weighted_rkd_loss": 0.006134686386212706 + }, + { + "epoch": 7.97788769275531, + "flow_matching_loss": 0.013750226702541113, + "grad_norm": 0.13641305267810822, + "learning_rate": 1.8345528182381314e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004167384165339172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001795519224833697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010130287567153572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5674698173999786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27420, + "weighted_flow_matching_loss": 0.013750226702541113, + "weighted_rkd_loss": 0.005748556088656187 + }, + { + "epoch": 7.980797206866454, + "flow_matching_loss": 0.011604152666404844, + "grad_norm": 0.1149100512266159, + "learning_rate": 1.8337468576868466e-05, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004066263185814023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018076821696013212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009940208587795495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5672104001045227, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27430, + "weighted_flow_matching_loss": 0.011604152666404844, + "weighted_rkd_loss": 0.005638211779296398 + }, + { + "epoch": 7.9837067209775965, + "flow_matching_loss": 0.011462199175730348, + "grad_norm": 0.12539923191070557, + "learning_rate": 1.8329407957520786e-05, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041526832152158025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018917636363767088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010197129985317587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5669509589672088, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27440, + "weighted_flow_matching_loss": 0.011462199175730348, + "weighted_rkd_loss": 0.005781248561106622 + }, + { + "epoch": 7.98661623508874, + "flow_matching_loss": 0.011902084806933999, + "grad_norm": 0.13257800042629242, + "learning_rate": 1.8321346326786872e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003968654363416135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016611549421213568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009598463820293546, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5666915059089661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27450, + "weighted_flow_matching_loss": 0.011902084806933999, + "weighted_rkd_loss": 0.00543928281404078 + }, + { + "epoch": 7.989525749199884, + "flow_matching_loss": 0.011674991529434919, + "grad_norm": 0.11324877291917801, + "learning_rate": 1.831328368711564e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038555639097467066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001644252985715866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00935538075864315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5664320409297943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27460, + "weighted_flow_matching_loss": 0.011674991529434919, + "weighted_rkd_loss": 0.005299184215255081 + }, + { + "epoch": 7.992435263311027, + "flow_matching_loss": 0.01230852659791708, + "grad_norm": 0.11296895891427994, + "learning_rate": 1.830522004095631e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037520661484450103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001621108513791114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009125240752473474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5661725401878357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27470, + "weighted_flow_matching_loss": 0.01230852659791708, + "weighted_rkd_loss": 0.005166479526087642 + }, + { + "epoch": 7.99534477742217, + "flow_matching_loss": 0.012200943380594253, + "grad_norm": 0.16382287442684174, + "learning_rate": 1.8297155390758394e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0045506984926760195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018242180114611984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010925615020096303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5659130275249481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27480, + "weighted_flow_matching_loss": 0.012200943380594253, + "weighted_rkd_loss": 0.006182901514694095 + }, + { + "epoch": 7.998254291533314, + "flow_matching_loss": 0.012517458479851484, + "grad_norm": 0.10967915505170822, + "learning_rate": 1.828908973897173e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039852165384218095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00166927328100428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009639706322923302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5656535029411316, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27490, + "weighted_flow_matching_loss": 0.012517458479851484, + "weighted_rkd_loss": 0.005452754977159202 + }, + { + "epoch": 8.001163805644458, + "flow_matching_loss": 0.010132744861766695, + "grad_norm": 0.09711921960115433, + "learning_rate": 1.828102308804645e-05, + "loss": 0.015, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003605455276556313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014010358543600887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008611946366727352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5653939545154572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27500, + "weighted_flow_matching_loss": 0.010132744861766695, + "weighted_rkd_loss": 0.004869134072214365 + }, + { + "epoch": 8.004073319755602, + "flow_matching_loss": 0.012445969693362712, + "grad_norm": 0.11552935093641281, + "learning_rate": 1.827295544043299e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004270287347026169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018667350406758488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010407309886068106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5651344001293183, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27510, + "weighted_flow_matching_loss": 0.012445969693362712, + "weighted_rkd_loss": 0.005881553189828992 + }, + { + "epoch": 8.006982833866743, + "flow_matching_loss": 0.012036665063351392, + "grad_norm": 0.11353741586208344, + "learning_rate": 1.826488679858207e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038138141855597494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001699166721664369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009326795069500803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5648748099803924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27520, + "weighted_flow_matching_loss": 0.012036665063351392, + "weighted_rkd_loss": 0.005268437834456563 + }, + { + "epoch": 8.009892347977887, + "flow_matching_loss": 0.012961410079151392, + "grad_norm": 0.13077373802661896, + "learning_rate": 1.8256817164944756e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004355827136896551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017654093913733958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010477063618600369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5646152079105378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27530, + "weighted_flow_matching_loss": 0.012961410079151392, + "weighted_rkd_loss": 0.005915517080575227 + }, + { + "epoch": 8.012801862089031, + "flow_matching_loss": 0.010417488170787692, + "grad_norm": 0.11293216049671173, + "learning_rate": 1.8248746541972377e-05, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004179252893663943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018322688119951636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010190774546936154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5643555998802186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27540, + "weighted_flow_matching_loss": 0.010417488170787692, + "weighted_rkd_loss": 0.005751189100556076 + }, + { + "epoch": 8.015711376200175, + "flow_matching_loss": 0.011952854972332717, + "grad_norm": 0.11252389848232269, + "learning_rate": 1.8240674932116575e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003670734423212707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015816648490726947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008923133742064238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5640959560871124, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27550, + "weighted_flow_matching_loss": 0.011952854972332717, + "weighted_rkd_loss": 0.0050334786530584095 + }, + { + "epoch": 8.018620890311318, + "flow_matching_loss": 0.012795516662299633, + "grad_norm": 0.13497582077980042, + "learning_rate": 1.8232602337829297e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004149731341749429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002067234553396702, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010366697050631046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5638363063335419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27560, + "weighted_flow_matching_loss": 0.012795516662299633, + "weighted_rkd_loss": 0.005845091445371509 + }, + { + "epoch": 8.021530404422462, + "flow_matching_loss": 0.013431584648787975, + "grad_norm": 0.10603304952383041, + "learning_rate": 1.8224528761562777e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041709610959514976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001734605839010328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010076528135687112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.563576626777649, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27570, + "weighted_flow_matching_loss": 0.013431584648787975, + "weighted_rkd_loss": 0.005678863497450948 + }, + { + "epoch": 8.024439918533606, + "flow_matching_loss": 0.011159949656575918, + "grad_norm": 0.12253927439451218, + "learning_rate": 1.821645420576956e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003955792146734893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016077277017757297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00951931206509471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.563316947221756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27580, + "weighted_flow_matching_loss": 0.011159949656575918, + "weighted_rkd_loss": 0.005362382857128978 + }, + { + "epoch": 8.027349432644748, + "flow_matching_loss": 0.010622299369424581, + "grad_norm": 0.13249178230762482, + "learning_rate": 1.820837867290248e-05, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036795435938984157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015043343417346477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00886342148296535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5630572438240051, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27590, + "weighted_flow_matching_loss": 0.010622299369424581, + "weighted_rkd_loss": 0.004990603844635189 + }, + { + "epoch": 8.030258946755891, + "flow_matching_loss": 0.013217673636972904, + "grad_norm": 0.12207143753767014, + "learning_rate": 1.820030216541467e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004520219983533025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019665235886350274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011006963532418012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5627975165843964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27600, + "weighted_flow_matching_loss": 0.013217673636972904, + "weighted_rkd_loss": 0.006194675946608186 + }, + { + "epoch": 8.033168460867035, + "flow_matching_loss": 0.01306612128391862, + "grad_norm": 0.12726011872291565, + "learning_rate": 1.8192224685759562e-05, + "loss": 0.0191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044486182508990165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018393402220681309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010736576840281486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5625377833843231, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27610, + "weighted_flow_matching_loss": 0.01306612128391862, + "weighted_rkd_loss": 0.006039789784699679 + }, + { + "epoch": 8.036077974978179, + "flow_matching_loss": 0.012034714594483376, + "grad_norm": 0.10306691378355026, + "learning_rate": 1.8184146236390883e-05, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003776705986820161, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015862320316955448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009139644028618932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.562278026342392, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27620, + "weighted_flow_matching_loss": 0.012034714594483376, + "weighted_rkd_loss": 0.005139002064242959 + }, + { + "epoch": 8.038987489089322, + "flow_matching_loss": 0.011096099810674786, + "grad_norm": 0.12274310737848282, + "learning_rate": 1.8176066819762644e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040334038436412815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001662326545920223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00972913447767496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5620182514190674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27630, + "weighted_flow_matching_loss": 0.011096099810674786, + "weighted_rkd_loss": 0.005467858258634805 + }, + { + "epoch": 8.041897003200466, + "flow_matching_loss": 0.011183625832200051, + "grad_norm": 0.1193549856543541, + "learning_rate": 1.8167986438329167e-05, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003724178811535239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016490039066411554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009097361657768487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5617584645748138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27640, + "weighted_flow_matching_loss": 0.011183625832200051, + "weighted_rkd_loss": 0.005110570369288326 + }, + { + "epoch": 8.044806517311608, + "flow_matching_loss": 0.01221402371302247, + "grad_norm": 0.1187860369682312, + "learning_rate": 1.815990509454506e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003998280875384808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016192771377973258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009615839086472988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5614986538887023, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27650, + "weighted_flow_matching_loss": 0.01221402371302247, + "weighted_rkd_loss": 0.005399380275048315 + }, + { + "epoch": 8.047716031422752, + "flow_matching_loss": 0.01251375423744321, + "grad_norm": 0.14391744136810303, + "learning_rate": 1.8151822790865218e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003687912644818425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001749036175897345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009124861564487219, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5612388253211975, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27660, + "weighted_flow_matching_loss": 0.01251375423744321, + "weighted_rkd_loss": 0.005121148424223065 + }, + { + "epoch": 8.050625545533896, + "flow_matching_loss": 0.013687949441373349, + "grad_norm": 0.1237754225730896, + "learning_rate": 1.8143739529744828e-05, + "loss": 0.0198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004482949315570295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002009542367886752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010975440917536616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5609790027141571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27670, + "weighted_flow_matching_loss": 0.013687949441373349, + "weighted_rkd_loss": 0.006156914541497827 + }, + { + "epoch": 8.05353505964504, + "flow_matching_loss": 0.011843278910964728, + "grad_norm": 0.09872480481863022, + "learning_rate": 1.813565531363937e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040700002806261185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001709753682371229, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009849754069000483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5607191324234009, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27680, + "weighted_flow_matching_loss": 0.011843278910964728, + "weighted_rkd_loss": 0.005523009900934994 + }, + { + "epoch": 8.056444573756183, + "flow_matching_loss": 0.010597552219405771, + "grad_norm": 0.1550564467906952, + "learning_rate": 1.8127570145004628e-05, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035426239715889095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015085193095728754, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008593767415732145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5604592561721802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27690, + "weighted_flow_matching_loss": 0.010597552219405771, + "weighted_rkd_loss": 0.004816431412473321 + }, + { + "epoch": 8.059354087867327, + "flow_matching_loss": 0.011188855860382318, + "grad_norm": 0.10378754138946533, + "learning_rate": 1.811948402629665e-05, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004128855164162814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016744579887017607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009932168293744326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5601993799209595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27700, + "weighted_flow_matching_loss": 0.011188855860382318, + "weighted_rkd_loss": 0.005563995707780123 + }, + { + "epoch": 8.06226360197847, + "flow_matching_loss": 0.012422305811196566, + "grad_norm": 0.10029447823762894, + "learning_rate": 1.811139695997179e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037413550075143577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001541269407607615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009023979352787138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5599394500255584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27710, + "weighted_flow_matching_loss": 0.012422305811196566, + "weighted_rkd_loss": 0.005052902828902006 + }, + { + "epoch": 8.065173116089612, + "flow_matching_loss": 0.012980118952691555, + "grad_norm": 0.0977218896150589, + "learning_rate": 1.810330894848669e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036504748743027447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001546783692901954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008847733354195952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5596795380115509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27720, + "weighted_flow_matching_loss": 0.012980118952691555, + "weighted_rkd_loss": 0.00495193381793797 + }, + { + "epoch": 8.068082630200756, + "flow_matching_loss": 0.011429806612432003, + "grad_norm": 0.12589691579341888, + "learning_rate": 1.8095219994298266e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003708299179561436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016731285373680293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009089726954698563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.559419596195221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27730, + "weighted_flow_matching_loss": 0.011429806612432003, + "weighted_rkd_loss": 0.005084990989416838 + }, + { + "epoch": 8.0709921443119, + "flow_matching_loss": 0.012300158012658357, + "grad_norm": 0.13636426627635956, + "learning_rate": 1.8087130099863728e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003814078704454005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015898236306384206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009217981016263366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.559159642457962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27740, + "weighted_flow_matching_loss": 0.012300158012658357, + "weighted_rkd_loss": 0.005154242319986224 + }, + { + "epoch": 8.073901658423043, + "flow_matching_loss": 0.010659575648605823, + "grad_norm": 0.11031531542539597, + "learning_rate": 1.8079039267640583e-05, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035365607356652616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014706430374644697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008543764520436526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5588996708393097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27750, + "weighted_flow_matching_loss": 0.010659575648605823, + "weighted_rkd_loss": 0.004775117500685155 + }, + { + "epoch": 8.076811172534187, + "flow_matching_loss": 0.013621837366372347, + "grad_norm": 0.10253437608480453, + "learning_rate": 1.80709475000866e-05, + "loss": 0.0193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004116492462344468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019142847973853349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010147269861772657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5586396753787994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27760, + "weighted_flow_matching_loss": 0.013621837366372347, + "weighted_rkd_loss": 0.005668621649965644 + }, + { + "epoch": 8.07972068664533, + "flow_matching_loss": 0.012795585207641125, + "grad_norm": 0.12717510759830475, + "learning_rate": 1.8062854799659855e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004384104744531214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019666312728077173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010734840668737889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5583796679973603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27770, + "weighted_flow_matching_loss": 0.012795585207641125, + "weighted_rkd_loss": 0.00599416196346283 + }, + { + "epoch": 8.082630200756475, + "flow_matching_loss": 0.01442401371896267, + "grad_norm": 0.13657955825328827, + "learning_rate": 1.805476116881869e-05, + "loss": 0.0196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003826107853092253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016034268075600267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009255642537027597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5581196546554565, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27780, + "weighted_flow_matching_loss": 0.01442401371896267, + "weighted_rkd_loss": 0.005165650183334947 + }, + { + "epoch": 8.085539714867616, + "flow_matching_loss": 0.011293888930231333, + "grad_norm": 0.1593467891216278, + "learning_rate": 1.804666661002174e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004225792642682791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019107065862044692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010362291894853114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5578596115112304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27790, + "weighted_flow_matching_loss": 0.011293888930231333, + "weighted_rkd_loss": 0.005780685925856233 + }, + { + "epoch": 8.08844922897876, + "flow_matching_loss": 0.010944255348294973, + "grad_norm": 0.09339550882577896, + "learning_rate": 1.8038571125727917e-05, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037105648312717675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001637865975499153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009058995731174945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5575995564460754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27800, + "weighted_flow_matching_loss": 0.010944255348294973, + "weighted_rkd_loss": 0.005051280604675412 + }, + { + "epoch": 8.091358743089904, + "flow_matching_loss": 0.012591219041496515, + "grad_norm": 0.10587174445390701, + "learning_rate": 1.8030474718396414e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003565840458031744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015430651954375207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008674746099859475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5573395013809204, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27810, + "weighted_flow_matching_loss": 0.012591219041496515, + "weighted_rkd_loss": 0.004834742029197514 + }, + { + "epoch": 8.094268257201048, + "flow_matching_loss": 0.012045336468145252, + "grad_norm": 0.11213185638189316, + "learning_rate": 1.8022377390486704e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038958174642175435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018823485355824233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009673983417451381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.557079404592514, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27820, + "weighted_flow_matching_loss": 0.012045336468145252, + "weighted_rkd_loss": 0.005389161896891892 + }, + { + "epoch": 8.097177771312191, + "flow_matching_loss": 0.011660795565694571, + "grad_norm": 0.14101503789424896, + "learning_rate": 1.8014279144458546e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003683337778784335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016410276293754578, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00900770309381187, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5568193078041077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27830, + "weighted_flow_matching_loss": 0.011660795565694571, + "weighted_rkd_loss": 0.005015593999996781 + }, + { + "epoch": 8.100087285423335, + "flow_matching_loss": 0.013386308308690786, + "grad_norm": 0.10218780487775803, + "learning_rate": 1.800617998277197e-05, + "loss": 0.0189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004111225437372923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017277708393521607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009950221609324216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5565591931343079, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27840, + "weighted_flow_matching_loss": 0.013386308308690786, + "weighted_rkd_loss": 0.005537881609052419 + }, + { + "epoch": 8.102996799534477, + "flow_matching_loss": 0.012341055506840348, + "grad_norm": 0.1285756528377533, + "learning_rate": 1.799807990788729e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00407922484446317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001595665269996971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009754114923998714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5562990605831146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27850, + "weighted_flow_matching_loss": 0.012341055506840348, + "weighted_rkd_loss": 0.005426220363005996 + }, + { + "epoch": 8.10590631364562, + "flow_matching_loss": 0.011721122218295932, + "grad_norm": 0.11209681630134583, + "learning_rate": 1.7989978922265094e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034455870976671577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015183306066319346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008409504825249315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5560389161109924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27860, + "weighted_flow_matching_loss": 0.011721122218295932, + "weighted_rkd_loss": 0.004676061193458736 + }, + { + "epoch": 8.108815827756764, + "flow_matching_loss": 0.012718487903475762, + "grad_norm": 0.13431943953037262, + "learning_rate": 1.7981877028366248e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004020783235318959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016807273146696388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009722293727099896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5557787418365479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27870, + "weighted_flow_matching_loss": 0.012718487903475762, + "weighted_rkd_loss": 0.0054035026114434 + }, + { + "epoch": 8.111725341867908, + "flow_matching_loss": 0.01149501372128725, + "grad_norm": 0.11206748336553574, + "learning_rate": 1.797377422865189e-05, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034135050140321256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014583602431230247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008285370282828807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5555185675621033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27880, + "weighted_flow_matching_loss": 0.01149501372128725, + "weighted_rkd_loss": 0.004602646478451789 + }, + { + "epoch": 8.114634855979052, + "flow_matching_loss": 0.011836744192987681, + "grad_norm": 0.10433812439441681, + "learning_rate": 1.796567052558344e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003861149377189577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001572028116788715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00929432692937553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5552583813667298, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27890, + "weighted_flow_matching_loss": 0.011836744192987681, + "weighted_rkd_loss": 0.005160730751231313 + }, + { + "epoch": 8.117544370090195, + "flow_matching_loss": 0.011958135478198528, + "grad_norm": 0.10905303061008453, + "learning_rate": 1.795756592162259e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037877146154642106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017500186106190086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009325447911396622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5549981772899628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27900, + "weighted_flow_matching_loss": 0.011958135478198528, + "weighted_rkd_loss": 0.005175677570514381 + }, + { + "epoch": 8.12045388420134, + "flow_matching_loss": 0.012375139351934194, + "grad_norm": 0.13197261095046997, + "learning_rate": 1.79494604192313e-05, + "loss": 0.0183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004390237829647958, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019448970211669802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010725372796878218, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5547379553318024, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27910, + "weighted_flow_matching_loss": 0.012375139351934194, + "weighted_rkd_loss": 0.0059496574103832245 + }, + { + "epoch": 8.123363398312481, + "flow_matching_loss": 0.010695005441084504, + "grad_norm": 0.13215696811676025, + "learning_rate": 1.7941354020871815e-05, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038276692852377892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016288356389850378, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00928417439572513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5544777154922486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27920, + "weighted_flow_matching_loss": 0.010695005441084504, + "weighted_rkd_loss": 0.005147755215875805 + }, + { + "epoch": 8.126272912423625, + "flow_matching_loss": 0.012088697031140327, + "grad_norm": 0.12445273995399475, + "learning_rate": 1.793324672900664e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036754487780854108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014486731495708227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0087995707988739, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5542174577713013, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27930, + "weighted_flow_matching_loss": 0.012088697031140327, + "weighted_rkd_loss": 0.004876897064968943 + }, + { + "epoch": 8.129182426534769, + "flow_matching_loss": 0.0126408358104527, + "grad_norm": 0.14370501041412354, + "learning_rate": 1.7925138546098553e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036649957532063127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015507991774939, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008880790695548058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.553957200050354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27940, + "weighted_flow_matching_loss": 0.0126408358104527, + "weighted_rkd_loss": 0.004919571499340236 + }, + { + "epoch": 8.132091940645912, + "flow_matching_loss": 0.012229628581553697, + "grad_norm": 0.12022142112255096, + "learning_rate": 1.791702947461061e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004055426362901926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017339650774374604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00984481773339212, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5536969125270843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27950, + "weighted_flow_matching_loss": 0.012229628581553697, + "weighted_rkd_loss": 0.005451138969510794 + }, + { + "epoch": 8.135001454757056, + "flow_matching_loss": 0.011073242733255028, + "grad_norm": 0.14580439031124115, + "learning_rate": 1.7908919517006135e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004097060882486403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001710532244760543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009904653998091817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5534366130828857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27960, + "weighted_flow_matching_loss": 0.011073242733255028, + "weighted_rkd_loss": 0.005481581832282245 + }, + { + "epoch": 8.1379109688682, + "flow_matching_loss": 0.011432171007618308, + "grad_norm": 0.11806945502758026, + "learning_rate": 1.7900808675748714e-05, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003373690601438284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014287890400737524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008176170103251935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5531763017177582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27970, + "weighted_flow_matching_loss": 0.011432171007618308, + "weighted_rkd_loss": 0.004522868664935231 + }, + { + "epoch": 8.140820482979342, + "flow_matching_loss": 0.01222587488591671, + "grad_norm": 0.10480811446905136, + "learning_rate": 1.7892696953302214e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003525823703967035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015319131780415773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008583560539409518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5529159724712371, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27980, + "weighted_flow_matching_loss": 0.01222587488591671, + "weighted_rkd_loss": 0.004745987872593105 + }, + { + "epoch": 8.143729997090485, + "flow_matching_loss": 0.013407604955136775, + "grad_norm": 0.11971595138311386, + "learning_rate": 1.7884584352130747e-05, + "loss": 0.019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041370243998244405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017856152961030602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010059664119035005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5526556372642517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27990, + "weighted_flow_matching_loss": 0.013407604955136775, + "weighted_rkd_loss": 0.005559427849948406 + }, + { + "epoch": 8.146639511201629, + "flow_matching_loss": 0.013057788368314505, + "grad_norm": 0.1662130504846573, + "learning_rate": 1.7876470874698722e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038237297907471658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015848055714741348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00923226522281766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5523952841758728, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28000, + "weighted_flow_matching_loss": 0.013057788368314505, + "weighted_rkd_loss": 0.005099832243286073 + }, + { + "epoch": 8.149549025312773, + "flow_matching_loss": 0.012430741172283887, + "grad_norm": 0.10113536566495895, + "learning_rate": 1.7868356523470796e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003816726175136864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017337380442768336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009367190301418304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5521349132061004, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28010, + "weighted_flow_matching_loss": 0.012430741172283887, + "weighted_rkd_loss": 0.0051719947718083855 + }, + { + "epoch": 8.152458539423916, + "flow_matching_loss": 0.013072058092802763, + "grad_norm": 0.14104531705379486, + "learning_rate": 1.7860241300911895e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004043372697196901, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017178805428557098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009804625995457172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5518745243549347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28020, + "weighted_flow_matching_loss": 0.013072058092802763, + "weighted_rkd_loss": 0.005410867789760232 + }, + { + "epoch": 8.15536805353506, + "flow_matching_loss": 0.01205417150631547, + "grad_norm": 0.10946214199066162, + "learning_rate": 1.7852125209487202e-05, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003535325080156326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015697089256718755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008640359109267592, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5516141295433045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28030, + "weighted_flow_matching_loss": 0.01205417150631547, + "weighted_rkd_loss": 0.004766172543168068 + }, + { + "epoch": 8.158277567646204, + "flow_matching_loss": 0.013412843504920603, + "grad_norm": 0.12161624431610107, + "learning_rate": 1.7844008251662175e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038205596385523675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016006711055524647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00924179032444954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5513537228107452, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28040, + "weighted_flow_matching_loss": 0.013412843504920603, + "weighted_rkd_loss": 0.0050955359824001786 + }, + { + "epoch": 8.161187081757346, + "flow_matching_loss": 0.012153738783672452, + "grad_norm": 0.11435321718454361, + "learning_rate": 1.7835890429902535e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003982264781370759, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018078705994412302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009772400092333555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5510932981967926, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28050, + "weighted_flow_matching_loss": 0.012153738783672452, + "weighted_rkd_loss": 0.005385546339675784 + }, + { + "epoch": 8.16409659586849, + "flow_matching_loss": 0.012081556767225266, + "grad_norm": 0.1265096515417099, + "learning_rate": 1.7827771746674257e-05, + "loss": 0.0172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003828868269920349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015990015002898873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009256738051772118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5508328378200531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28060, + "weighted_flow_matching_loss": 0.012081556767225266, + "weighted_rkd_loss": 0.005098877847194672 + }, + { + "epoch": 8.167006109979633, + "flow_matching_loss": 0.012607546895742417, + "grad_norm": 0.14824441075325012, + "learning_rate": 1.7819652204443593e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004139195266179741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017352920491248368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01001368248835206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5505724072456359, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28070, + "weighted_flow_matching_loss": 0.012607546895742417, + "weighted_rkd_loss": 0.0055133656598627566 + }, + { + "epoch": 8.169915624090777, + "flow_matching_loss": 0.014478034898638725, + "grad_norm": 0.1345762312412262, + "learning_rate": 1.7811531805677033e-05, + "loss": 0.0195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037661694455891847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016377199441194533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009170058881863952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.550311952829361, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28080, + "weighted_flow_matching_loss": 0.014478034898638725, + "weighted_rkd_loss": 0.005046393582597375 + }, + { + "epoch": 8.17282513820192, + "flow_matching_loss": 0.011821899749338626, + "grad_norm": 0.11089836061000824, + "learning_rate": 1.7803410552841346e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004247768898494542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017899902537465095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010285528004169464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5500514388084412, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28090, + "weighted_flow_matching_loss": 0.011821899749338626, + "weighted_rkd_loss": 0.005657493602484465 + }, + { + "epoch": 8.175734652313064, + "flow_matching_loss": 0.012569843232631684, + "grad_norm": 0.1395169198513031, + "learning_rate": 1.7795288448403548e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036004109540954234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001423413585871458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008624235494062304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5497909486293793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28100, + "weighted_flow_matching_loss": 0.012569843232631684, + "weighted_rkd_loss": 0.004741471284069121 + }, + { + "epoch": 8.178644166424208, + "flow_matching_loss": 0.011572767700999975, + "grad_norm": 0.13616009056568146, + "learning_rate": 1.7787165494830928e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004037517379038036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018061437527649104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009881178522482515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.549530440568924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28110, + "weighted_flow_matching_loss": 0.011572767700999975, + "weighted_rkd_loss": 0.005430000927299261 + }, + { + "epoch": 8.18155368053535, + "flow_matching_loss": 0.011825356585904955, + "grad_norm": 0.13579709827899933, + "learning_rate": 1.7779041694591024e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004356179037131369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002162000897806138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01087435893714428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5492699205875397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28120, + "weighted_flow_matching_loss": 0.011825356585904955, + "weighted_rkd_loss": 0.005972962593659759 + }, + { + "epoch": 8.184463194646494, + "flow_matching_loss": 0.013805798999965192, + "grad_norm": 0.12259112298488617, + "learning_rate": 1.7770917050151628e-05, + "loss": 0.0194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004084469377994537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019476060522720217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010116544784978032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.549009382724762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28130, + "weighted_flow_matching_loss": 0.013805798999965192, + "weighted_rkd_loss": 0.005554074048995971 + }, + { + "epoch": 8.187372708757637, + "flow_matching_loss": 0.01166987307369709, + "grad_norm": 0.14405904710292816, + "learning_rate": 1.7762791563980795e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004001856362447142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001648509514052421, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009652222180739045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5487488389015198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28140, + "weighted_flow_matching_loss": 0.01166987307369709, + "weighted_rkd_loss": 0.005296623869799078 + }, + { + "epoch": 8.190282222868781, + "flow_matching_loss": 0.013878711499273777, + "grad_norm": 0.1343349814414978, + "learning_rate": 1.7754665238546828e-05, + "loss": 0.0192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003963335696607828, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016932484577409923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009619919816032052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5484882652759552, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28150, + "weighted_flow_matching_loss": 0.013878711499273777, + "weighted_rkd_loss": 0.005276468489319086 + }, + { + "epoch": 8.193191736979925, + "flow_matching_loss": 0.012669370044022799, + "grad_norm": 0.1422097086906433, + "learning_rate": 1.77465380763183e-05, + "loss": 0.0186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0044424392515793444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001977822929620743, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010862701386213303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5482276916503906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28160, + "weighted_flow_matching_loss": 0.012669370044022799, + "weighted_rkd_loss": 0.005955157219432294 + }, + { + "epoch": 8.196101251091068, + "flow_matching_loss": 0.013287305273115635, + "grad_norm": 0.12323345243930817, + "learning_rate": 1.773841007976402e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00340124003123492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014485560939647257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008251036098226906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5479671061038971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28170, + "weighted_flow_matching_loss": 0.013287305273115635, + "weighted_rkd_loss": 0.004521305719390511 + }, + { + "epoch": 8.19901076520221, + "flow_matching_loss": 0.011816492164507508, + "grad_norm": 0.13489505648612976, + "learning_rate": 1.773028125135306e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038813760969787837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016375655890442432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00940031772479415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5477065026760102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28180, + "weighted_flow_matching_loss": 0.011816492164507508, + "weighted_rkd_loss": 0.005148583743721247 + }, + { + "epoch": 8.201920279313354, + "flow_matching_loss": 0.013038720656186342, + "grad_norm": 0.11854919791221619, + "learning_rate": 1.7722151593554743e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036503458861261604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016326928394846617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008933384716510773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5474458873271942, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28190, + "weighted_flow_matching_loss": 0.013038720656186342, + "weighted_rkd_loss": 0.0048904427327215675 + }, + { + "epoch": 8.204829793424498, + "flow_matching_loss": 0.011760516185313463, + "grad_norm": 0.12885773181915283, + "learning_rate": 1.7714021108838652e-05, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003809648659080267, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016825485625304281, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009301845822483302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5471852600574494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28200, + "weighted_flow_matching_loss": 0.011760516185313463, + "weighted_rkd_loss": 0.005089817498810589 + }, + { + "epoch": 8.207739307535642, + "flow_matching_loss": 0.010845315549522639, + "grad_norm": 0.11588593572378159, + "learning_rate": 1.77058897996746e-05, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003794857230968773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015357067924924195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009125421242788433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5469246149063111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28210, + "weighted_flow_matching_loss": 0.010845315549522639, + "weighted_rkd_loss": 0.004990897001698613 + }, + { + "epoch": 8.210648821646785, + "flow_matching_loss": 0.011658692266792059, + "grad_norm": 0.1152595728635788, + "learning_rate": 1.7697757668532677e-05, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033880980918183923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014336724299937487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008209868520498275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5466639578342438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28220, + "weighted_flow_matching_loss": 0.011658692266792059, + "weighted_rkd_loss": 0.0044880366884171964 + }, + { + "epoch": 8.213558335757929, + "flow_matching_loss": 0.011676200199872255, + "grad_norm": 0.13044504821300507, + "learning_rate": 1.7689624717883204e-05, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003545553027652204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014515323331579567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008542638458311557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5464032888412476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28230, + "weighted_flow_matching_loss": 0.011676200199872255, + "weighted_rkd_loss": 0.004667728766798973 + }, + { + "epoch": 8.216467849869073, + "flow_matching_loss": 0.01278136931359768, + "grad_norm": 0.10040982812643051, + "learning_rate": 1.768149095019676e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003504279744811356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014123917324468493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008420951152220368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5461426138877868, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28240, + "weighted_flow_matching_loss": 0.01278136931359768, + "weighted_rkd_loss": 0.0045991417486220595 + }, + { + "epoch": 8.219377363980215, + "flow_matching_loss": 0.010648526344448328, + "grad_norm": 0.10177077353000641, + "learning_rate": 1.7673356367944162e-05, + "loss": 0.0148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003138518682681024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013534619938582182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007630499405786395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5458819210529328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28250, + "weighted_flow_matching_loss": 0.010648526344448328, + "weighted_rkd_loss": 0.004165369388647378 + }, + { + "epoch": 8.222286878091358, + "flow_matching_loss": 0.010035679815337062, + "grad_norm": 0.09697581827640533, + "learning_rate": 1.7665220973596484e-05, + "loss": 0.0144, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033644892275333403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013557500089518726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008084728522226215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5456212103366852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28260, + "weighted_flow_matching_loss": 0.010035679815337062, + "weighted_rkd_loss": 0.004411249281838536 + }, + { + "epoch": 8.225196392202502, + "flow_matching_loss": 0.011768645793199538, + "grad_norm": 0.11313635855913162, + "learning_rate": 1.7657084769625046e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003653599671088159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001572185882832855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008879385190084577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5453604936599732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28270, + "weighted_flow_matching_loss": 0.011768645793199538, + "weighted_rkd_loss": 0.004842483787797391 + }, + { + "epoch": 8.228105906313646, + "flow_matching_loss": 0.01161558199673891, + "grad_norm": 0.09432046860456467, + "learning_rate": 1.7648947758501414e-05, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0031736820470541715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013565641769673676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007703928230330348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5450997650623322, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28280, + "weighted_flow_matching_loss": 0.01161558199673891, + "weighted_rkd_loss": 0.004199401917867362 + }, + { + "epoch": 8.23101542042479, + "flow_matching_loss": 0.012536035617813468, + "grad_norm": 0.09877294301986694, + "learning_rate": 1.7640809942697394e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003822256554849446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016626775031909347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00930719063617289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5448390245437622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28290, + "weighted_flow_matching_loss": 0.012536035617813468, + "weighted_rkd_loss": 0.005070895841345191 + }, + { + "epoch": 8.233924934535933, + "flow_matching_loss": 0.012616968248039484, + "grad_norm": 0.11585503816604614, + "learning_rate": 1.7632671324685037e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004603958455845714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021110827918164434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011318999715149403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5445782721042634, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28300, + "weighted_flow_matching_loss": 0.012616968248039484, + "weighted_rkd_loss": 0.0061641751090064645 + }, + { + "epoch": 8.236834448647077, + "flow_matching_loss": 0.012813945394009351, + "grad_norm": 0.10503078252077103, + "learning_rate": 1.7624531906936647e-05, + "loss": 0.0188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004499361664056778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019274109858088196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01092613423243165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5443175137042999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28310, + "weighted_flow_matching_loss": 0.012813945394009351, + "weighted_rkd_loss": 0.005947320954874158 + }, + { + "epoch": 8.239743962758219, + "flow_matching_loss": 0.010356832016259431, + "grad_norm": 0.11705781519412994, + "learning_rate": 1.761639169192476e-05, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003590276720933616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014522691606543958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008632822660729289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5440567135810852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28320, + "weighted_flow_matching_loss": 0.010356832016259431, + "weighted_rkd_loss": 0.004696780419908464 + }, + { + "epoch": 8.242653476869362, + "flow_matching_loss": 0.012338489294052124, + "grad_norm": 0.1390075832605362, + "learning_rate": 1.7608250682122157e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004259568359702826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017728164209984242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010291953105479479, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.543795931339264, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28330, + "weighted_flow_matching_loss": 0.012338489294052124, + "weighted_rkd_loss": 0.005596631346270442 + }, + { + "epoch": 8.245562990980506, + "flow_matching_loss": 0.011496973130851984, + "grad_norm": 0.09671258926391602, + "learning_rate": 1.760010888000186e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00373212608974427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015906325657851993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009054884780198336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5435351252555847, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28340, + "weighted_flow_matching_loss": 0.011496973130851984, + "weighted_rkd_loss": 0.004921718733385205 + }, + { + "epoch": 8.24847250509165, + "flow_matching_loss": 0.011659663822501897, + "grad_norm": 0.14191021025180817, + "learning_rate": 1.7591966288037142e-05, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003344994504004717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014104693487752228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008100458420813084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.543274313211441, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28350, + "weighted_flow_matching_loss": 0.011659663822501897, + "weighted_rkd_loss": 0.004400686523877084 + }, + { + "epoch": 8.251382019202794, + "flow_matching_loss": 0.011696252413094044, + "grad_norm": 0.10202650725841522, + "learning_rate": 1.75838229087015e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00374620845541358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001589543663430959, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009081960516050458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5430134892463684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28360, + "weighted_flow_matching_loss": 0.011696252413094044, + "weighted_rkd_loss": 0.0049315819051116705 + }, + { + "epoch": 8.254291533313937, + "flow_matching_loss": 0.011300201015546917, + "grad_norm": 0.09469927102327347, + "learning_rate": 1.7575678744468684e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003920472785830497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00161323162028566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009454177180305123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5427526473999024, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28370, + "weighted_flow_matching_loss": 0.011300201015546917, + "weighted_rkd_loss": 0.005131252435967326 + }, + { + "epoch": 8.25720104742508, + "flow_matching_loss": 0.013003051187843085, + "grad_norm": 0.10987228900194168, + "learning_rate": 1.756753379781267e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033590053906664254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016242517158389091, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008342262543737888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5424917936325073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28380, + "weighted_flow_matching_loss": 0.013003051187843085, + "weighted_rkd_loss": 0.004525602562353015 + }, + { + "epoch": 8.260110561536223, + "flow_matching_loss": 0.012048943433910608, + "grad_norm": 0.13396741449832916, + "learning_rate": 1.7559388071207682e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037496935576200483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001596899819560349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009096286818385124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5422309339046478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28390, + "weighted_flow_matching_loss": 0.012048943433910608, + "weighted_rkd_loss": 0.004932215088047087 + }, + { + "epoch": 8.263020075647367, + "flow_matching_loss": 0.011257670586928725, + "grad_norm": 0.1148805096745491, + "learning_rate": 1.7551241567128173e-05, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034251746255904436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001388027868233621, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008238377049565315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5419700622558594, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28400, + "weighted_flow_matching_loss": 0.011257670586928725, + "weighted_rkd_loss": 0.004464949783869088 + }, + { + "epoch": 8.26592958975851, + "flow_matching_loss": 0.012722550146281719, + "grad_norm": 0.13193978369235992, + "learning_rate": 1.7543094288048836e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004130265628919006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001920227613300085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010180758824571968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5417091727256775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28410, + "weighted_flow_matching_loss": 0.012722550146281719, + "weighted_rkd_loss": 0.0055150641594082115 + }, + { + "epoch": 8.268839103869654, + "flow_matching_loss": 0.011200186237692833, + "grad_norm": 0.11638514697551727, + "learning_rate": 1.7534946236444602e-05, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003350244788452983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014014792628586291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008101968746632337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5414482772350311, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28420, + "weighted_flow_matching_loss": 0.011200186237692833, + "weighted_rkd_loss": 0.00438678553327918 + }, + { + "epoch": 8.271748617980798, + "flow_matching_loss": 0.01099720373749733, + "grad_norm": 0.10730431228876114, + "learning_rate": 1.7526797414790633e-05, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003795518050901592, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001655528019182384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00924656419083476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5411873638629914, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28430, + "weighted_flow_matching_loss": 0.01099720373749733, + "weighted_rkd_loss": 0.005004108045250177 + }, + { + "epoch": 8.274658132091941, + "flow_matching_loss": 0.012383783794939518, + "grad_norm": 0.12275408953428268, + "learning_rate": 1.751864782556233e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004348524869419635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019820353714749216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010679085226729512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5409264445304871, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28440, + "weighted_flow_matching_loss": 0.012383783794939518, + "weighted_rkd_loss": 0.005776523496024311 + }, + { + "epoch": 8.277567646203083, + "flow_matching_loss": 0.011787464190274478, + "grad_norm": 0.0939917266368866, + "learning_rate": 1.7510497471235314e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003994498238898814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017741960939019919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009763192525133491, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5406655132770538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28450, + "weighted_flow_matching_loss": 0.011787464190274478, + "weighted_rkd_loss": 0.005278656026348472 + }, + { + "epoch": 8.280477160314227, + "flow_matching_loss": 0.012067575380206108, + "grad_norm": 0.11824415624141693, + "learning_rate": 1.7502346354285455e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004153051786124707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018610831582918763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010167186707258224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5404045760631562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28460, + "weighted_flow_matching_loss": 0.012067575380206108, + "weighted_rkd_loss": 0.005494438135065139 + }, + { + "epoch": 8.28338667442537, + "flow_matching_loss": 0.012056306283921004, + "grad_norm": 0.1109221875667572, + "learning_rate": 1.7494194477188838e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00330009029712528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013653263682499527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007965507078915834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.540143620967865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28470, + "weighted_flow_matching_loss": 0.012056306283921004, + "weighted_rkd_loss": 0.004302533599548042 + }, + { + "epoch": 8.286296188536515, + "flow_matching_loss": 0.011746387742459775, + "grad_norm": 0.10240652412176132, + "learning_rate": 1.74860418424218e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004071017843671143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018111875746399165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009953223261982202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5398826599121094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28480, + "weighted_flow_matching_loss": 0.011746387742459775, + "weighted_rkd_loss": 0.005373601382598281 + }, + { + "epoch": 8.289205702647658, + "flow_matching_loss": 0.012291453685611487, + "grad_norm": 0.16040858626365662, + "learning_rate": 1.7477888452460885e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003904713410884142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015940000885166228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009403426805511117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5396216809749603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28490, + "weighted_flow_matching_loss": 0.012291453685611487, + "weighted_rkd_loss": 0.005074194935150445 + }, + { + "epoch": 8.292115216758802, + "flow_matching_loss": 0.010941836470738053, + "grad_norm": 0.0917978286743164, + "learning_rate": 1.746973430978289e-05, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003493922995403409, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013297672965563833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008317613322287799, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5393606901168824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28500, + "weighted_flow_matching_loss": 0.010941836470738053, + "weighted_rkd_loss": 0.004486179212108254 + }, + { + "epoch": 8.295024730869944, + "flow_matching_loss": 0.011253710743039847, + "grad_norm": 0.13368529081344604, + "learning_rate": 1.7461579416864814e-05, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00372550745960325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015532066812738777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009004221670329571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5390997052192688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28510, + "weighted_flow_matching_loss": 0.011253710743039847, + "weighted_rkd_loss": 0.004854136309586465 + }, + { + "epoch": 8.297934244981088, + "flow_matching_loss": 0.012802305119112133, + "grad_norm": 0.11151257157325745, + "learning_rate": 1.7453423776183907e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003744300827383995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016270905500277877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009115692181512713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5388386905193329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28520, + "weighted_flow_matching_loss": 0.012802305119112133, + "weighted_rkd_loss": 0.004911886458285153 + }, + { + "epoch": 8.300843759092231, + "flow_matching_loss": 0.011727397050708532, + "grad_norm": 0.10471589118242264, + "learning_rate": 1.744526739021764e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037308072438463568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00148929808055982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008950912579894065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.538577675819397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28530, + "weighted_flow_matching_loss": 0.011727397050708532, + "weighted_rkd_loss": 0.004820747417397797 + }, + { + "epoch": 8.303753273203375, + "flow_matching_loss": 0.01191618456505239, + "grad_norm": 0.14237017929553986, + "learning_rate": 1.74371102614437e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003588140057399869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014889604528434575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008665240649133921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5383166551589966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28540, + "weighted_flow_matching_loss": 0.01191618456505239, + "weighted_rkd_loss": 0.004664677497930825 + }, + { + "epoch": 8.306662787314519, + "flow_matching_loss": 0.013021722342818976, + "grad_norm": 0.1312599927186966, + "learning_rate": 1.742895239234001e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038779553724452853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016533357789739966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009409246500581503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5380555987358093, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28550, + "weighted_flow_matching_loss": 0.013021722342818976, + "weighted_rkd_loss": 0.005062724254094065 + }, + { + "epoch": 8.309572301425662, + "flow_matching_loss": 0.012035626731812955, + "grad_norm": 0.13367408514022827, + "learning_rate": 1.742079378538472e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004093333147466183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016306808451190592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009817347303032875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5377945482730866, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28560, + "weighted_flow_matching_loss": 0.012035626731812955, + "weighted_rkd_loss": 0.005279631400480866 + }, + { + "epoch": 8.312481815536806, + "flow_matching_loss": 0.012687514908611775, + "grad_norm": 0.10656318068504333, + "learning_rate": 1.7412634443056192e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004061873606406152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018167833972256632, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009940530546009541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5375335216522217, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28570, + "weighted_flow_matching_loss": 0.012687514908611775, + "weighted_rkd_loss": 0.005343320383690298 + }, + { + "epoch": 8.315391329647948, + "flow_matching_loss": 0.012327193934470414, + "grad_norm": 0.11553794145584106, + "learning_rate": 1.7404474367833024e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0042437577154487375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017739662784151734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010261481581255793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5372724533081055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28580, + "weighted_flow_matching_loss": 0.012327193934470414, + "weighted_rkd_loss": 0.005513105471618473 + }, + { + "epoch": 8.318300843759092, + "flow_matching_loss": 0.012058603810146451, + "grad_norm": 0.14722567796707153, + "learning_rate": 1.739631356219403e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038642714731395244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015891047776676714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009317647852003574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5370113253593445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28590, + "weighted_flow_matching_loss": 0.012058603810146451, + "weighted_rkd_loss": 0.005003657634370029 + }, + { + "epoch": 8.321210357870235, + "flow_matching_loss": 0.012101463740691543, + "grad_norm": 0.12889592349529266, + "learning_rate": 1.7388152028618244e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038004158530384304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001622296217828989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009223127970471978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5367502570152283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28600, + "weighted_flow_matching_loss": 0.012101463740691543, + "weighted_rkd_loss": 0.00495057525113225 + }, + { + "epoch": 8.32411987198138, + "flow_matching_loss": 0.012298649363219737, + "grad_norm": 0.1371632069349289, + "learning_rate": 1.737998976958493e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004200694942846894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001958530570846051, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010359920468181372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5364891529083252, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28610, + "weighted_flow_matching_loss": 0.012298649363219737, + "weighted_rkd_loss": 0.005557833588682115 + }, + { + "epoch": 8.327029386092523, + "flow_matching_loss": 0.010305123776197434, + "grad_norm": 0.11435182392597198, + "learning_rate": 1.7371826787573562e-05, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038903261767700313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016900059883482755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009470658376812935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5362280488014222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28620, + "weighted_flow_matching_loss": 0.010305123776197434, + "weighted_rkd_loss": 0.00507850379217416 + }, + { + "epoch": 8.329938900203667, + "flow_matching_loss": 0.012215308099985122, + "grad_norm": 0.10749328881502151, + "learning_rate": 1.736366308506384e-05, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035878811730071904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015030999667942524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008678862173110247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5359669268131256, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28630, + "weighted_flow_matching_loss": 0.012215308099985122, + "weighted_rkd_loss": 0.004651575582101941 + }, + { + "epoch": 8.33284841431481, + "flow_matching_loss": 0.012641425523906947, + "grad_norm": 0.11786844581365585, + "learning_rate": 1.7355498664535682e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0041325647383928295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018125051050446928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010077634640038014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5357057929039002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28640, + "weighted_flow_matching_loss": 0.012641425523906947, + "weighted_rkd_loss": 0.005398649489507079 + }, + { + "epoch": 8.335757928425952, + "flow_matching_loss": 0.013508330192416906, + "grad_norm": 0.1483946591615677, + "learning_rate": 1.7347333528469225e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0031753325602039695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013723926735110582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007723057735711336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5354446530342102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28650, + "weighted_flow_matching_loss": 0.013508330192416906, + "weighted_rkd_loss": 0.004135247156955302 + }, + { + "epoch": 8.338667442537096, + "flow_matching_loss": 0.012780320830643177, + "grad_norm": 0.12649813294410706, + "learning_rate": 1.733916767934482e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038120792945846916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016262735356576742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009250432159751654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5351835072040558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28660, + "weighted_flow_matching_loss": 0.012780320830643177, + "weighted_rkd_loss": 0.004950595856644213 + }, + { + "epoch": 8.34157695664824, + "flow_matching_loss": 0.012326352950185537, + "grad_norm": 0.10975487530231476, + "learning_rate": 1.7331001119643034e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004191186721436679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016383789596147835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010020752437412739, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5349223375320434, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28670, + "weighted_flow_matching_loss": 0.012326352950185537, + "weighted_rkd_loss": 0.005360371107235551 + }, + { + "epoch": 8.344486470759383, + "flow_matching_loss": 0.013337595760822296, + "grad_norm": 0.11733101308345795, + "learning_rate": 1.7322833851844653e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038491340121254323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017106507439166307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009408918721601368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5346611738204956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28680, + "weighted_flow_matching_loss": 0.013337595760822296, + "weighted_rkd_loss": 0.005030624172650278 + }, + { + "epoch": 8.347395984870527, + "flow_matching_loss": 0.01119828512892127, + "grad_norm": 0.11722900718450546, + "learning_rate": 1.731466587843068e-05, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034988828701898456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001467549952212721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00846531568095088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5343999922275543, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28690, + "weighted_flow_matching_loss": 0.01119828512892127, + "weighted_rkd_loss": 0.004523925296962261 + }, + { + "epoch": 8.35030549898167, + "flow_matching_loss": 0.01195892971009016, + "grad_norm": 0.11688727140426636, + "learning_rate": 1.7306497201882334e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034166966332122685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015748232952319085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008408216526731849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5341388046741485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28700, + "weighted_flow_matching_loss": 0.01195892971009016, + "weighted_rkd_loss": 0.0044911210658028725 + }, + { + "epoch": 8.353215013092813, + "flow_matching_loss": 0.01190285747870803, + "grad_norm": 0.10849541425704956, + "learning_rate": 1.7298327824681032e-05, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003396417712792754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014254848181735724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008218320133164526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5338776111602783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28710, + "weighted_flow_matching_loss": 0.01190285747870803, + "weighted_rkd_loss": 0.00438753894995898 + }, + { + "epoch": 8.356124527203956, + "flow_matching_loss": 0.011143300775438547, + "grad_norm": 0.12487281858921051, + "learning_rate": 1.729015774930842e-05, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00394661114551127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001643426134251058, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009536648448556662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5336164116859436, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28720, + "weighted_flow_matching_loss": 0.011143300775438547, + "weighted_rkd_loss": 0.0050889563979581 + }, + { + "epoch": 8.3590340413151, + "flow_matching_loss": 0.01172167668119073, + "grad_norm": 0.1178264245390892, + "learning_rate": 1.7281986978246353e-05, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003499908884987235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001446514914277941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008446332579478622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5333551883697509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28730, + "weighted_flow_matching_loss": 0.01172167668119073, + "weighted_rkd_loss": 0.004504885617643594 + }, + { + "epoch": 8.361943555426244, + "flow_matching_loss": 0.01140011679381132, + "grad_norm": 0.11422386765480042, + "learning_rate": 1.7273815513976894e-05, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034034091513603926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013971309992484749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008203949313610792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5330939650535583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28740, + "weighted_flow_matching_loss": 0.01140011679381132, + "weighted_rkd_loss": 0.0043734937440603975 + }, + { + "epoch": 8.364853069537387, + "flow_matching_loss": 0.013052857853472233, + "grad_norm": 0.16289499402046204, + "learning_rate": 1.7265643358982318e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003978735976852476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016236974042840302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00958116939291358, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5328327298164368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28750, + "weighted_flow_matching_loss": 0.013052857853472233, + "weighted_rkd_loss": 0.00510513405315578 + }, + { + "epoch": 8.367762583648531, + "flow_matching_loss": 0.010904861520975829, + "grad_norm": 0.11681164801120758, + "learning_rate": 1.7257470515745117e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004465272068046033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020296409842558206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010960184969007969, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5325714826583863, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28760, + "weighted_flow_matching_loss": 0.010904861520975829, + "weighted_rkd_loss": 0.005837057903409004 + }, + { + "epoch": 8.370672097759675, + "flow_matching_loss": 0.011028130166232586, + "grad_norm": 0.10639975965023041, + "learning_rate": 1.724929698674797e-05, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033583194948732854, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00150276857893914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008219407545402646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5323102355003357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28770, + "weighted_flow_matching_loss": 0.011028130166232586, + "weighted_rkd_loss": 0.004375304817222059 + }, + { + "epoch": 8.373581611870817, + "flow_matching_loss": 0.011865699803456664, + "grad_norm": 0.1173563078045845, + "learning_rate": 1.7241122774473788e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004000665992498398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016897548804990948, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00969108697026968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5320489704608917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28780, + "weighted_flow_matching_loss": 0.011865699803456664, + "weighted_rkd_loss": 0.005156125617213547 + }, + { + "epoch": 8.37649112598196, + "flow_matching_loss": 0.011511640436947346, + "grad_norm": 0.12018667161464691, + "learning_rate": 1.7232947881405684e-05, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003556042304262519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015332377515733242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00864532245323062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5317877054214477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28790, + "weighted_flow_matching_loss": 0.011511640436947346, + "weighted_rkd_loss": 0.004597425763495267 + }, + { + "epoch": 8.379400640093104, + "flow_matching_loss": 0.010784067353233695, + "grad_norm": 0.10735528916120529, + "learning_rate": 1.722477231002697e-05, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038573616882786156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014977234648540616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00921244672499597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5315264225006103, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28800, + "weighted_flow_matching_loss": 0.010784067353233695, + "weighted_rkd_loss": 0.004896696284413337 + }, + { + "epoch": 8.382310154204248, + "flow_matching_loss": 0.012413382530212402, + "grad_norm": 0.08989965915679932, + "learning_rate": 1.721659606282117e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003884263942018151, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016343831724952907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009402910992503167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.531265139579773, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28810, + "weighted_flow_matching_loss": 0.012413382530212402, + "weighted_rkd_loss": 0.004995367792434991 + }, + { + "epoch": 8.385219668315392, + "flow_matching_loss": 0.012065216433256865, + "grad_norm": 0.10528947412967682, + "learning_rate": 1.7208419142272012e-05, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003638075781054795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015848681796342134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00886101983487606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5310038447380065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28820, + "weighted_flow_matching_loss": 0.012065216433256865, + "weighted_rkd_loss": 0.004705290868878364 + }, + { + "epoch": 8.388129182426535, + "flow_matching_loss": 0.010011275112628937, + "grad_norm": 0.09821464121341705, + "learning_rate": 1.720024155086343e-05, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004068915382958949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016230598092079162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009760890668258072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5307425379753112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28830, + "weighted_flow_matching_loss": 0.010011275112628937, + "weighted_rkd_loss": 0.005180518189445138 + }, + { + "epoch": 8.391038696537677, + "flow_matching_loss": 0.011843134835362434, + "grad_norm": 0.11307995766401291, + "learning_rate": 1.7192063291079555e-05, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003936621826142073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015374596463516355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009410703368484975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5304812252521515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28840, + "weighted_flow_matching_loss": 0.011843134835362434, + "weighted_rkd_loss": 0.004992139362730086 + }, + { + "epoch": 8.393948210648821, + "flow_matching_loss": 0.011451030615717173, + "grad_norm": 0.12769781053066254, + "learning_rate": 1.7183884365404738e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038293357007205485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016389340453315526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009297605697065591, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5302199065685272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28850, + "weighted_flow_matching_loss": 0.011451030615717173, + "weighted_rkd_loss": 0.004929803940467536 + }, + { + "epoch": 8.396857724759965, + "flow_matching_loss": 0.010579572711139917, + "grad_norm": 0.09994084388017654, + "learning_rate": 1.7175704776323504e-05, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038626931607723238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017262609617318958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00945164724253118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5299585700035095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28860, + "weighted_flow_matching_loss": 0.010579572711139917, + "weighted_rkd_loss": 0.005009046383202076 + }, + { + "epoch": 8.399767238871108, + "flow_matching_loss": 0.012571576982736588, + "grad_norm": 0.12674206495285034, + "learning_rate": 1.7167524526320612e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038798874709755184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017396634095348418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009499438386410475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5296972393989563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28870, + "weighted_flow_matching_loss": 0.012571576982736588, + "weighted_rkd_loss": 0.005031806463375687 + }, + { + "epoch": 8.402676752982252, + "flow_matching_loss": 0.010859473049640656, + "grad_norm": 0.12330717593431473, + "learning_rate": 1.7159343617881e-05, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003935547056607902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016285046585835516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009499598760157824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5294358909130097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28880, + "weighted_flow_matching_loss": 0.010859473049640656, + "weighted_rkd_loss": 0.005029392894357443 + }, + { + "epoch": 8.405586267093396, + "flow_matching_loss": 0.012217167392373085, + "grad_norm": 0.1348506063222885, + "learning_rate": 1.71511620534898e-05, + "loss": 0.0176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004085492575541139, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001917456416413188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010088441707193852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5291745364665985, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28890, + "weighted_flow_matching_loss": 0.012217167392373085, + "weighted_rkd_loss": 0.005338453315198421 + }, + { + "epoch": 8.40849578120454, + "flow_matching_loss": 0.01224088091403246, + "grad_norm": 0.11302877217531204, + "learning_rate": 1.7142979835632376e-05, + "loss": 0.0179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004343697475269437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019412299618124962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010628624772652983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5289131700992584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28900, + "weighted_flow_matching_loss": 0.01224088091403246, + "weighted_rkd_loss": 0.0056215660646557804 + }, + { + "epoch": 8.411405295315681, + "flow_matching_loss": 0.009908679872751236, + "grad_norm": 0.1211889237165451, + "learning_rate": 1.7134796966794254e-05, + "loss": 0.015, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003954050410538912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017089407076127828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009617041423916816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5286518037319183, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28910, + "weighted_flow_matching_loss": 0.009908679872751236, + "weighted_rkd_loss": 0.005084090679883957 + }, + { + "epoch": 8.414314809426825, + "flow_matching_loss": 0.01260184971615672, + "grad_norm": 0.098661407828331, + "learning_rate": 1.7126613449461186e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003672009985893965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015788060030899943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008922825939953328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5283904373645782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28920, + "weighted_flow_matching_loss": 0.01260184971615672, + "weighted_rkd_loss": 0.004714793199673295 + }, + { + "epoch": 8.417224323537969, + "flow_matching_loss": 0.011478900164365768, + "grad_norm": 0.11701178550720215, + "learning_rate": 1.7118429286119098e-05, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037152070086449384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015020691324025392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008932483056560159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5281290411949158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28930, + "weighted_flow_matching_loss": 0.011478900164365768, + "weighted_rkd_loss": 0.004717495059594512 + }, + { + "epoch": 8.420133837649113, + "flow_matching_loss": 0.010783889889717102, + "grad_norm": 0.10279557853937149, + "learning_rate": 1.7110244479254126e-05, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003864703211002052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015441265306435526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00927353291772306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5278676509857178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28940, + "weighted_flow_matching_loss": 0.010783889889717102, + "weighted_rkd_loss": 0.0048951847478747364 + }, + { + "epoch": 8.423043351760256, + "flow_matching_loss": 0.01094059101305902, + "grad_norm": 0.12725071609020233, + "learning_rate": 1.71020590313526e-05, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00365125541575253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015486934804357587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008851204114034773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5276062488555908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28950, + "weighted_flow_matching_loss": 0.01094059101305902, + "weighted_rkd_loss": 0.004669903311878443 + }, + { + "epoch": 8.4259528658714, + "flow_matching_loss": 0.01155036841519177, + "grad_norm": 0.11946893483400345, + "learning_rate": 1.7093872944901044e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003968668496236205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001615277148084715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009552614111453295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5273448407649994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28960, + "weighted_flow_matching_loss": 0.01155036841519177, + "weighted_rkd_loss": 0.0050374764483422044 + }, + { + "epoch": 8.428862379982544, + "flow_matching_loss": 0.011406130529940129, + "grad_norm": 0.09503871202468872, + "learning_rate": 1.7085686222386172e-05, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003512944281101227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014518738025799394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008477762434631586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5270834267139435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28970, + "weighted_flow_matching_loss": 0.011406130529940129, + "weighted_rkd_loss": 0.004468500218354166 + }, + { + "epoch": 8.431771894093686, + "flow_matching_loss": 0.010840542986989022, + "grad_norm": 0.10776461660861969, + "learning_rate": 1.7077498866294895e-05, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004180777911096811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001730798836797476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010092354845255613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5268220007419586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28980, + "weighted_flow_matching_loss": 0.010840542986989022, + "weighted_rkd_loss": 0.005316973477602005 + }, + { + "epoch": 8.43468140820483, + "flow_matching_loss": 0.011028820462524892, + "grad_norm": 0.3179199695587158, + "learning_rate": 1.7069310879114313e-05, + "loss": 0.0155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035125073743984105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014547845581546425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0084797992836684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5265605747699738, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28990, + "weighted_flow_matching_loss": 0.011028820462524892, + "weighted_rkd_loss": 0.004465116909705103 + }, + { + "epoch": 8.437590922315973, + "flow_matching_loss": 0.01288790493272245, + "grad_norm": 0.1414519101381302, + "learning_rate": 1.7061122263331727e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037907524965703487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015487473690882325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009130252432078123, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5262991309165954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29000, + "weighted_flow_matching_loss": 0.01288790493272245, + "weighted_rkd_loss": 0.004805227275937796 + }, + { + "epoch": 8.440500436427117, + "flow_matching_loss": 0.01198449032381177, + "grad_norm": 0.10707300156354904, + "learning_rate": 1.7052933021434617e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003623149124905467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001702702243346721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008949000388383865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5260376930236816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29010, + "weighted_flow_matching_loss": 0.01198449032381177, + "weighted_rkd_loss": 0.0047075555892661216 + }, + { + "epoch": 8.44340995053826, + "flow_matching_loss": 0.01183855002745986, + "grad_norm": 0.13209709525108337, + "learning_rate": 1.7044743155910658e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003580577881075442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014718155027367175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008632971299812197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5257762372493744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29020, + "weighted_flow_matching_loss": 0.01183855002745986, + "weighted_rkd_loss": 0.004539060150273144 + }, + { + "epoch": 8.446319464649404, + "flow_matching_loss": 0.011271201632916927, + "grad_norm": 0.11277856677770615, + "learning_rate": 1.703655266924772e-05, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003558128117583692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015399672440253199, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008656223444268108, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5255147814750671, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29030, + "weighted_flow_matching_loss": 0.011271201632916927, + "weighted_rkd_loss": 0.004549032705835998 + }, + { + "epoch": 8.449228978760548, + "flow_matching_loss": 0.011457614414393903, + "grad_norm": 0.09091167151927948, + "learning_rate": 1.7028361563933846e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003948760102503002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016707799164578318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009568300284445285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.525253313779831, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29040, + "weighted_flow_matching_loss": 0.011457614414393903, + "weighted_rkd_loss": 0.0050258016213774685 + }, + { + "epoch": 8.45213849287169, + "flow_matching_loss": 0.011912946216762066, + "grad_norm": 0.11228177696466446, + "learning_rate": 1.702016984245729e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003768120030872524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015990366693586112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009135276731103658, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5249918401241302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29050, + "weighted_flow_matching_loss": 0.011912946216762066, + "weighted_rkd_loss": 0.004796025808900595 + }, + { + "epoch": 8.455048006982834, + "flow_matching_loss": 0.010838846722617746, + "grad_norm": 0.12214438617229462, + "learning_rate": 1.701197750730647e-05, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033435314195230602, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014084207417909057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00809548357501626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5247303605079651, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29060, + "weighted_flow_matching_loss": 0.010838846722617746, + "weighted_rkd_loss": 0.0042479472467675805 + }, + { + "epoch": 8.457957521093977, + "flow_matching_loss": 0.012611238565295935, + "grad_norm": 0.12700295448303223, + "learning_rate": 1.7003784560970013e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003711443115025759, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017084598424844445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009131346084177493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5244688749313354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29070, + "weighted_flow_matching_loss": 0.012611238565295935, + "weighted_rkd_loss": 0.004789162939414382 + }, + { + "epoch": 8.460867035205121, + "flow_matching_loss": 0.013071237411350013, + "grad_norm": 0.11889434605836868, + "learning_rate": 1.6995591005936714e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003835720825009048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016982851026114076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009369726665318013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5242073833942413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29080, + "weighted_flow_matching_loss": 0.013071237411350013, + "weighted_rkd_loss": 0.004911738820374012 + }, + { + "epoch": 8.463776549316265, + "flow_matching_loss": 0.01162552870810032, + "grad_norm": 0.1259576827287674, + "learning_rate": 1.6987396844695555e-05, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004175603482872247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016859513125382364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010037158336490392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5239458799362182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29090, + "weighted_flow_matching_loss": 0.01162552870810032, + "weighted_rkd_loss": 0.005258933873847127 + }, + { + "epoch": 8.466686063427408, + "flow_matching_loss": 0.01103509971871972, + "grad_norm": 0.1292671263217926, + "learning_rate": 1.6979202079735708e-05, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003708407492376864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014916853280737997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008908500289544463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5236843705177308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29100, + "weighted_flow_matching_loss": 0.01103509971871972, + "weighted_rkd_loss": 0.004665217665024102 + }, + { + "epoch": 8.46959557753855, + "flow_matching_loss": 0.010011422447860241, + "grad_norm": 0.12341785430908203, + "learning_rate": 1.697100671354653e-05, + "loss": 0.0147, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036709291860461237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016445599729195237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008986418321728706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5234228670597076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29110, + "weighted_flow_matching_loss": 0.010011422447860241, + "weighted_rkd_loss": 0.004703653417527676 + }, + { + "epoch": 8.472505091649694, + "flow_matching_loss": 0.013369739335030318, + "grad_norm": 0.12867408990859985, + "learning_rate": 1.6962810748617548e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037323371740058064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015726673416793346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009037341736257077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5231613337993621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29120, + "weighted_flow_matching_loss": 0.013369739335030318, + "weighted_rkd_loss": 0.004728011018596589 + }, + { + "epoch": 8.475414605760838, + "flow_matching_loss": 0.012252179067581892, + "grad_norm": 0.13746598362922668, + "learning_rate": 1.695461418743849e-05, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036218164721503856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014908452052623034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008734478196129204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5228998184204101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29130, + "weighted_flow_matching_loss": 0.012252179067581892, + "weighted_rkd_loss": 0.004567217500880361 + }, + { + "epoch": 8.478324119871981, + "flow_matching_loss": 0.012296954542398453, + "grad_norm": 0.10353224724531174, + "learning_rate": 1.6946417032499252e-05, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036753602558746933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001500435429625213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008851155964657664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5226382851600647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29140, + "weighted_flow_matching_loss": 0.012296954542398453, + "weighted_rkd_loss": 0.004625958460383117 + }, + { + "epoch": 8.481233633983125, + "flow_matching_loss": 0.01202009990811348, + "grad_norm": 0.08674901723861694, + "learning_rate": 1.6938219286289908e-05, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037643149262294173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016990339267067612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009227663883939386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5223767399787903, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29150, + "weighted_flow_matching_loss": 0.01202009990811348, + "weighted_rkd_loss": 0.004820357006974518 + }, + { + "epoch": 8.484143148094269, + "flow_matching_loss": 0.011799757927656173, + "grad_norm": 0.10112140327692032, + "learning_rate": 1.693002095130072e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004118766961619258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001822434552013874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010059968521818518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5221151947975159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29160, + "weighted_flow_matching_loss": 0.011799757927656173, + "weighted_rkd_loss": 0.005252444068901241 + }, + { + "epoch": 8.487052662205413, + "flow_matching_loss": 0.0109462377615273, + "grad_norm": 0.09877507388591766, + "learning_rate": 1.6921822030022132e-05, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003386452607810497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014379667642060666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008210871974006296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5218536496162415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29170, + "weighted_flow_matching_loss": 0.0109462377615273, + "weighted_rkd_loss": 0.004284798027947545 + }, + { + "epoch": 8.489962176316554, + "flow_matching_loss": 0.010587573889642954, + "grad_norm": 0.119660384953022, + "learning_rate": 1.691362252494475e-05, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035965964198112488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015748320263810455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008768024854362011, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5215920865535736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29180, + "weighted_flow_matching_loss": 0.010587573889642954, + "weighted_rkd_loss": 0.004573352914303541 + }, + { + "epoch": 8.492871690427698, + "flow_matching_loss": 0.011123346397653222, + "grad_norm": 0.11999751627445221, + "learning_rate": 1.690542243855937e-05, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004052952281199396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018285750644281506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009934479556977748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5213305294513703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29190, + "weighted_flow_matching_loss": 0.011123346397653222, + "weighted_rkd_loss": 0.005179050820879638 + }, + { + "epoch": 8.495781204538842, + "flow_matching_loss": 0.01159296976402402, + "grad_norm": 0.0947643294930458, + "learning_rate": 1.6897221773356963e-05, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034441100200638176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014314956846646965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008319715736433864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5210689604282379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29200, + "weighted_flow_matching_loss": 0.01159296976402402, + "weighted_rkd_loss": 0.004335130983963609 + }, + { + "epoch": 8.498690718649986, + "flow_matching_loss": 0.011700992798432707, + "grad_norm": 0.10024300962686539, + "learning_rate": 1.6889020531828676e-05, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003374133352190256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001400035829283297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008148302556946874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5208073854446411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29210, + "weighted_flow_matching_loss": 0.011700992798432707, + "weighted_rkd_loss": 0.004243671195581555 + }, + { + "epoch": 8.50160023276113, + "flow_matching_loss": 0.01080407933332026, + "grad_norm": 0.0992923453450203, + "learning_rate": 1.688081871646583e-05, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003603271977044642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015372620429843665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008743806043639779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5205458045005799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29220, + "weighted_flow_matching_loss": 0.01080407933332026, + "weighted_rkd_loss": 0.004551660409197212 + }, + { + "epoch": 8.504509746872273, + "flow_matching_loss": 0.01105934837833047, + "grad_norm": 0.10515929758548737, + "learning_rate": 1.6872616329759917e-05, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037064680131152273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014735936070792376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008886529738083482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5202842235565186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29230, + "weighted_flow_matching_loss": 0.01105934837833047, + "weighted_rkd_loss": 0.004623510502278804 + }, + { + "epoch": 8.507419260983415, + "flow_matching_loss": 0.012780412333086133, + "grad_norm": 0.11297722160816193, + "learning_rate": 1.6864413374202605e-05, + "loss": 0.0175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003739557764492929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016145303379744291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00909364577382803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5200226306915283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29240, + "weighted_flow_matching_loss": 0.012780412333086133, + "weighted_rkd_loss": 0.004728962294757366 + }, + { + "epoch": 8.510328775094559, + "flow_matching_loss": 0.012172635830938815, + "grad_norm": 0.11124370992183685, + "learning_rate": 1.685620985228574e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004067900846712292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017117233714088798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009847524994984268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5197610318660736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29250, + "weighted_flow_matching_loss": 0.012172635830938815, + "weighted_rkd_loss": 0.005118375620804727 + }, + { + "epoch": 8.513238289205702, + "flow_matching_loss": 0.013265698589384556, + "grad_norm": 0.10211177170276642, + "learning_rate": 1.6848005766501326e-05, + "loss": 0.0181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038712866837158797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016358697437681257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009378443099558353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5194994330406189, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29260, + "weighted_flow_matching_loss": 0.013265698589384556, + "weighted_rkd_loss": 0.004872219683602452 + }, + { + "epoch": 8.516147803316846, + "flow_matching_loss": 0.010555168986320496, + "grad_norm": 0.12561962008476257, + "learning_rate": 1.683980111934156e-05, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036621870007365943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014634199440479278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00878779380582273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5192378282546997, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29270, + "weighted_flow_matching_loss": 0.010555168986320496, + "weighted_rkd_loss": 0.004562907735817134 + }, + { + "epoch": 8.51905731742799, + "flow_matching_loss": 0.012999131809920073, + "grad_norm": 0.11629759520292282, + "learning_rate": 1.6831595913298787e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037961835972964765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016506321029737593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009242999320849776, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5189762115478516, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29280, + "weighted_flow_matching_loss": 0.012999131809920073, + "weighted_rkd_loss": 0.004796927655115723 + }, + { + "epoch": 8.521966831539133, + "flow_matching_loss": 0.01103389048948884, + "grad_norm": 0.10317876189947128, + "learning_rate": 1.682339015086554e-05, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0031913533806800843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014455119089689106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007828218722715974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5187145888805389, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29290, + "weighted_flow_matching_loss": 0.01103389048948884, + "weighted_rkd_loss": 0.004060602397657931 + }, + { + "epoch": 8.524876345650277, + "flow_matching_loss": 0.011418353812769056, + "grad_norm": 0.11804696172475815, + "learning_rate": 1.6815183834534507e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004701941530220211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019222549977712332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011326138023287058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5184529721736908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29300, + "weighted_flow_matching_loss": 0.011418353812769056, + "weighted_rkd_loss": 0.005872031440958381 + }, + { + "epoch": 8.52778585976142, + "flow_matching_loss": 0.013564590364694595, + "grad_norm": 0.11461450904607773, + "learning_rate": 1.680697696679855e-05, + "loss": 0.0184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003817249764688313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001605513843242079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009240013314411044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5181913495063781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29310, + "weighted_flow_matching_loss": 0.013564590364694595, + "weighted_rkd_loss": 0.0047880826285108926 + }, + { + "epoch": 8.530695373872563, + "flow_matching_loss": 0.012404267769306898, + "grad_norm": 0.15298107266426086, + "learning_rate": 1.6798769550150702e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036932995077222585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001636860566213727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009023459628224373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5179297029972076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29320, + "weighted_flow_matching_loss": 0.012404267769306898, + "weighted_rkd_loss": 0.004673403571359813 + }, + { + "epoch": 8.533604887983707, + "flow_matching_loss": 0.011550392396748066, + "grad_norm": 0.10188909620046616, + "learning_rate": 1.679056158708416e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004331159475259483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020031407242640854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010665459698066116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.517668080329895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29330, + "weighted_flow_matching_loss": 0.011550392396748066, + "weighted_rkd_loss": 0.0055212285602465275 + }, + { + "epoch": 8.53651440209485, + "flow_matching_loss": 0.011154014663770796, + "grad_norm": 0.11140982806682587, + "learning_rate": 1.6782353080092283e-05, + "loss": 0.0153, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032663657097145916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014222270052414387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007954958407208324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5174064338207245, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29340, + "weighted_flow_matching_loss": 0.011154014663770796, + "weighted_rkd_loss": 0.0041159266140311955 + }, + { + "epoch": 8.539423916205994, + "flow_matching_loss": 0.012041833903640508, + "grad_norm": 0.11720173805952072, + "learning_rate": 1.6774144031668603e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036557718645781278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017013938282616436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009012937685474754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5171447694301605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29350, + "weighted_flow_matching_loss": 0.012041833903640508, + "weighted_rkd_loss": 0.004661042802035808 + }, + { + "epoch": 8.542333430317138, + "flow_matching_loss": 0.012601515278220177, + "grad_norm": 0.1213986724615097, + "learning_rate": 1.6765934444306804e-05, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034083656733855604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014577156049199401, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008274446986615659, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5168831169605255, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29360, + "weighted_flow_matching_loss": 0.012601515278220177, + "weighted_rkd_loss": 0.0042769860476255415 + }, + { + "epoch": 8.545242944428281, + "flow_matching_loss": 0.013258678838610649, + "grad_norm": 0.10081489384174347, + "learning_rate": 1.6757724320500753e-05, + "loss": 0.018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00375974893104285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001618527399841696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009138025296851992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5166215002536774, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29370, + "weighted_flow_matching_loss": 0.013258678838610649, + "weighted_rkd_loss": 0.004720943165011704 + }, + { + "epoch": 8.548152458539423, + "flow_matching_loss": 0.012666859664022923, + "grad_norm": 0.11417841911315918, + "learning_rate": 1.674951366274447e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035563949728384613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014950352371670306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008607825171202422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5163598358631134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29380, + "weighted_flow_matching_loss": 0.012666859664022923, + "weighted_rkd_loss": 0.0044447075808420776 + }, + { + "epoch": 8.551061972650567, + "flow_matching_loss": 0.012043064087629318, + "grad_norm": 0.11302907019853592, + "learning_rate": 1.674130247353212e-05, + "loss": 0.0163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00342578764539212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001423425425309688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008275000797584652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5160981714725494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29390, + "weighted_flow_matching_loss": 0.012043064087629318, + "weighted_rkd_loss": 0.004270763718523085 + }, + { + "epoch": 8.55397148676171, + "flow_matching_loss": 0.010786883905529976, + "grad_norm": 0.10105515271425247, + "learning_rate": 1.6733090755358056e-05, + "loss": 0.015, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033633311511948706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015353873022831976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008262049639597535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5158364832401275, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29400, + "weighted_flow_matching_loss": 0.010786883905529976, + "weighted_rkd_loss": 0.004261843883432448 + }, + { + "epoch": 8.556881000872854, + "flow_matching_loss": 0.0109453191049397, + "grad_norm": 0.10537927597761154, + "learning_rate": 1.672487851071678e-05, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035315498244017363, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001557168090948835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008620267827063799, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5155747830867767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29410, + "weighted_flow_matching_loss": 0.0109453191049397, + "weighted_rkd_loss": 0.004444385669194162 + }, + { + "epoch": 8.559790514983998, + "flow_matching_loss": 0.012941953353583812, + "grad_norm": 0.09508627653121948, + "learning_rate": 1.6716665742102956e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003611130337230861, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013850633869878948, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008607324119657278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5153131186962128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29420, + "weighted_flow_matching_loss": 0.012941953353583812, + "weighted_rkd_loss": 0.004435446322895586 + }, + { + "epoch": 8.562700029095142, + "flow_matching_loss": 0.011922716116532683, + "grad_norm": 0.10194166749715805, + "learning_rate": 1.670845245201141e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003616066835820675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014865696197375656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008718703361228108, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5150514543056488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29430, + "weighted_flow_matching_loss": 0.011922716116532683, + "weighted_rkd_loss": 0.004490615730173886 + }, + { + "epoch": 8.565609543206284, + "flow_matching_loss": 0.012395379412919284, + "grad_norm": 0.10052067041397095, + "learning_rate": 1.670023864293711e-05, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0030533440178260205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0012350382341537625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007341726263985038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5147897481918335, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29440, + "weighted_flow_matching_loss": 0.012395379412919284, + "weighted_rkd_loss": 0.0037794855888932943 + }, + { + "epoch": 8.568519057317427, + "flow_matching_loss": 0.011775982100516558, + "grad_norm": 0.12685969471931458, + "learning_rate": 1.669202431737521e-05, + "loss": 0.0166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038182145915925505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017017673584632576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009338196460157632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5145280659198761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29450, + "weighted_flow_matching_loss": 0.011775982100516558, + "weighted_rkd_loss": 0.004804784106090665 + }, + { + "epoch": 8.571428571428571, + "flow_matching_loss": 0.011095537152141333, + "grad_norm": 0.1200067400932312, + "learning_rate": 1.6683809477820995e-05, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004040328087285161, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017420431599020959, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009822699334472417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5142663657665253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29460, + "weighted_flow_matching_loss": 0.011095537152141333, + "weighted_rkd_loss": 0.005051532294601202 + }, + { + "epoch": 8.574338085539715, + "flow_matching_loss": 0.010557331796735524, + "grad_norm": 0.12030191719532013, + "learning_rate": 1.6675594126769916e-05, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003886623401194811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017249871394596993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009498233953490854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5140046775341034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29470, + "weighted_flow_matching_loss": 0.010557331796735524, + "weighted_rkd_loss": 0.004882206209003925 + }, + { + "epoch": 8.577247599650859, + "flow_matching_loss": 0.010854111239314079, + "grad_norm": 0.12616416811943054, + "learning_rate": 1.6667378266717586e-05, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038326709531247617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016287950042169542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009294136893004179, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5137429594993591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29480, + "weighted_flow_matching_loss": 0.010854111239314079, + "weighted_rkd_loss": 0.0047747053438797595 + }, + { + "epoch": 8.580157113762002, + "flow_matching_loss": 0.012490166071802377, + "grad_norm": 0.14535707235336304, + "learning_rate": 1.665916190015976e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038947428576648234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017420061165466904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00953149190172553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5134812593460083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29490, + "weighted_flow_matching_loss": 0.012490166071802377, + "weighted_rkd_loss": 0.004894281993620098 + }, + { + "epoch": 8.583066627873146, + "flow_matching_loss": 0.012082244921475648, + "grad_norm": 0.13280300796031952, + "learning_rate": 1.665094502959236e-05, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038652915740385652, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015510784811340272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009281661594286561, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.513219553232193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29500, + "weighted_flow_matching_loss": 0.012082244921475648, + "weighted_rkd_loss": 0.004763548821210861 + }, + { + "epoch": 8.585976141984288, + "flow_matching_loss": 0.012520974082872272, + "grad_norm": 0.14087428152561188, + "learning_rate": 1.664272765751145e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003362050489522517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013284828048199416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008052583783864975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5129578351974488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29510, + "weighted_flow_matching_loss": 0.012520974082872272, + "weighted_rkd_loss": 0.00413065745960921 + }, + { + "epoch": 8.588885656095432, + "flow_matching_loss": 0.010708479909226298, + "grad_norm": 0.10893648862838745, + "learning_rate": 1.6634509786413248e-05, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036557268584147097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001540387119166553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00885184076614678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5126961171627045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29520, + "weighted_flow_matching_loss": 0.010708479909226298, + "weighted_rkd_loss": 0.004538316652178765 + }, + { + "epoch": 8.591795170206575, + "flow_matching_loss": 0.009976301668211817, + "grad_norm": 0.10960351675748825, + "learning_rate": 1.6626291418794135e-05, + "loss": 0.0153, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004261780832894147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019117637886665762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010435325372964144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5124343931674957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29530, + "weighted_flow_matching_loss": 0.009976301668211817, + "weighted_rkd_loss": 0.005347356502898038 + }, + { + "epoch": 8.594704684317719, + "flow_matching_loss": 0.011326869949698448, + "grad_norm": 0.12817764282226562, + "learning_rate": 1.6618072557150625e-05, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003436602861620486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013809502124786377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008254155935719609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5121726751327514, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29540, + "weighted_flow_matching_loss": 0.011326869949698448, + "weighted_rkd_loss": 0.00422749207355082 + }, + { + "epoch": 8.597614198428863, + "flow_matching_loss": 0.011795865092426538, + "grad_norm": 0.094078429043293, + "learning_rate": 1.6609853203979398e-05, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003535429551266134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015894333366304636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008660292392596603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5119109392166138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29550, + "weighted_flow_matching_loss": 0.011795865092426538, + "weighted_rkd_loss": 0.004433292080648243 + }, + { + "epoch": 8.600523712540006, + "flow_matching_loss": 0.013462858367711305, + "grad_norm": 0.106764055788517, + "learning_rate": 1.6601633361777273e-05, + "loss": 0.0182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038291791919618846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016392719116993248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009297630330547691, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5116492092609406, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29560, + "weighted_flow_matching_loss": 0.013462858367711305, + "weighted_rkd_loss": 0.004757091635838151 + }, + { + "epoch": 8.603433226651148, + "flow_matching_loss": 0.013422505743801594, + "grad_norm": 0.12327409535646439, + "learning_rate": 1.6593413033041215e-05, + "loss": 0.0178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035120835760608315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014761562459170818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008500323491171002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5113874793052673, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29570, + "weighted_flow_matching_loss": 0.013422505743801594, + "weighted_rkd_loss": 0.00434700024779886 + }, + { + "epoch": 8.606342740762292, + "flow_matching_loss": 0.01133893383666873, + "grad_norm": 0.11657421290874481, + "learning_rate": 1.6585192220268356e-05, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036394458962604403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015131102758459748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008792002126574517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5111257433891296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29580, + "weighted_flow_matching_loss": 0.01133893383666873, + "weighted_rkd_loss": 0.004493933357298374 + }, + { + "epoch": 8.609252254873436, + "flow_matching_loss": 0.01081292750313878, + "grad_norm": 0.1064760833978653, + "learning_rate": 1.6576970925955956e-05, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035252192756161094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015255228674504906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008575961412861943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.510864007472992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29590, + "weighted_flow_matching_loss": 0.01081292750313878, + "weighted_rkd_loss": 0.00438119200989604 + }, + { + "epoch": 8.61216176898458, + "flow_matching_loss": 0.01016699206084013, + "grad_norm": 0.09533174335956573, + "learning_rate": 1.6568749152601422e-05, + "loss": 0.0146, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003554616984911263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015549796167761087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008664213679730892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5106022536754609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29600, + "weighted_flow_matching_loss": 0.01016699206084013, + "weighted_rkd_loss": 0.004424070101231336 + }, + { + "epoch": 8.615071283095723, + "flow_matching_loss": 0.011618515755981208, + "grad_norm": 0.12323558330535889, + "learning_rate": 1.6560526902702317e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0040352630196139215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015908760717138648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009661402273923158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5103405117988586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29610, + "weighted_flow_matching_loss": 0.011618515755981208, + "weighted_rkd_loss": 0.004930609907023609 + }, + { + "epoch": 8.617980797206867, + "flow_matching_loss": 0.011651655565947295, + "grad_norm": 0.09872311353683472, + "learning_rate": 1.6552304178756337e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004307725303806365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018388764234259724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010454327007755638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.510078763961792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29620, + "weighted_flow_matching_loss": 0.011651655565947295, + "weighted_rkd_loss": 0.005332502210512758 + }, + { + "epoch": 8.62089031131801, + "flow_matching_loss": 0.011735721491277218, + "grad_norm": 0.12051183730363846, + "learning_rate": 1.654408098326134e-05, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034490513615310194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014809693326242267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008379071950912476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5098170101642608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29630, + "weighted_flow_matching_loss": 0.011735721491277218, + "weighted_rkd_loss": 0.00427174586802721 + }, + { + "epoch": 8.623799825429153, + "flow_matching_loss": 0.011877215141430498, + "grad_norm": 0.11376828700304031, + "learning_rate": 1.6535857318715307e-05, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033702436136081814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014271920255851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008167679188773036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5095552563667297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29640, + "weighted_flow_matching_loss": 0.011877215141430498, + "weighted_rkd_loss": 0.004161916370503604 + }, + { + "epoch": 8.626709339540296, + "flow_matching_loss": 0.012724266573786735, + "grad_norm": 0.09206031262874603, + "learning_rate": 1.6527633187616376e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032217752654105426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013185012037865817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0077620516531169415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5092934966087341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29650, + "weighted_flow_matching_loss": 0.012724266573786735, + "weighted_rkd_loss": 0.003953158110380173 + }, + { + "epoch": 8.62961885365144, + "flow_matching_loss": 0.010513555724173784, + "grad_norm": 0.09775429219007492, + "learning_rate": 1.6519408592462818e-05, + "loss": 0.0154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0039289883803576234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016913600848056376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00954933688044548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5090317368507385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29660, + "weighted_flow_matching_loss": 0.010513555724173784, + "weighted_rkd_loss": 0.00486090718768537 + }, + { + "epoch": 8.632528367762584, + "flow_matching_loss": 0.010904714884236454, + "grad_norm": 0.11917541176080704, + "learning_rate": 1.6511183535753044e-05, + "loss": 0.0156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003819016204215586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001571517379488796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009209549846127629, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5087699830532074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29670, + "weighted_flow_matching_loss": 0.010904714884236454, + "weighted_rkd_loss": 0.004685524618253112 + }, + { + "epoch": 8.635437881873727, + "flow_matching_loss": 0.013455824181437492, + "grad_norm": 0.11220242828130722, + "learning_rate": 1.650295801998561e-05, + "loss": 0.0177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003489193366840482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014578569447621704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00843624365516007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5085082113742828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29680, + "weighted_flow_matching_loss": 0.013455824181437492, + "weighted_rkd_loss": 0.004289898136630654 + }, + { + "epoch": 8.638347395984871, + "flow_matching_loss": 0.010751823196187615, + "grad_norm": 0.11042128503322601, + "learning_rate": 1.6494732047659217e-05, + "loss": 0.015, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003452848270535469, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014814264606684447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008387123001739383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5082464396953583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29690, + "weighted_flow_matching_loss": 0.010751823196187615, + "weighted_rkd_loss": 0.00426267774309963 + }, + { + "epoch": 8.641256910096015, + "flow_matching_loss": 0.010646887682378292, + "grad_norm": 0.11045314371585846, + "learning_rate": 1.6486505621272693e-05, + "loss": 0.015, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003562178392894566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014911895617842675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008615546347573399, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5079846739768982, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29700, + "weighted_flow_matching_loss": 0.010646887682378292, + "weighted_rkd_loss": 0.004376573208719492 + }, + { + "epoch": 8.644166424207157, + "flow_matching_loss": 0.014146391209214926, + "grad_norm": 0.12157837301492691, + "learning_rate": 1.647827874332501e-05, + "loss": 0.0185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00353134055621922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014846078120172024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0085472890175879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5077229082584381, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29710, + "weighted_flow_matching_loss": 0.014146391209214926, + "weighted_rkd_loss": 0.0043396950932219625 + }, + { + "epoch": 8.6470759383183, + "flow_matching_loss": 0.011938921362161636, + "grad_norm": 0.11749202013015747, + "learning_rate": 1.6470051416315277e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037488159723579884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001484451990108937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008982083899900318, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5074611306190491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29720, + "weighted_flow_matching_loss": 0.011938921362161636, + "weighted_rkd_loss": 0.004558043647557497 + }, + { + "epoch": 8.649985452429444, + "flow_matching_loss": 0.011897313594818115, + "grad_norm": 0.10786585509777069, + "learning_rate": 1.6461823642742734e-05, + "loss": 0.0162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034506853204220533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015579398721456527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00845931051298976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.50719935297966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29730, + "weighted_flow_matching_loss": 0.011897313594818115, + "weighted_rkd_loss": 0.0042906234506517645 + }, + { + "epoch": 8.652894966540588, + "flow_matching_loss": 0.012016305420547723, + "grad_norm": 0.16140516102313995, + "learning_rate": 1.645359542510677e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004099010163918138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002185750345233828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010383770801126957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.506937575340271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29740, + "weighted_flow_matching_loss": 0.012016305420547723, + "weighted_rkd_loss": 0.005263905110768974 + }, + { + "epoch": 8.655804480651732, + "flow_matching_loss": 0.012577881291508674, + "grad_norm": 0.1372486650943756, + "learning_rate": 1.6445366765906897e-05, + "loss": 0.0168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034048233879730105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015290956012904643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008338742284104228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.506675797700882, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29750, + "weighted_flow_matching_loss": 0.012577881291508674, + "weighted_rkd_loss": 0.004224943369626999 + }, + { + "epoch": 8.658713994762875, + "flow_matching_loss": 0.012324912752956153, + "grad_norm": 0.1187266930937767, + "learning_rate": 1.643713766764276e-05, + "loss": 0.0174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004176740488037467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016937361098825932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010047217085957526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5064140200614929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29760, + "weighted_flow_matching_loss": 0.012324912752956153, + "weighted_rkd_loss": 0.0050880284514278175 + }, + { + "epoch": 8.661623508874017, + "flow_matching_loss": 0.012349226139485836, + "grad_norm": 0.10818785429000854, + "learning_rate": 1.6428908132814146e-05, + "loss": 0.017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037195730488747357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016649376833811403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009104083711281418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5061522364616394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29770, + "weighted_flow_matching_loss": 0.012349226139485836, + "weighted_rkd_loss": 0.004608091549016535 + }, + { + "epoch": 8.664533022985161, + "flow_matching_loss": 0.011835556663572788, + "grad_norm": 0.10541702806949615, + "learning_rate": 1.642067816392097e-05, + "loss": 0.0165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003736380022019148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001742170297075063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009214930329471826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5058904528617859, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29780, + "weighted_flow_matching_loss": 0.011835556663572788, + "weighted_rkd_loss": 0.004661748534999788 + }, + { + "epoch": 8.667442537096305, + "flow_matching_loss": 0.011419681645929814, + "grad_norm": 0.09485401213169098, + "learning_rate": 1.6412447763463275e-05, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003585662809200585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015849445015192031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008756270166486502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5056286573410034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29790, + "weighted_flow_matching_loss": 0.011419681645929814, + "weighted_rkd_loss": 0.004427402000874281 + }, + { + "epoch": 8.670352051207448, + "flow_matching_loss": 0.010677800793200732, + "grad_norm": 0.08891338855028152, + "learning_rate": 1.6404216933941247e-05, + "loss": 0.0152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036904102889820933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015005790104623885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008881399501115083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5053668737411499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29800, + "weighted_flow_matching_loss": 0.010677800793200732, + "weighted_rkd_loss": 0.004488362860865891 + }, + { + "epoch": 8.673261565318592, + "flow_matching_loss": 0.011850041057914495, + "grad_norm": 0.11099628359079361, + "learning_rate": 1.639598567785519e-05, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034298452781513335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001460678392322734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008320368872955442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5051050841808319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29810, + "weighted_flow_matching_loss": 0.011850041057914495, + "weighted_rkd_loss": 0.004202658962458372 + }, + { + "epoch": 8.676171079429736, + "flow_matching_loss": 0.011498821340501309, + "grad_norm": 0.11635763943195343, + "learning_rate": 1.6387753997705547e-05, + "loss": 0.0167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004217871651053429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018864751909859478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010322218621149659, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5048432946205139, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29820, + "weighted_flow_matching_loss": 0.011498821340501309, + "weighted_rkd_loss": 0.005211169365793467 + }, + { + "epoch": 8.67908059354088, + "flow_matching_loss": 0.010983042372390627, + "grad_norm": 0.11154895275831223, + "learning_rate": 1.637952189599287e-05, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003967560548335314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018666468560695649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009801767952740192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5045814990997315, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29830, + "weighted_flow_matching_loss": 0.010983042372390627, + "weighted_rkd_loss": 0.004945706319995224 + }, + { + "epoch": 8.681990107652021, + "flow_matching_loss": 0.011248499900102616, + "grad_norm": 0.10036143660545349, + "learning_rate": 1.6371289375217873e-05, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0038039424922317266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015535605722106992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00916144554503262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5043197095394134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29840, + "weighted_flow_matching_loss": 0.011248499900102616, + "weighted_rkd_loss": 0.004620274435728789 + }, + { + "epoch": 8.684899621763165, + "flow_matching_loss": 0.010234928410500287, + "grad_norm": 0.11890721321105957, + "learning_rate": 1.6363056437881372e-05, + "loss": 0.0148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037595112808048725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001586539892014116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00910556255839765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.504057914018631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29850, + "weighted_flow_matching_loss": 0.010234928410500287, + "weighted_rkd_loss": 0.004589679138734937 + }, + { + "epoch": 8.687809135874309, + "flow_matching_loss": 0.012756581138819455, + "grad_norm": 0.09148385375738144, + "learning_rate": 1.6354823086484315e-05, + "loss": 0.0173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0036581186344847084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016095955797936767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008925832947716117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5037961184978486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29860, + "weighted_flow_matching_loss": 0.012756581138819455, + "weighted_rkd_loss": 0.004496830725111067 + }, + { + "epoch": 8.690718649985453, + "flow_matching_loss": 0.01049762018956244, + "grad_norm": 0.10594145208597183, + "learning_rate": 1.634658932352777e-05, + "loss": 0.0147, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034688357729464768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014302869560196996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008367958525195719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.503534322977066, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29870, + "weighted_flow_matching_loss": 0.01049762018956244, + "weighted_rkd_loss": 0.004213603469543159 + }, + { + "epoch": 8.693628164096596, + "flow_matching_loss": 0.01183232180774212, + "grad_norm": 0.1041228249669075, + "learning_rate": 1.633835515151295e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037850958295166494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001490000879857689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009060192434117199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5032725214958191, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29880, + "weighted_flow_matching_loss": 0.01183232180774212, + "weighted_rkd_loss": 0.004559821821749211 + }, + { + "epoch": 8.69653767820774, + "flow_matching_loss": 0.012176007311791182, + "grad_norm": 0.09452757239341736, + "learning_rate": 1.6330120572941166e-05, + "loss": 0.0164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003464641608297825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013687995611689984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008298082882538438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5030107259750366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29890, + "weighted_flow_matching_loss": 0.012176007311791182, + "weighted_rkd_loss": 0.004174096556380391 + }, + { + "epoch": 8.699447192318882, + "flow_matching_loss": 0.01164389094337821, + "grad_norm": 0.0795823335647583, + "learning_rate": 1.6321885590313876e-05, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033383323578163983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015253177902195604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008201982453465461, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5027489244937897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29900, + "weighted_flow_matching_loss": 0.01164389094337821, + "weighted_rkd_loss": 0.004123588022775948 + }, + { + "epoch": 8.702356706430026, + "flow_matching_loss": 0.011511122807860374, + "grad_norm": 0.10362618416547775, + "learning_rate": 1.6313650206132643e-05, + "loss": 0.0161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003771976660937071, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0016499591758474708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009193912660703064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5024871289730072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29910, + "weighted_flow_matching_loss": 0.011511122807860374, + "weighted_rkd_loss": 0.00461985154543072 + }, + { + "epoch": 8.70526622054117, + "flow_matching_loss": 0.011993542965501546, + "grad_norm": 0.09494026750326157, + "learning_rate": 1.6305414422899163e-05, + "loss": 0.0169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003951838612556458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0018233833718113602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009727060655131936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5022253215312957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29920, + "weighted_flow_matching_loss": 0.011993542965501546, + "weighted_rkd_loss": 0.0048853048589080576 + }, + { + "epoch": 8.708175734652313, + "flow_matching_loss": 0.01219228133559227, + "grad_norm": 0.10743598639965057, + "learning_rate": 1.629717824311525e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004063545959070325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0017310785362496973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009858170431107283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5019635260105133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29930, + "weighted_flow_matching_loss": 0.01219228133559227, + "weighted_rkd_loss": 0.004948424431495369 + }, + { + "epoch": 8.711085248763457, + "flow_matching_loss": 0.011795541644096375, + "grad_norm": 0.10286310315132141, + "learning_rate": 1.6288941669282834e-05, + "loss": 0.016, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0034694421803578733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014434423355851322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008382326597347856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5017017245292663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29940, + "weighted_flow_matching_loss": 0.011795541644096375, + "weighted_rkd_loss": 0.004205545363947749 + }, + { + "epoch": 8.7139947628746, + "flow_matching_loss": 0.011611560452729463, + "grad_norm": 0.11032941192388535, + "learning_rate": 1.628070470390398e-05, + "loss": 0.0157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0033559359842911364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015264582994859665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.00823833029717207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.501439917087555, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29950, + "weighted_flow_matching_loss": 0.011611560452729463, + "weighted_rkd_loss": 0.004131011641584336 + }, + { + "epoch": 8.716904276985744, + "flow_matching_loss": 0.011787842772901059, + "grad_norm": 0.08734644204378128, + "learning_rate": 1.627246734948084e-05, + "loss": 0.0159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003423511632718146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014071904239244758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008254213584586978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.501178115606308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29960, + "weighted_flow_matching_loss": 0.011787842772901059, + "weighted_rkd_loss": 0.004136815969832242 + }, + { + "epoch": 8.719813791096886, + "flow_matching_loss": 0.012573604378849268, + "grad_norm": 0.097365602850914, + "learning_rate": 1.6264229608515725e-05, + "loss": 0.0171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0037407117430120706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015864964516367763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.009067919990047812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.500916314125061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29970, + "weighted_flow_matching_loss": 0.012573604378849268, + "weighted_rkd_loss": 0.00454229605384171 + }, + { + "epoch": 8.72272330520803, + "flow_matching_loss": 0.010747326537966728, + "grad_norm": 0.09672122448682785, + "learning_rate": 1.6255991483511035e-05, + "loss": 0.0151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.003605054272338748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0014500482589937746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008660156885161996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5006545066833497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29980, + "weighted_flow_matching_loss": 0.010747326537966728, + "weighted_rkd_loss": 0.004335703770630061 + }, + { + "epoch": 8.725632819319173, + "flow_matching_loss": 0.011423486471176147, + "grad_norm": 0.10426770150661469, + "learning_rate": 1.6247752976969288e-05, + "loss": 0.0158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0035957696149125696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0015147670870646834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.008706306293606757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5003927052021027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29990, + "weighted_flow_matching_loss": 0.011423486471176147, + "weighted_rkd_loss": 0.00435655266046524 + }, + { + "epoch": 8.728542333430317, + "flow_matching_loss": 0.010536189191043377, + "grad_norm": 0.10687262564897537, + "learning_rate": 1.6239514091393138e-05, + "loss": 0.0144, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0032264549052342774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0013712016050703823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.007824111543595792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5001309037208557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 30000, + "weighted_flow_matching_loss": 0.010536189191043377, + "weighted_rkd_loss": 0.003913067979738116 + } + ], + "logging_steps": 10, + "max_steps": 60000, + "num_input_tokens_seen": 0, + "num_train_epochs": 18, + "save_steps": 30000, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 128, + "trial_name": null, + "trial_params": null +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..064785cc03d1e2223bf2981a1b22674a7e445646 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.0, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/experiment_cfg/metadata.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..300eb070963337016e16465f50ba8351ea2eb25c --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.3139495849609375, + 0.4876587688922882, + 0.7196521759033203 + ], + "min": [ + -4.94293737411499, + -6.890198230743408, + 0.6996827125549316 + ], + "mean": [ + 2.749288365724937, + -2.0455216239909983, + 0.700532551020533 + ], + "std": [ + 1.6786295897046262, + 1.312897969434244, + 0.00133751758906258 + ], + "q01": [ + -1.4241907881148037, + -5.1716547935950885, + 0.6999670898768614 + ], + "q99": [ + 6.198111545831555, + -0.5873531334104489, + 0.7038833014242587 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.23084999587450247, + 0.6238431088581847 + ], + "std": [ + 0.0, + 0.0, + 0.6556404371644047, + 0.3572252105732042 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 1.7482354593273349e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.9014528393745422, + 0.8003877401351929, + 0.9829942584037781 + ], + "min": [ + -0.37471598386764526, + -0.8472502827644348, + -0.25070279836654663 + ], + "mean": [ + 0.28667332601863704, + -0.038315473170876704, + 0.45974658906777444 + ], + "std": [ + 0.17067058828305154, + 0.23569952037784195, + 0.21950702354181487 + ], + "q01": [ + 0.015172948963784228, + -0.44450502063220615, + 0.23314444103329338 + ], + "q99": [ + 0.5609069689572412, + 0.38454179128009547, + 0.7028102736473852 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999998807907104, + 0.9984455108642578, + 0.9434727430343628, + 0.9062229990959167 + ], + "min": [ + -0.9999930262565613, + -0.9988337755203247, + -0.9618149995803833, + 2.1872274658107926e-07 + ], + "mean": [ + -0.25672297401336, + 0.024425335806687216, + -0.08740977164483847, + 0.16608739644018397 + ], + "std": [ + 0.7932585208392383, + 0.3102260195580416, + 0.3772472576315148, + 0.17451959300158773 + ], + "q01": [ + -0.99379006987882, + -0.5478102050038828, + -0.6101777952971636, + 0.002274085014134748 + ], + "q99": [ + 0.8804856038896288, + 0.5910611092873037, + 0.5216058042993562, + 0.5231965974012255 + ] + }, + "gripper_qpos": { + "max": [ + 0.055869169533252716, + 0.010916369967162609 + ], + "min": [ + -0.011436971835792065, + -0.05664053186774254 + ], + "mean": [ + 0.031651551516052555, + -0.03161481387920421 + ], + "std": [ + 0.013119769222217526, + 0.01306078225192402 + ], + "q01": [ + 0.0060999246446777336, + -0.04062601653964198 + ], + "q99": [ + 0.04054891621517283, + -0.0059163567113456345 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.008144896753718373, + -0.00018893135146078263, + -0.0008739062727221845, + 0.0 + ], + "std": [ + 0.11030045424364411, + 0.10148082570313594, + 0.08975698373467506, + 0.0 + ], + "q01": [ + -0.07287373067581518, + -0.09446994199569948, + -0.07702818259249572, + 0.0 + ], + "q99": [ + 0.04485485414407898, + 0.09626941605965177, + 0.07610665906122742, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9221974313425939 + ], + "std": [ + 0.38670194342612674 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.384693946883099 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.01976129923017491, + -0.020870631344490034, + -0.05993476517349181 + ], + "std": [ + 0.4347024990804053, + 0.4221662976569997, + 0.3845506843044066 + ], + "q01": [ + -0.8835792939767807, + -0.9280919251713778, + -0.783361071470956 + ], + "q99": [ + 0.7622084438449307, + 0.8625125252609077, + 0.7470974753250706 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008089673068544335, + -0.026498254627927348, + 0.0029083551743059005 + ], + "std": [ + 0.11216734503733773, + 0.13206002613798629, + 0.12615870412692864 + ], + "q01": [ + -0.2814627972846672, + -0.4342386516115439, + -0.34489755628340574 + ], + "q99": [ + 0.31879353266939386, + 0.28411797683220563, + 0.3597718671669894 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3824228874769045 + ], + "std": [ + 0.9240015096469786 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.5597272305813592 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/config.json new file mode 100644 index 0000000000000000000000000000000000000000..064785cc03d1e2223bf2981a1b22674a7e445646 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.0, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/resolved_config.yaml b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/resolved_config.yaml index cd9b9e931490dbb9df4388f531d29349dbc58231..97f9e7810fa6b2a081dc09f8f42d2b1e29c03b5c 100644 --- a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/resolved_config.yaml +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_Atomic26Baseline60k/default/resolved_config.yaml @@ -9,6 +9,7 @@ training: num_gpus: 1 batch_size: 128 seed: 42 + report_to: wandb model: rkd_teacher_source: timewarp_action rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46 diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k.yaml b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k.yaml new file mode 100644 index 0000000000000000000000000000000000000000..475eff4555be3c3bee29df8686e740ed568c7982 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k.yaml @@ -0,0 +1,40 @@ +experiment_name: RKDv3 TimewarpVAE DA, Student:APHead, single-phase FM+RKD cosine + decay from GR00T base to 120k +policy_type: groot_RKDv3 +base_model_path: nvidia/GR00T-N1.5-3B +output_root: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k +dataset: + dataset_soup: my_atomic26_human +training: + num_gpus: 1 + batch_size: 128 + seed: 42 + report_to: wandb +model: + rkd_teacher_source: timewarp_action + rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46 + rkd_relation_mode: token_pair_mean + rkd_loss_type: distance_angle + rkd_student_reconstruction_head_pooling: attention + rkd_student_reconstruction_head_enabled: true + rkd_student_reconstruction_head_dim: 512 + rkd_student_reconstruction_head_hidden_dim: 512 + rkd_distance_loss_weight: 1.0 + rkd_angle_loss_weight: 2.0 + rkd_exclude_diagonal: true +phases: +- name: fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k + max_steps: 120000 + save_steps: 30000 + trainable: + tune_llm: false + tune_visual: false + tune_projector: true + tune_diffusion_model: true + losses: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 1.0 + rkd_loss_weight_schedule: cosine_decay + rkd_loss_weight_start: 1.0 + rkd_loss_weight_end: 0.0 diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..064785cc03d1e2223bf2981a1b22674a7e445646 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.0, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/experiment_cfg/metadata.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..300eb070963337016e16465f50ba8351ea2eb25c --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.3139495849609375, + 0.4876587688922882, + 0.7196521759033203 + ], + "min": [ + -4.94293737411499, + -6.890198230743408, + 0.6996827125549316 + ], + "mean": [ + 2.749288365724937, + -2.0455216239909983, + 0.700532551020533 + ], + "std": [ + 1.6786295897046262, + 1.312897969434244, + 0.00133751758906258 + ], + "q01": [ + -1.4241907881148037, + -5.1716547935950885, + 0.6999670898768614 + ], + "q99": [ + 6.198111545831555, + -0.5873531334104489, + 0.7038833014242587 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.23084999587450247, + 0.6238431088581847 + ], + "std": [ + 0.0, + 0.0, + 0.6556404371644047, + 0.3572252105732042 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 1.7482354593273349e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.9014528393745422, + 0.8003877401351929, + 0.9829942584037781 + ], + "min": [ + -0.37471598386764526, + -0.8472502827644348, + -0.25070279836654663 + ], + "mean": [ + 0.28667332601863704, + -0.038315473170876704, + 0.45974658906777444 + ], + "std": [ + 0.17067058828305154, + 0.23569952037784195, + 0.21950702354181487 + ], + "q01": [ + 0.015172948963784228, + -0.44450502063220615, + 0.23314444103329338 + ], + "q99": [ + 0.5609069689572412, + 0.38454179128009547, + 0.7028102736473852 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999998807907104, + 0.9984455108642578, + 0.9434727430343628, + 0.9062229990959167 + ], + "min": [ + -0.9999930262565613, + -0.9988337755203247, + -0.9618149995803833, + 2.1872274658107926e-07 + ], + "mean": [ + -0.25672297401336, + 0.024425335806687216, + -0.08740977164483847, + 0.16608739644018397 + ], + "std": [ + 0.7932585208392383, + 0.3102260195580416, + 0.3772472576315148, + 0.17451959300158773 + ], + "q01": [ + -0.99379006987882, + -0.5478102050038828, + -0.6101777952971636, + 0.002274085014134748 + ], + "q99": [ + 0.8804856038896288, + 0.5910611092873037, + 0.5216058042993562, + 0.5231965974012255 + ] + }, + "gripper_qpos": { + "max": [ + 0.055869169533252716, + 0.010916369967162609 + ], + "min": [ + -0.011436971835792065, + -0.05664053186774254 + ], + "mean": [ + 0.031651551516052555, + -0.03161481387920421 + ], + "std": [ + 0.013119769222217526, + 0.01306078225192402 + ], + "q01": [ + 0.0060999246446777336, + -0.04062601653964198 + ], + "q99": [ + 0.04054891621517283, + -0.0059163567113456345 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.008144896753718373, + -0.00018893135146078263, + -0.0008739062727221845, + 0.0 + ], + "std": [ + 0.11030045424364411, + 0.10148082570313594, + 0.08975698373467506, + 0.0 + ], + "q01": [ + -0.07287373067581518, + -0.09446994199569948, + -0.07702818259249572, + 0.0 + ], + "q99": [ + 0.04485485414407898, + 0.09626941605965177, + 0.07610665906122742, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9221974313425939 + ], + "std": [ + 0.38670194342612674 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.384693946883099 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.01976129923017491, + -0.020870631344490034, + -0.05993476517349181 + ], + "std": [ + 0.4347024990804053, + 0.4221662976569997, + 0.3845506843044066 + ], + "q01": [ + -0.8835792939767807, + -0.9280919251713778, + -0.783361071470956 + ], + "q99": [ + 0.7622084438449307, + 0.8625125252609077, + 0.7470974753250706 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008089673068544335, + -0.026498254627927348, + 0.0029083551743059005 + ], + "std": [ + 0.11216734503733773, + 0.13206002613798629, + 0.12615870412692864 + ], + "q01": [ + -0.2814627972846672, + -0.4342386516115439, + -0.34489755628340574 + ], + "q99": [ + 0.31879353266939386, + 0.28411797683220563, + 0.3597718671669894 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3824228874769045 + ], + "std": [ + 0.9240015096469786 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.5597272305813592 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-120000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..b714fae314b5f241bdf8ee8aff1130f71c2be39b --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.8535603325875607, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/experiment_cfg/metadata.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..300eb070963337016e16465f50ba8351ea2eb25c --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.3139495849609375, + 0.4876587688922882, + 0.7196521759033203 + ], + "min": [ + -4.94293737411499, + -6.890198230743408, + 0.6996827125549316 + ], + "mean": [ + 2.749288365724937, + -2.0455216239909983, + 0.700532551020533 + ], + "std": [ + 1.6786295897046262, + 1.312897969434244, + 0.00133751758906258 + ], + "q01": [ + -1.4241907881148037, + -5.1716547935950885, + 0.6999670898768614 + ], + "q99": [ + 6.198111545831555, + -0.5873531334104489, + 0.7038833014242587 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.23084999587450247, + 0.6238431088581847 + ], + "std": [ + 0.0, + 0.0, + 0.6556404371644047, + 0.3572252105732042 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 1.7482354593273349e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.9014528393745422, + 0.8003877401351929, + 0.9829942584037781 + ], + "min": [ + -0.37471598386764526, + -0.8472502827644348, + -0.25070279836654663 + ], + "mean": [ + 0.28667332601863704, + -0.038315473170876704, + 0.45974658906777444 + ], + "std": [ + 0.17067058828305154, + 0.23569952037784195, + 0.21950702354181487 + ], + "q01": [ + 0.015172948963784228, + -0.44450502063220615, + 0.23314444103329338 + ], + "q99": [ + 0.5609069689572412, + 0.38454179128009547, + 0.7028102736473852 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999998807907104, + 0.9984455108642578, + 0.9434727430343628, + 0.9062229990959167 + ], + "min": [ + -0.9999930262565613, + -0.9988337755203247, + -0.9618149995803833, + 2.1872274658107926e-07 + ], + "mean": [ + -0.25672297401336, + 0.024425335806687216, + -0.08740977164483847, + 0.16608739644018397 + ], + "std": [ + 0.7932585208392383, + 0.3102260195580416, + 0.3772472576315148, + 0.17451959300158773 + ], + "q01": [ + -0.99379006987882, + -0.5478102050038828, + -0.6101777952971636, + 0.002274085014134748 + ], + "q99": [ + 0.8804856038896288, + 0.5910611092873037, + 0.5216058042993562, + 0.5231965974012255 + ] + }, + "gripper_qpos": { + "max": [ + 0.055869169533252716, + 0.010916369967162609 + ], + "min": [ + -0.011436971835792065, + -0.05664053186774254 + ], + "mean": [ + 0.031651551516052555, + -0.03161481387920421 + ], + "std": [ + 0.013119769222217526, + 0.01306078225192402 + ], + "q01": [ + 0.0060999246446777336, + -0.04062601653964198 + ], + "q99": [ + 0.04054891621517283, + -0.0059163567113456345 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.008144896753718373, + -0.00018893135146078263, + -0.0008739062727221845, + 0.0 + ], + "std": [ + 0.11030045424364411, + 0.10148082570313594, + 0.08975698373467506, + 0.0 + ], + "q01": [ + -0.07287373067581518, + -0.09446994199569948, + -0.07702818259249572, + 0.0 + ], + "q99": [ + 0.04485485414407898, + 0.09626941605965177, + 0.07610665906122742, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9221974313425939 + ], + "std": [ + 0.38670194342612674 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.384693946883099 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.01976129923017491, + -0.020870631344490034, + -0.05993476517349181 + ], + "std": [ + 0.4347024990804053, + 0.4221662976569997, + 0.3845506843044066 + ], + "q01": [ + -0.8835792939767807, + -0.9280919251713778, + -0.783361071470956 + ], + "q99": [ + 0.7622084438449307, + 0.8625125252609077, + 0.7470974753250706 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008089673068544335, + -0.026498254627927348, + 0.0029083551743059005 + ], + "std": [ + 0.11216734503733773, + 0.13206002613798629, + 0.12615870412692864 + ], + "q01": [ + -0.2814627972846672, + -0.4342386516115439, + -0.34489755628340574 + ], + "q99": [ + 0.31879353266939386, + 0.28411797683220563, + 0.3597718671669894 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3824228874769045 + ], + "std": [ + 0.9240015096469786 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.5597272305813592 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/trainer_state.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..95f9487d1dea69f0aed6a5de34d973510b423e5e --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-30000/trainer_state.json @@ -0,0 +1,138034 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 8.728542333430317, + "eval_steps": 500, + "global_step": 30000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.002909514111143439, + "flow_matching_loss": 0.5737217128276825, + "grad_norm": 6.403800964355469, + "learning_rate": 4.5e-08, + "loss": 0.8651, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10369824916124344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0839690662920475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2913655638694763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 1.0, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10, + "weighted_flow_matching_loss": 0.5737217128276825, + "weighted_rkd_loss": 0.2913655638694763 + }, + { + "epoch": 0.005819028222286878, + "flow_matching_loss": 0.5561041116714478, + "grad_norm": 5.974508762359619, + "learning_rate": 9.5e-08, + "loss": 0.8582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10607568621635437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08991435542702675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.30206572711467744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999999642372132, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20, + "weighted_flow_matching_loss": 0.5561041116714478, + "weighted_rkd_loss": 0.302065709233284 + }, + { + "epoch": 0.008728542333430317, + "flow_matching_loss": 0.5545377016067505, + "grad_norm": 5.052680015563965, + "learning_rate": 1.4500000000000001e-07, + "loss": 0.8542, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10525789856910706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08917311057448388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29968890845775603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999998986721039, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 30, + "weighted_flow_matching_loss": 0.5545377016067505, + "weighted_rkd_loss": 0.29968887865543364 + }, + { + "epoch": 0.011638056444573755, + "flow_matching_loss": 0.5808596134185791, + "grad_norm": 5.544183254241943, + "learning_rate": 1.9499999999999999e-07, + "loss": 0.8806, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10543257817625999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08885328844189644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29971844255924224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999997913837433, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 40, + "weighted_flow_matching_loss": 0.5808596134185791, + "weighted_rkd_loss": 0.299718376994133 + }, + { + "epoch": 0.014547570555717195, + "flow_matching_loss": 0.5410171538591385, + "grad_norm": 6.105470180511475, + "learning_rate": 2.45e-07, + "loss": 0.8393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10493538156151772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08837539330124855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2982461541891098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999996542930603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 50, + "weighted_flow_matching_loss": 0.5410171538591385, + "weighted_rkd_loss": 0.2982460469007492 + }, + { + "epoch": 0.017457084666860634, + "flow_matching_loss": 0.5402100414037705, + "grad_norm": 5.401076316833496, + "learning_rate": 2.95e-07, + "loss": 0.8305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10308770015835762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08410486578941345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29028027057647704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999994933605194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 60, + "weighted_flow_matching_loss": 0.5402100414037705, + "weighted_rkd_loss": 0.2902801215648651 + }, + { + "epoch": 0.020366598778004074, + "flow_matching_loss": 0.5209941565990448, + "grad_norm": 5.495227813720703, + "learning_rate": 3.45e-07, + "loss": 0.8157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10446898788213729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08577501624822617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29471299052238464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999992847442627, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 70, + "weighted_flow_matching_loss": 0.5209941565990448, + "weighted_rkd_loss": 0.2947127789258957 + }, + { + "epoch": 0.02327611288914751, + "flow_matching_loss": 0.5188608229160309, + "grad_norm": 5.4039082527160645, + "learning_rate": 3.9500000000000003e-07, + "loss": 0.809, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10313423350453377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08389574736356735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2901642143726349, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999999052286148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 80, + "weighted_flow_matching_loss": 0.5188608229160309, + "weighted_rkd_loss": 0.29016393423080444 + }, + { + "epoch": 0.02618562700029095, + "flow_matching_loss": 0.4622833788394928, + "grad_norm": 4.486055850982666, + "learning_rate": 4.4500000000000003e-07, + "loss": 0.7543, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10323422849178314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08558745831251144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29205591678619386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999987781047821, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 90, + "weighted_flow_matching_loss": 0.4622833788394928, + "weighted_rkd_loss": 0.29205556213855743 + }, + { + "epoch": 0.02909514111143439, + "flow_matching_loss": 0.4313195139169693, + "grad_norm": 4.220290660858154, + "learning_rate": 4.95e-07, + "loss": 0.7265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10463635548949242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08586051240563393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2951332300901413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999984741210938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 100, + "weighted_flow_matching_loss": 0.4313195139169693, + "weighted_rkd_loss": 0.295132777094841 + }, + { + "epoch": 0.03200465522257783, + "flow_matching_loss": 0.40186394155025484, + "grad_norm": 3.971691608428955, + "learning_rate": 5.450000000000001e-07, + "loss": 0.6925, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10277294293045998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08511824384331704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29066413044929507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999998128414154, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 110, + "weighted_flow_matching_loss": 0.40186394155025484, + "weighted_rkd_loss": 0.29066358506679535 + }, + { + "epoch": 0.03491416933372127, + "flow_matching_loss": 0.3133441314101219, + "grad_norm": 1.8667128086090088, + "learning_rate": 5.95e-07, + "loss": 0.5998, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10259379521012306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08128911629319191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.28647670745849607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999997752904892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 120, + "weighted_flow_matching_loss": 0.3133441314101219, + "weighted_rkd_loss": 0.2864760607481003 + }, + { + "epoch": 0.037823683444864704, + "flow_matching_loss": 0.2448175296187401, + "grad_norm": 1.644705891609192, + "learning_rate": 6.45e-07, + "loss": 0.5358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10321366116404533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08456038162112237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2909876972436905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999973475933075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 130, + "weighted_flow_matching_loss": 0.2448175296187401, + "weighted_rkd_loss": 0.2909869283437729 + }, + { + "epoch": 0.04073319755600815, + "flow_matching_loss": 0.24459341168403625, + "grad_norm": 1.2942585945129395, + "learning_rate": 6.95e-07, + "loss": 0.5392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1041433036327362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08627613335847854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.29456273913383485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999968945980072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 140, + "weighted_flow_matching_loss": 0.24459341168403625, + "weighted_rkd_loss": 0.2945618212223053 + }, + { + "epoch": 0.043642711667151585, + "flow_matching_loss": 0.21617774963378905, + "grad_norm": 0.887003481388092, + "learning_rate": 7.45e-07, + "loss": 0.4942, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09907905384898186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07984766066074371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2780057698488235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999964237213135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 150, + "weighted_flow_matching_loss": 0.21617774963378905, + "weighted_rkd_loss": 0.2780047744512558 + }, + { + "epoch": 0.04655222577829502, + "flow_matching_loss": 0.188312891125679, + "grad_norm": 0.6389485001564026, + "learning_rate": 7.95e-07, + "loss": 0.4735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10164806023240089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08190092742443085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.28519704937934875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999995905160904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 160, + "weighted_flow_matching_loss": 0.188312891125679, + "weighted_rkd_loss": 0.2851958841085434 + }, + { + "epoch": 0.049461739889438465, + "flow_matching_loss": 0.18851989209651948, + "grad_norm": 0.6190875768661499, + "learning_rate": 8.45e-07, + "loss": 0.4639, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09865621253848075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07810960114002227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2754220277070999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999995368719101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 170, + "weighted_flow_matching_loss": 0.18851989209651948, + "weighted_rkd_loss": 0.2754207491874695 + }, + { + "epoch": 0.0523712540005819, + "flow_matching_loss": 0.174469231069088, + "grad_norm": 0.5561557412147522, + "learning_rate": 8.95e-07, + "loss": 0.4459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09751037582755089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07642504721879959, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.27144579887390136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999947845935822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 180, + "weighted_flow_matching_loss": 0.174469231069088, + "weighted_rkd_loss": 0.2714443773031235 + }, + { + "epoch": 0.055280768111725345, + "flow_matching_loss": 0.16975222080945968, + "grad_norm": 0.5568479895591736, + "learning_rate": 9.450000000000001e-07, + "loss": 0.4389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09722834378480912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07469995468854904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2691566407680511, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999941706657409, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 190, + "weighted_flow_matching_loss": 0.16975222080945968, + "weighted_rkd_loss": 0.26915507167577746 + }, + { + "epoch": 0.05819028222286878, + "flow_matching_loss": 0.14826540797948837, + "grad_norm": 0.5543729662895203, + "learning_rate": 9.95e-07, + "loss": 0.416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09636104553937912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07504480481147766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2677668958902359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999935150146484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 200, + "weighted_flow_matching_loss": 0.14826540797948837, + "weighted_rkd_loss": 0.26776515543460844 + }, + { + "epoch": 0.06109979633401222, + "flow_matching_loss": 0.1506043404340744, + "grad_norm": 0.5701209902763367, + "learning_rate": 1.045e-06, + "loss": 0.4166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09624321386218071, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07354565635323525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2660320907831192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999992835521698, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 210, + "weighted_flow_matching_loss": 0.1506043404340744, + "weighted_rkd_loss": 0.26603018045425414 + }, + { + "epoch": 0.06400931044515566, + "flow_matching_loss": 0.15071430802345276, + "grad_norm": 0.5476171374320984, + "learning_rate": 1.0949999999999999e-06, + "loss": 0.41, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0946278914809227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07002539560198784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2592811778187752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999921143054962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 220, + "weighted_flow_matching_loss": 0.15071430802345276, + "weighted_rkd_loss": 0.2592791348695755 + }, + { + "epoch": 0.06691882455629909, + "flow_matching_loss": 0.14814706295728683, + "grad_norm": 0.5319710373878479, + "learning_rate": 1.145e-06, + "loss": 0.4009, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0926702156662941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06743392422795295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2527743548154831, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999991363286972, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 230, + "weighted_flow_matching_loss": 0.14814706295728683, + "weighted_rkd_loss": 0.2527721717953682 + }, + { + "epoch": 0.06982833866744254, + "flow_matching_loss": 0.14610999152064325, + "grad_norm": 0.5947344303131104, + "learning_rate": 1.195e-06, + "loss": 0.3966, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09194172844290734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06661853082478046, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.25050198584794997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999905824661255, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 240, + "weighted_flow_matching_loss": 0.14610999152064325, + "weighted_rkd_loss": 0.250499626994133 + }, + { + "epoch": 0.07273785277858598, + "flow_matching_loss": 0.14442772120237352, + "grad_norm": 0.4710013270378113, + "learning_rate": 1.2450000000000002e-06, + "loss": 0.3933, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09156592562794685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06578052714467049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24891237914562225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999897539615631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 250, + "weighted_flow_matching_loss": 0.14442772120237352, + "weighted_rkd_loss": 0.24890982806682588 + }, + { + "epoch": 0.07564736688972941, + "flow_matching_loss": 0.14084165468811988, + "grad_norm": 0.4943302273750305, + "learning_rate": 1.295e-06, + "loss": 0.3953, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09304517060518265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06840209811925888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.25449243783950803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999889016151429, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 260, + "weighted_flow_matching_loss": 0.14084165468811988, + "weighted_rkd_loss": 0.2544896125793457 + }, + { + "epoch": 0.07855688100087285, + "flow_matching_loss": 0.14127179533243178, + "grad_norm": 0.5258878469467163, + "learning_rate": 1.345e-06, + "loss": 0.3931, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09232835173606872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06712558344006539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.251782289147377, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999880194664001, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 270, + "weighted_flow_matching_loss": 0.14127179533243178, + "weighted_rkd_loss": 0.2517792761325836 + }, + { + "epoch": 0.0814663951120163, + "flow_matching_loss": 0.13854421377182008, + "grad_norm": 0.5124452114105225, + "learning_rate": 1.395e-06, + "loss": 0.3868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09082813411951066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06664861105382443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24830487966537476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999870836734772, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 280, + "weighted_flow_matching_loss": 0.13854421377182008, + "weighted_rkd_loss": 0.24830167293548583 + }, + { + "epoch": 0.08437590922315973, + "flow_matching_loss": 0.13457257449626922, + "grad_norm": 0.5596188902854919, + "learning_rate": 1.4450000000000001e-06, + "loss": 0.3781, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08952665403485298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0644586943089962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24351200163364412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999861299991608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 290, + "weighted_flow_matching_loss": 0.13457257449626922, + "weighted_rkd_loss": 0.2435086265206337 + }, + { + "epoch": 0.08728542333430317, + "flow_matching_loss": 0.12803267762064935, + "grad_norm": 0.5214979648590088, + "learning_rate": 1.495e-06, + "loss": 0.3706, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08905257806181907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0644542571157217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24255941212177276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999851405620575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 300, + "weighted_flow_matching_loss": 0.12803267762064935, + "weighted_rkd_loss": 0.2425558090209961 + }, + { + "epoch": 0.09019493744544661, + "flow_matching_loss": 0.12196605280041695, + "grad_norm": 0.46032965183258057, + "learning_rate": 1.545e-06, + "loss": 0.3701, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09081107079982757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06655174307525158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24817388653755187, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999841153621674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 310, + "weighted_flow_matching_loss": 0.12196605280041695, + "weighted_rkd_loss": 0.2481699451804161 + }, + { + "epoch": 0.09310445155659004, + "flow_matching_loss": 0.12123346850275993, + "grad_norm": 0.433512419462204, + "learning_rate": 1.595e-06, + "loss": 0.364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0889467641711235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06492478549480438, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24281831085681915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999830484390259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 320, + "weighted_flow_matching_loss": 0.12123346850275993, + "weighted_rkd_loss": 0.24281420111656188 + }, + { + "epoch": 0.09601396566773349, + "flow_matching_loss": 0.1246294029057026, + "grad_norm": 0.6288943886756897, + "learning_rate": 1.645e-06, + "loss": 0.3666, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08866680487990379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06468086130917072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24201447367668152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999819576740265, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 330, + "weighted_flow_matching_loss": 0.1246294029057026, + "weighted_rkd_loss": 0.24201010763645173 + }, + { + "epoch": 0.09892347977887693, + "flow_matching_loss": 0.12302199602127076, + "grad_norm": 0.4926513433456421, + "learning_rate": 1.695e-06, + "loss": 0.365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08875568434596062, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06443985216319562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24195122271776198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999808251857758, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 340, + "weighted_flow_matching_loss": 0.12302199602127076, + "weighted_rkd_loss": 0.24194658398628235 + }, + { + "epoch": 0.10183299389002037, + "flow_matching_loss": 0.12470095157623291, + "grad_norm": 0.5615043044090271, + "learning_rate": 1.745e-06, + "loss": 0.3637, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0884881243109703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06204789020121097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23902413845062256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999796748161316, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 350, + "weighted_flow_matching_loss": 0.12470095157623291, + "weighted_rkd_loss": 0.23901928067207337 + }, + { + "epoch": 0.1047425080011638, + "flow_matching_loss": 0.11879483461380005, + "grad_norm": 0.5437239408493042, + "learning_rate": 1.7950000000000002e-06, + "loss": 0.3581, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08817570060491561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06295398995280266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23930539041757584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999784767627716, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 360, + "weighted_flow_matching_loss": 0.11879483461380005, + "weighted_rkd_loss": 0.23930024206638337 + }, + { + "epoch": 0.10765202211230725, + "flow_matching_loss": 0.11552066579461098, + "grad_norm": 0.4707695245742798, + "learning_rate": 1.8450000000000001e-06, + "loss": 0.347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08549066111445427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06050843633711338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23148976117372513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999772310256958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 370, + "weighted_flow_matching_loss": 0.11552066579461098, + "weighted_rkd_loss": 0.23148448914289474 + }, + { + "epoch": 0.11056153622345069, + "flow_matching_loss": 0.11901139318943024, + "grad_norm": 0.4747121334075928, + "learning_rate": 1.8949999999999999e-06, + "loss": 0.3527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08657917901873588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.060504211112856865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23366256803274155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999975961446762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 380, + "weighted_flow_matching_loss": 0.11901139318943024, + "weighted_rkd_loss": 0.23365694880485535 + }, + { + "epoch": 0.11347105033459412, + "flow_matching_loss": 0.11856654509902001, + "grad_norm": 0.5021275877952576, + "learning_rate": 1.9450000000000004e-06, + "loss": 0.345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08406960964202881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.058284978941082956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22642419636249542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999746680259705, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 390, + "weighted_flow_matching_loss": 0.11856654509902001, + "weighted_rkd_loss": 0.22641846239566804 + }, + { + "epoch": 0.11638056444573756, + "flow_matching_loss": 0.11770113483071327, + "grad_norm": 0.546765923500061, + "learning_rate": 1.995e-06, + "loss": 0.3455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08470815643668175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05836289674043656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22777920663356782, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999733328819275, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 400, + "weighted_flow_matching_loss": 0.11770113483071327, + "weighted_rkd_loss": 0.22777313590049744 + }, + { + "epoch": 0.11929007855688101, + "flow_matching_loss": 0.11463892310857773, + "grad_norm": 0.6406611204147339, + "learning_rate": 2.045e-06, + "loss": 0.3368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0827507160604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05671004019677639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22221147418022155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999719560146332, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 410, + "weighted_flow_matching_loss": 0.11463892310857773, + "weighted_rkd_loss": 0.2222052425146103 + }, + { + "epoch": 0.12219959266802444, + "flow_matching_loss": 0.11378192603588104, + "grad_norm": 0.5256585478782654, + "learning_rate": 2.095e-06, + "loss": 0.3432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08528148457407951, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05885979682207108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2294227659702301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999705612659454, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 420, + "weighted_flow_matching_loss": 0.11378192603588104, + "weighted_rkd_loss": 0.22941601127386094 + }, + { + "epoch": 0.12510910677916787, + "flow_matching_loss": 0.11215269416570664, + "grad_norm": 0.4803306460380554, + "learning_rate": 2.145e-06, + "loss": 0.3395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08424516394734383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05883433036506176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22732465714216232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999691188335419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 430, + "weighted_flow_matching_loss": 0.11215269416570664, + "weighted_rkd_loss": 0.2273176357150078 + }, + { + "epoch": 0.12801862089031132, + "flow_matching_loss": 0.10847183912992478, + "grad_norm": 0.474737286567688, + "learning_rate": 2.195e-06, + "loss": 0.3446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08726069331169128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06162208653986454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23614347279071807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999676525592804, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 440, + "weighted_flow_matching_loss": 0.10847183912992478, + "weighted_rkd_loss": 0.2361358344554901 + }, + { + "epoch": 0.13092813500145475, + "flow_matching_loss": 0.1059174083173275, + "grad_norm": 0.46820786595344543, + "learning_rate": 2.245e-06, + "loss": 0.3353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08485040888190269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05969025082886219, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22939106971025466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999661386013031, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 450, + "weighted_flow_matching_loss": 0.1059174083173275, + "weighted_rkd_loss": 0.22938330322504044 + }, + { + "epoch": 0.13383764911259818, + "flow_matching_loss": 0.10910382121801376, + "grad_norm": 0.5320130586624146, + "learning_rate": 2.295e-06, + "loss": 0.3337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08320427015423774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05818249471485615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2245910346508026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999646008014679, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 460, + "weighted_flow_matching_loss": 0.10910382121801376, + "weighted_rkd_loss": 0.22458308190107346 + }, + { + "epoch": 0.13674716322374164, + "flow_matching_loss": 0.10526451990008354, + "grad_norm": 0.5238983035087585, + "learning_rate": 2.345e-06, + "loss": 0.3251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0817720003426075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05628750436007977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21983150541782379, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999630272388458, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 470, + "weighted_flow_matching_loss": 0.10526451990008354, + "weighted_rkd_loss": 0.21982337683439254 + }, + { + "epoch": 0.13965667733488507, + "flow_matching_loss": 0.10387463271617889, + "grad_norm": 0.44965898990631104, + "learning_rate": 2.3950000000000004e-06, + "loss": 0.3303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08382733911275864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.058825019374489786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22647969722747802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999614179134368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 480, + "weighted_flow_matching_loss": 0.10387463271617889, + "weighted_rkd_loss": 0.22647095769643782 + }, + { + "epoch": 0.1425661914460285, + "flow_matching_loss": 0.09914228543639184, + "grad_norm": 0.6081563234329224, + "learning_rate": 2.4450000000000003e-06, + "loss": 0.3189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08177263662219048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.056262578815221786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21980785131454467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999597787857055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 490, + "weighted_flow_matching_loss": 0.09914228543639184, + "weighted_rkd_loss": 0.2197990134358406 + }, + { + "epoch": 0.14547570555717196, + "flow_matching_loss": 0.10508745610713958, + "grad_norm": 0.5450594425201416, + "learning_rate": 2.4950000000000003e-06, + "loss": 0.3208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08080345094203949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05414633564651013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21575323790311812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999580979347229, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 500, + "weighted_flow_matching_loss": 0.10508745610713958, + "weighted_rkd_loss": 0.21574419289827346 + }, + { + "epoch": 0.1483852196683154, + "flow_matching_loss": 0.09971612617373467, + "grad_norm": 0.489124059677124, + "learning_rate": 2.5449999999999998e-06, + "loss": 0.3206, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08209460899233818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05666870437562466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22085792273283006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999563872814179, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 510, + "weighted_flow_matching_loss": 0.09971612617373467, + "weighted_rkd_loss": 0.2208482876420021 + }, + { + "epoch": 0.15129473377945882, + "flow_matching_loss": 0.09784649387001991, + "grad_norm": 0.561657190322876, + "learning_rate": 2.5949999999999997e-06, + "loss": 0.32, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08264525160193444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05683711059391498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22212761640548706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999546468257904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 520, + "weighted_flow_matching_loss": 0.09784649387001991, + "weighted_rkd_loss": 0.22211754620075225 + }, + { + "epoch": 0.15420424789060228, + "flow_matching_loss": 0.09803689643740654, + "grad_norm": 0.44686922430992126, + "learning_rate": 2.645e-06, + "loss": 0.3179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08208770528435708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.055648038536310195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2198234513401985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999952882528305, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 530, + "weighted_flow_matching_loss": 0.09803689643740654, + "weighted_rkd_loss": 0.21981309354305267 + }, + { + "epoch": 0.1571137620017457, + "flow_matching_loss": 0.09569604173302651, + "grad_norm": 0.5454907417297363, + "learning_rate": 2.695e-06, + "loss": 0.314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08152997344732285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05526325516402721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21832320094108582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999951034784317, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 540, + "weighted_flow_matching_loss": 0.09569604173302651, + "weighted_rkd_loss": 0.21831251084804534 + }, + { + "epoch": 0.16002327611288913, + "flow_matching_loss": 0.10124593079090119, + "grad_norm": 0.49220237135887146, + "learning_rate": 2.745e-06, + "loss": 0.3156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08026279956102371, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05381502136588097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21434062272310256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999492049217225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 550, + "weighted_flow_matching_loss": 0.10124593079090119, + "weighted_rkd_loss": 0.21432973444461823 + }, + { + "epoch": 0.1629327902240326, + "flow_matching_loss": 0.09673213213682175, + "grad_norm": 0.5224379301071167, + "learning_rate": 2.795e-06, + "loss": 0.3114, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08017460778355598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05433070324361324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21467991918325424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999947315454483, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 560, + "weighted_flow_matching_loss": 0.09673213213682175, + "weighted_rkd_loss": 0.2146686092019081 + }, + { + "epoch": 0.16584230433517602, + "flow_matching_loss": 0.0997069962322712, + "grad_norm": 0.5108600854873657, + "learning_rate": 2.8450000000000003e-06, + "loss": 0.3134, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08020409494638443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.053314324095845224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21372251361608505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999453961849213, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 570, + "weighted_flow_matching_loss": 0.0997069962322712, + "weighted_rkd_loss": 0.21371084302663804 + }, + { + "epoch": 0.16875181844631945, + "flow_matching_loss": 0.09654248654842376, + "grad_norm": 0.5994459390640259, + "learning_rate": 2.8950000000000002e-06, + "loss": 0.3084, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07922307029366493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05344288423657417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21188902407884597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999434471130371, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 580, + "weighted_flow_matching_loss": 0.09654248654842376, + "weighted_rkd_loss": 0.21187703758478166 + }, + { + "epoch": 0.1716613325574629, + "flow_matching_loss": 0.09500396698713302, + "grad_norm": 0.5412867069244385, + "learning_rate": 2.945e-06, + "loss": 0.3075, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07947369068861007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05353306792676449, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21248045116662978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999414622783661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 590, + "weighted_flow_matching_loss": 0.09500396698713302, + "weighted_rkd_loss": 0.2124680131673813 + }, + { + "epoch": 0.17457084666860634, + "flow_matching_loss": 0.09535243436694145, + "grad_norm": 0.5644533634185791, + "learning_rate": 2.995e-06, + "loss": 0.3059, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07930486500263215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0519989725202322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21060870140790938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999394416809082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 600, + "weighted_flow_matching_loss": 0.09535243436694145, + "weighted_rkd_loss": 0.21059594601392745 + }, + { + "epoch": 0.17748036077974977, + "flow_matching_loss": 0.0955127276480198, + "grad_norm": 0.4822477400302887, + "learning_rate": 3.0450000000000005e-06, + "loss": 0.3041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07848285436630249, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05161047391593456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20857618600130082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999373853206635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 610, + "weighted_flow_matching_loss": 0.0955127276480198, + "weighted_rkd_loss": 0.2085631236433983 + }, + { + "epoch": 0.18038987489089323, + "flow_matching_loss": 0.09671695530414581, + "grad_norm": 0.735806941986084, + "learning_rate": 3.095e-06, + "loss": 0.3039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07779621109366416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05157471261918545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2071671336889267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999352991580963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 620, + "weighted_flow_matching_loss": 0.09671695530414581, + "weighted_rkd_loss": 0.2071537271142006 + }, + { + "epoch": 0.18329938900203666, + "flow_matching_loss": 0.0898235209286213, + "grad_norm": 0.6739233136177063, + "learning_rate": 3.145e-06, + "loss": 0.2994, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07875544354319572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.052035973966121675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.209546859562397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999331712722779, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 630, + "weighted_flow_matching_loss": 0.0898235209286213, + "weighted_rkd_loss": 0.20953285545110703 + }, + { + "epoch": 0.18620890311318009, + "flow_matching_loss": 0.0911832682788372, + "grad_norm": 0.6197074055671692, + "learning_rate": 3.195e-06, + "loss": 0.303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0794224426150322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.052941114827990535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21178599894046785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999310135841369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 640, + "weighted_flow_matching_loss": 0.0911832682788372, + "weighted_rkd_loss": 0.21177138686180114 + }, + { + "epoch": 0.18911841722432354, + "flow_matching_loss": 0.09259459599852563, + "grad_norm": 0.4756271541118622, + "learning_rate": 3.245e-06, + "loss": 0.3033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07885557040572166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05299187041819096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21070300936698913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999288260936737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 650, + "weighted_flow_matching_loss": 0.09259459599852563, + "weighted_rkd_loss": 0.21068801432847978 + }, + { + "epoch": 0.19202793133546697, + "flow_matching_loss": 0.08833198398351669, + "grad_norm": 0.6852693557739258, + "learning_rate": 3.2950000000000002e-06, + "loss": 0.2992, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0789560280740261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05292240157723427, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21083445698022843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999265968799591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 660, + "weighted_flow_matching_loss": 0.08833198398351669, + "weighted_rkd_loss": 0.21081898212432862 + }, + { + "epoch": 0.1949374454466104, + "flow_matching_loss": 0.08614316806197167, + "grad_norm": 0.5937833189964294, + "learning_rate": 3.345e-06, + "loss": 0.2973, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0786509096622467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05388017669320107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21118199676275254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999243378639221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 670, + "weighted_flow_matching_loss": 0.08614316806197167, + "weighted_rkd_loss": 0.21116601675748825 + }, + { + "epoch": 0.19784695955775386, + "flow_matching_loss": 0.09015217125415802, + "grad_norm": 0.6100872159004211, + "learning_rate": 3.395e-06, + "loss": 0.2949, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07678105756640434, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05121423341333866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20477634966373442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999220490455627, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 680, + "weighted_flow_matching_loss": 0.09015217125415802, + "weighted_rkd_loss": 0.2047603815793991 + }, + { + "epoch": 0.2007564736688973, + "flow_matching_loss": 0.08791607171297074, + "grad_norm": 0.5763606429100037, + "learning_rate": 3.445e-06, + "loss": 0.2901, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07637171149253845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04944418258965015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20218760371208191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999197125434875, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 690, + "weighted_flow_matching_loss": 0.08791607171297074, + "weighted_rkd_loss": 0.20217137336730956 + }, + { + "epoch": 0.20366598778004075, + "flow_matching_loss": 0.0882195420563221, + "grad_norm": 0.569847583770752, + "learning_rate": 3.4950000000000004e-06, + "loss": 0.2882, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07547706440091133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04899582639336586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19994995296001433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999173283576965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 700, + "weighted_flow_matching_loss": 0.0882195420563221, + "weighted_rkd_loss": 0.199933423101902 + }, + { + "epoch": 0.20657550189118418, + "flow_matching_loss": 0.08962792679667472, + "grad_norm": 0.5277986526489258, + "learning_rate": 3.5450000000000004e-06, + "loss": 0.2932, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07670732215046883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0501865740865469, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2036012202501297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999149441719055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 710, + "weighted_flow_matching_loss": 0.08962792679667472, + "weighted_rkd_loss": 0.20358389765024185 + }, + { + "epoch": 0.2094850160023276, + "flow_matching_loss": 0.08426154851913452, + "grad_norm": 0.5213105082511902, + "learning_rate": 3.595e-06, + "loss": 0.2877, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07673632949590684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05000234618782997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20347500294446946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999125242233277, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 720, + "weighted_flow_matching_loss": 0.08426154851913452, + "weighted_rkd_loss": 0.2034572035074234 + }, + { + "epoch": 0.21239453011347106, + "flow_matching_loss": 0.08454644381999969, + "grad_norm": 0.5175310373306274, + "learning_rate": 3.645e-06, + "loss": 0.2816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07471510246396065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04763279855251312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19706300050020217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999100565910339, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 730, + "weighted_flow_matching_loss": 0.08454644381999969, + "weighted_rkd_loss": 0.1970452770590782 + }, + { + "epoch": 0.2153040442246145, + "flow_matching_loss": 0.08501040861010552, + "grad_norm": 0.4858770966529846, + "learning_rate": 3.6949999999999998e-06, + "loss": 0.2841, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07501617893576622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04906984567642212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19910220205783843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999075591564178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 740, + "weighted_flow_matching_loss": 0.08501040861010552, + "weighted_rkd_loss": 0.1990837946534157 + }, + { + "epoch": 0.21821355833575792, + "flow_matching_loss": 0.08983534276485443, + "grad_norm": 0.45997336506843567, + "learning_rate": 3.745e-06, + "loss": 0.2838, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07364512532949448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.046698933094739915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1939891830086708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999050259590149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 750, + "weighted_flow_matching_loss": 0.08983534276485443, + "weighted_rkd_loss": 0.19397075921297074 + }, + { + "epoch": 0.22112307244690138, + "flow_matching_loss": 0.08681272342801094, + "grad_norm": 0.5401157140731812, + "learning_rate": 3.795e-06, + "loss": 0.2821, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07381276264786721, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04767441898584366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19529994279146196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999024569988251, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 760, + "weighted_flow_matching_loss": 0.08681272342801094, + "weighted_rkd_loss": 0.19528089314699174 + }, + { + "epoch": 0.2240325865580448, + "flow_matching_loss": 0.08497177362442017, + "grad_norm": 0.5683760046958923, + "learning_rate": 3.8450000000000005e-06, + "loss": 0.2764, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07240133732557297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.046616044268012045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19141871929168702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998998522758484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 770, + "weighted_flow_matching_loss": 0.08497177362442017, + "weighted_rkd_loss": 0.191399547457695 + }, + { + "epoch": 0.22694210066918824, + "flow_matching_loss": 0.0803778313100338, + "grad_norm": 0.6218329071998596, + "learning_rate": 3.895e-06, + "loss": 0.2714, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07246446013450622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.046123959869146344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19105288088321687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998972177505493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 780, + "weighted_flow_matching_loss": 0.0803778313100338, + "weighted_rkd_loss": 0.19103324115276338 + }, + { + "epoch": 0.2298516147803317, + "flow_matching_loss": 0.08288460895419121, + "grad_norm": 0.6256872415542603, + "learning_rate": 3.945e-06, + "loss": 0.2734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0724123127758503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.045686669275164606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1905112937092781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998945534229279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 790, + "weighted_flow_matching_loss": 0.08288460895419121, + "weighted_rkd_loss": 0.19049120396375657 + }, + { + "epoch": 0.23276112889147513, + "flow_matching_loss": 0.08272862508893013, + "grad_norm": 0.6145538687705994, + "learning_rate": 3.995e-06, + "loss": 0.2783, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07419610172510147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04721040613949299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19560261070728302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998918414115906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 800, + "weighted_flow_matching_loss": 0.08272862508893013, + "weighted_rkd_loss": 0.19558145701885224 + }, + { + "epoch": 0.23567064300261856, + "flow_matching_loss": 0.08722126483917236, + "grad_norm": 0.5504571199417114, + "learning_rate": 4.045e-06, + "loss": 0.2873, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07526344954967498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04956614971160889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.200093050301075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998891055583954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 810, + "weighted_flow_matching_loss": 0.08722126483917236, + "weighted_rkd_loss": 0.20007086247205735 + }, + { + "epoch": 0.23858015711376201, + "flow_matching_loss": 0.0833832360804081, + "grad_norm": 0.4709477126598358, + "learning_rate": 4.095000000000001e-06, + "loss": 0.2742, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07225504368543625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.046324358135461804, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19083444476127626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998863279819489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 820, + "weighted_flow_matching_loss": 0.0833832360804081, + "weighted_rkd_loss": 0.19081275016069413 + }, + { + "epoch": 0.24148967122490544, + "flow_matching_loss": 0.08110717087984085, + "grad_norm": 0.5248776078224182, + "learning_rate": 4.145e-06, + "loss": 0.2727, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0723581776022911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04688061252236366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19159696847200394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998835146427154, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 830, + "weighted_flow_matching_loss": 0.08110717087984085, + "weighted_rkd_loss": 0.1915746495127678 + }, + { + "epoch": 0.24439918533604887, + "flow_matching_loss": 0.08250362053513527, + "grad_norm": 0.5177388787269592, + "learning_rate": 4.1950000000000005e-06, + "loss": 0.2753, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0728741578757763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04704448841512203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1927928015589714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998806774616241, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 840, + "weighted_flow_matching_loss": 0.08250362053513527, + "weighted_rkd_loss": 0.19276979565620422 + }, + { + "epoch": 0.24730869944719233, + "flow_matching_loss": 0.07850465178489685, + "grad_norm": 0.48424312472343445, + "learning_rate": 4.245e-06, + "loss": 0.2652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07100682258605957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.044750921055674554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18676456660032273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999877804517746, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 850, + "weighted_flow_matching_loss": 0.07850465178489685, + "weighted_rkd_loss": 0.18674174547195435 + }, + { + "epoch": 0.25021821355833573, + "flow_matching_loss": 0.08460175096988679, + "grad_norm": 0.49639734625816345, + "learning_rate": 4.295e-06, + "loss": 0.2726, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07158516570925713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04489554651081562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18806587755680085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998748898506165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 860, + "weighted_flow_matching_loss": 0.08460175096988679, + "weighted_rkd_loss": 0.18804234713315965 + }, + { + "epoch": 0.2531277276694792, + "flow_matching_loss": 0.0794167809188366, + "grad_norm": 0.5560476779937744, + "learning_rate": 4.345000000000001e-06, + "loss": 0.2713, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07267894893884659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04658915102481842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1919470489025116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998719394207001, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 870, + "weighted_flow_matching_loss": 0.0794167809188366, + "weighted_rkd_loss": 0.19192246049642564 + }, + { + "epoch": 0.25603724178062265, + "flow_matching_loss": 0.07824064455926419, + "grad_norm": 0.5967847108840942, + "learning_rate": 4.395e-06, + "loss": 0.2638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07047800049185753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0446726355701685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18562863469123841, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998689591884613, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 880, + "weighted_flow_matching_loss": 0.07824064455926419, + "weighted_rkd_loss": 0.1856043040752411 + }, + { + "epoch": 0.2589467558917661, + "flow_matching_loss": 0.08226599395275117, + "grad_norm": 0.6336823105812073, + "learning_rate": 4.445e-06, + "loss": 0.2655, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06979618743062019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.043694930523633956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1832873061299324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998659551143646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 890, + "weighted_flow_matching_loss": 0.08226599395275117, + "weighted_rkd_loss": 0.1832627385854721 + }, + { + "epoch": 0.2618562700029095, + "flow_matching_loss": 0.07749143987894058, + "grad_norm": 0.599888026714325, + "learning_rate": 4.495e-06, + "loss": 0.2618, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07013794481754303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.044098547846078875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.184374438226223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998629033565521, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 900, + "weighted_flow_matching_loss": 0.07749143987894058, + "weighted_rkd_loss": 0.1843491569161415 + }, + { + "epoch": 0.26476578411405294, + "flow_matching_loss": 0.08080033883452416, + "grad_norm": 0.5401078462600708, + "learning_rate": 4.545e-06, + "loss": 0.2635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06967909187078476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04336307384073734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18272125869989395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998598217964172, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 910, + "weighted_flow_matching_loss": 0.08080033883452416, + "weighted_rkd_loss": 0.18269564658403398 + }, + { + "epoch": 0.26767529822519637, + "flow_matching_loss": 0.07794601619243621, + "grad_norm": 0.5362082123756409, + "learning_rate": 4.595e-06, + "loss": 0.2642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07088533118367195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04446708932518959, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18623775243759155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998567044734955, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 920, + "weighted_flow_matching_loss": 0.07794601619243621, + "weighted_rkd_loss": 0.18621106594800949 + }, + { + "epoch": 0.27058481233633985, + "flow_matching_loss": 0.07864850386977196, + "grad_norm": 0.5836489200592041, + "learning_rate": 4.6449999999999996e-06, + "loss": 0.2622, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0694857344031334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.044578717648983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18355018347501756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998535513877869, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 930, + "weighted_flow_matching_loss": 0.07864850386977196, + "weighted_rkd_loss": 0.18352330327033997 + }, + { + "epoch": 0.2734943264474833, + "flow_matching_loss": 0.07927864119410515, + "grad_norm": 0.6146027445793152, + "learning_rate": 4.695e-06, + "loss": 0.2629, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07006957158446311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04347199909389019, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1836111441254616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998503684997558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 940, + "weighted_flow_matching_loss": 0.07927864119410515, + "weighted_rkd_loss": 0.18358366340398788 + }, + { + "epoch": 0.2764038405586267, + "flow_matching_loss": 0.07836099341511726, + "grad_norm": 0.5101514458656311, + "learning_rate": 4.745e-06, + "loss": 0.2608, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0692834310233593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04392227865755558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18248913884162904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999847149848938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 950, + "weighted_flow_matching_loss": 0.07836099341511726, + "weighted_rkd_loss": 0.18246124535799027 + }, + { + "epoch": 0.27931335466977014, + "flow_matching_loss": 0.07928932160139084, + "grad_norm": 0.5424975752830505, + "learning_rate": 4.795e-06, + "loss": 0.2628, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06960703954100608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.044290094077587126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18350417166948318, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998438954353333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 960, + "weighted_flow_matching_loss": 0.07928932160139084, + "weighted_rkd_loss": 0.183475524187088 + }, + { + "epoch": 0.28222286878091357, + "flow_matching_loss": 0.07574984058737755, + "grad_norm": 0.5275181531906128, + "learning_rate": 4.845e-06, + "loss": 0.2574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06924681514501571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04318289831280708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18167652785778046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998406112194062, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 970, + "weighted_flow_matching_loss": 0.07574984058737755, + "weighted_rkd_loss": 0.18164757043123245 + }, + { + "epoch": 0.285132382892057, + "flow_matching_loss": 0.0762649081647396, + "grad_norm": 0.5173529386520386, + "learning_rate": 4.895e-06, + "loss": 0.2561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06848668456077575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04290928207337856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17988265305757523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998372852802276, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 980, + "weighted_flow_matching_loss": 0.0762649081647396, + "weighted_rkd_loss": 0.17985338866710662 + }, + { + "epoch": 0.2880418970032005, + "flow_matching_loss": 0.07472844682633877, + "grad_norm": 0.5191641449928284, + "learning_rate": 4.945e-06, + "loss": 0.2531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0682908795773983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04177856482565403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1783603236079216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998339295387269, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 990, + "weighted_flow_matching_loss": 0.07472844682633877, + "weighted_rkd_loss": 0.1783307045698166 + }, + { + "epoch": 0.2909514111143439, + "flow_matching_loss": 0.07585284411907196, + "grad_norm": 0.6133706569671631, + "learning_rate": 4.9950000000000005e-06, + "loss": 0.2559, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06870491728186608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04268012680113316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18008995950222015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998305380344391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1000, + "weighted_flow_matching_loss": 0.07585284411907196, + "weighted_rkd_loss": 0.1800594300031662 + }, + { + "epoch": 0.29386092522548735, + "flow_matching_loss": 0.0773917242884636, + "grad_norm": 0.5385105609893799, + "learning_rate": 5.045e-06, + "loss": 0.2563, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06815950050950051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.042602423951029776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1789214253425598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998271107673645, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1010, + "weighted_flow_matching_loss": 0.0773917242884636, + "weighted_rkd_loss": 0.1788904920220375 + }, + { + "epoch": 0.2967704393366308, + "flow_matching_loss": 0.0736224502325058, + "grad_norm": 0.567859411239624, + "learning_rate": 5.095e-06, + "loss": 0.2532, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06871055141091346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.042166759446263316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17958786487579345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999823659658432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1020, + "weighted_flow_matching_loss": 0.0736224502325058, + "weighted_rkd_loss": 0.17955619394779204 + }, + { + "epoch": 0.2996799534477742, + "flow_matching_loss": 0.0725199893116951, + "grad_norm": 0.6103768348693848, + "learning_rate": 5.145000000000001e-06, + "loss": 0.2521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06834366619586944, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04292744472622871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1796147793531418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998201608657837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1030, + "weighted_flow_matching_loss": 0.0725199893116951, + "weighted_rkd_loss": 0.17958247661590576 + }, + { + "epoch": 0.30258946755891764, + "flow_matching_loss": 0.07369378432631493, + "grad_norm": 0.5530255436897278, + "learning_rate": 5.195e-06, + "loss": 0.2469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06644493341445923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040345488116145134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1732353538274765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998166382312774, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1040, + "weighted_flow_matching_loss": 0.07369378432631493, + "weighted_rkd_loss": 0.17320359349250794 + }, + { + "epoch": 0.3054989816700611, + "flow_matching_loss": 0.07045650593936444, + "grad_norm": 0.5088580250740051, + "learning_rate": 5.245000000000001e-06, + "loss": 0.246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06700631007552146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04151693433523178, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17552955448627472, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998130798339844, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1050, + "weighted_flow_matching_loss": 0.07045650593936444, + "weighted_rkd_loss": 0.17549674361944198 + }, + { + "epoch": 0.30840849578120455, + "flow_matching_loss": 0.07271011024713517, + "grad_norm": 0.5144979357719421, + "learning_rate": 5.295e-06, + "loss": 0.2488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06750005409121514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.041124117001891135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1761242240667343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998094797134399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1060, + "weighted_flow_matching_loss": 0.07271011024713517, + "weighted_rkd_loss": 0.17609067261219025 + }, + { + "epoch": 0.311318009892348, + "flow_matching_loss": 0.07369350120425225, + "grad_norm": 0.5047844648361206, + "learning_rate": 5.3450000000000005e-06, + "loss": 0.2508, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06757325939834118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04198368601500988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17713020592927933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998058438301086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1070, + "weighted_flow_matching_loss": 0.07369350120425225, + "weighted_rkd_loss": 0.17709581404924393 + }, + { + "epoch": 0.3142275240034914, + "flow_matching_loss": 0.0714143730700016, + "grad_norm": 0.5518739819526672, + "learning_rate": 5.395000000000001e-06, + "loss": 0.2426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06569396965205669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03985601216554642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17124395221471786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998021781444549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1080, + "weighted_flow_matching_loss": 0.0714143730700016, + "weighted_rkd_loss": 0.17121006846427916 + }, + { + "epoch": 0.31713703811463484, + "flow_matching_loss": 0.07083910405635833, + "grad_norm": 0.5322349071502686, + "learning_rate": 5.445e-06, + "loss": 0.2399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06476588919758797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03957023806869984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16910201460123062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997984766960144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1090, + "weighted_flow_matching_loss": 0.07083910405635833, + "weighted_rkd_loss": 0.16906794160604477 + }, + { + "epoch": 0.32004655222577827, + "flow_matching_loss": 0.07371394783258438, + "grad_norm": 0.6063311696052551, + "learning_rate": 5.495e-06, + "loss": 0.2428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06473901383578777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.039666487276554106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16914451569318772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997947454452515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1100, + "weighted_flow_matching_loss": 0.07371394783258438, + "weighted_rkd_loss": 0.16910979598760606 + }, + { + "epoch": 0.32295606633692175, + "flow_matching_loss": 0.06910721287131309, + "grad_norm": 0.608826756477356, + "learning_rate": 5.5449999999999994e-06, + "loss": 0.2394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06518277041614055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03992423601448536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1702897772192955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997909784317016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1110, + "weighted_flow_matching_loss": 0.06910721287131309, + "weighted_rkd_loss": 0.17025418132543563 + }, + { + "epoch": 0.3258655804480652, + "flow_matching_loss": 0.0702458955347538, + "grad_norm": 0.5070309042930603, + "learning_rate": 5.595e-06, + "loss": 0.2373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06409506760537624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03891272135078907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1671028569340706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999787175655365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1120, + "weighted_flow_matching_loss": 0.0702458955347538, + "weighted_rkd_loss": 0.16706729978322982 + }, + { + "epoch": 0.3287750945592086, + "flow_matching_loss": 0.0734496284276247, + "grad_norm": 0.5757617354393005, + "learning_rate": 5.645e-06, + "loss": 0.2443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06519027538597584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04048117138445377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1708617240190506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997833371162415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1130, + "weighted_flow_matching_loss": 0.0734496284276247, + "weighted_rkd_loss": 0.17082470506429673 + }, + { + "epoch": 0.33168460867035204, + "flow_matching_loss": 0.07025745995342732, + "grad_norm": 0.5578871965408325, + "learning_rate": 5.695e-06, + "loss": 0.2422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0656081810593605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04072230085730553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1719386622309685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999779462814331, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1140, + "weighted_flow_matching_loss": 0.07025745995342732, + "weighted_rkd_loss": 0.171900737285614 + }, + { + "epoch": 0.3345941227814955, + "flow_matching_loss": 0.07419702112674713, + "grad_norm": 0.5173911452293396, + "learning_rate": 5.745e-06, + "loss": 0.247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06622696258127689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04039231613278389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17284624129533768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997755706310272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1150, + "weighted_flow_matching_loss": 0.07419702112674713, + "weighted_rkd_loss": 0.17280745059251784 + }, + { + "epoch": 0.3375036368926389, + "flow_matching_loss": 0.07229251861572265, + "grad_norm": 0.6461354494094849, + "learning_rate": 5.7949999999999996e-06, + "loss": 0.2404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0645807471126318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03900366611778736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16816516071558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997716307640075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1160, + "weighted_flow_matching_loss": 0.07229251861572265, + "weighted_rkd_loss": 0.16812676191329956 + }, + { + "epoch": 0.3404131510037824, + "flow_matching_loss": 0.06941298395395279, + "grad_norm": 0.5555940866470337, + "learning_rate": 5.845e-06, + "loss": 0.2381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06447470188140869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03973294347524643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1686823457479477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997676551342011, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1170, + "weighted_flow_matching_loss": 0.06941298395395279, + "weighted_rkd_loss": 0.16864315569400787 + }, + { + "epoch": 0.3433226651149258, + "flow_matching_loss": 0.07442597411572933, + "grad_norm": 0.5672925710678101, + "learning_rate": 5.895e-06, + "loss": 0.2415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06401483304798603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03905444480478763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16708411127328873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997636497020721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1180, + "weighted_flow_matching_loss": 0.07442597411572933, + "weighted_rkd_loss": 0.16704462319612504 + }, + { + "epoch": 0.34623217922606925, + "flow_matching_loss": 0.07075450383126736, + "grad_norm": 0.5880147218704224, + "learning_rate": 5.945e-06, + "loss": 0.2381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06385873667895794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.039658546820282935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16737602055072784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997596025466919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1190, + "weighted_flow_matching_loss": 0.07075450383126736, + "weighted_rkd_loss": 0.16733578890562056 + }, + { + "epoch": 0.3491416933372127, + "flow_matching_loss": 0.07053215801715851, + "grad_norm": 0.5888411998748779, + "learning_rate": 5.995e-06, + "loss": 0.2371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06382154077291488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.038949859142303464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16659294068813324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997555315494537, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1200, + "weighted_flow_matching_loss": 0.07053215801715851, + "weighted_rkd_loss": 0.16655221730470657 + }, + { + "epoch": 0.3520512074483561, + "flow_matching_loss": 0.07029289528727531, + "grad_norm": 0.566745400428772, + "learning_rate": 6.0450000000000006e-06, + "loss": 0.235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0628094106912613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03916594982147217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16478476971387862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997514188289642, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1210, + "weighted_flow_matching_loss": 0.07029289528727531, + "weighted_rkd_loss": 0.1647438019514084 + }, + { + "epoch": 0.35496072155949954, + "flow_matching_loss": 0.06885041818022727, + "grad_norm": 0.5218459963798523, + "learning_rate": 6.095e-06, + "loss": 0.2325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06285245642066002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.037948434427380565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1636533483862877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997472822666168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1220, + "weighted_flow_matching_loss": 0.06885041818022727, + "weighted_rkd_loss": 0.1636119842529297 + }, + { + "epoch": 0.357870235670643, + "flow_matching_loss": 0.06867560930550098, + "grad_norm": 0.5254201889038086, + "learning_rate": 6.1450000000000005e-06, + "loss": 0.2336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06305710934102535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.038821053132414816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16493526995182037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997430920600892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1230, + "weighted_flow_matching_loss": 0.06867560930550098, + "weighted_rkd_loss": 0.16489290148019792 + }, + { + "epoch": 0.36077974978178645, + "flow_matching_loss": 0.06775167994201184, + "grad_norm": 0.54854816198349, + "learning_rate": 6.195e-06, + "loss": 0.228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.061686261743307116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03691702485084534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16028954982757568, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997388839721679, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1240, + "weighted_flow_matching_loss": 0.06775167994201184, + "weighted_rkd_loss": 0.16024770140647887 + }, + { + "epoch": 0.3636892638929299, + "flow_matching_loss": 0.06657882817089558, + "grad_norm": 0.5349016189575195, + "learning_rate": 6.245e-06, + "loss": 0.2314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06330144591629505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03825621604919434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1648591101169586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99973464012146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1250, + "weighted_flow_matching_loss": 0.06657882817089558, + "weighted_rkd_loss": 0.1648153617978096 + }, + { + "epoch": 0.3665987780040733, + "flow_matching_loss": 0.06744631938636303, + "grad_norm": 0.5194674134254456, + "learning_rate": 6.295000000000001e-06, + "loss": 0.2288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06156456358730793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0382913913577795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16142051964998244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999730360507965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1260, + "weighted_flow_matching_loss": 0.06744631938636303, + "weighted_rkd_loss": 0.1613769918680191 + }, + { + "epoch": 0.36950829211521674, + "flow_matching_loss": 0.06777779161930084, + "grad_norm": 0.557066023349762, + "learning_rate": 6.345e-06, + "loss": 0.2281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06164591237902641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03705928847193718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16035111248493195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997260391712188, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1270, + "weighted_flow_matching_loss": 0.06777779161930084, + "weighted_rkd_loss": 0.1603071853518486 + }, + { + "epoch": 0.37241780622636017, + "flow_matching_loss": 0.06982056722044945, + "grad_norm": 0.5262861847877502, + "learning_rate": 6.395000000000001e-06, + "loss": 0.2322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06216110698878765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03806359581649303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1623858094215393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997216880321502, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1280, + "weighted_flow_matching_loss": 0.06982056722044945, + "weighted_rkd_loss": 0.1623406082391739 + }, + { + "epoch": 0.37532732033750366, + "flow_matching_loss": 0.06888850666582584, + "grad_norm": 0.4991404116153717, + "learning_rate": 6.445e-06, + "loss": 0.2275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06116843782365322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03628031872212887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1586171954870224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997173070907592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1290, + "weighted_flow_matching_loss": 0.06888850666582584, + "weighted_rkd_loss": 0.15857236087322235 + }, + { + "epoch": 0.3782368344486471, + "flow_matching_loss": 0.06648623161017894, + "grad_norm": 0.6345030665397644, + "learning_rate": 6.4950000000000005e-06, + "loss": 0.228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06223501041531563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03708642423152923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1615564450621605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997128903865814, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1300, + "weighted_flow_matching_loss": 0.06648623161017894, + "weighted_rkd_loss": 0.1615100622177124 + }, + { + "epoch": 0.3811463485597905, + "flow_matching_loss": 0.06610469743609429, + "grad_norm": 0.5499760508537292, + "learning_rate": 6.545e-06, + "loss": 0.2276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06175419762730598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.038000814989209174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16150921136140822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997084438800812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1310, + "weighted_flow_matching_loss": 0.06610469743609429, + "weighted_rkd_loss": 0.16146211326122284 + }, + { + "epoch": 0.38405586267093395, + "flow_matching_loss": 0.06715149693191051, + "grad_norm": 0.6076885461807251, + "learning_rate": 6.5949999999999995e-06, + "loss": 0.2293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06254096254706383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.037157300114631656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1622392252087593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997039496898651, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1320, + "weighted_flow_matching_loss": 0.06715149693191051, + "weighted_rkd_loss": 0.16219118982553482 + }, + { + "epoch": 0.3869653767820774, + "flow_matching_loss": 0.06785603538155556, + "grad_norm": 0.7049685716629028, + "learning_rate": 6.645e-06, + "loss": 0.2214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.059063471108675006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03542300723493099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15354994982481002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996994256973266, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1330, + "weighted_flow_matching_loss": 0.06785603538155556, + "weighted_rkd_loss": 0.15350379496812822 + }, + { + "epoch": 0.3898748908932208, + "flow_matching_loss": 0.06327464617788792, + "grad_norm": 0.5152153968811035, + "learning_rate": 6.695e-06, + "loss": 0.2213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06071299239993096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03669433631002903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1581203207373619, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996948719024659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1340, + "weighted_flow_matching_loss": 0.06327464617788792, + "weighted_rkd_loss": 0.1580720752477646 + }, + { + "epoch": 0.3927844050043643, + "flow_matching_loss": 0.06699395887553691, + "grad_norm": 0.5535424947738647, + "learning_rate": 6.745e-06, + "loss": 0.2307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06229067482054233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03913618959486485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16371753811836243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996902883052826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1350, + "weighted_flow_matching_loss": 0.06699395887553691, + "weighted_rkd_loss": 0.16366683840751647 + }, + { + "epoch": 0.3956939191155077, + "flow_matching_loss": 0.06599124856293201, + "grad_norm": 0.5967103242874146, + "learning_rate": 6.795e-06, + "loss": 0.2244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06070224419236183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03703798279166222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15844247192144395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996856570243835, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1360, + "weighted_flow_matching_loss": 0.06599124856293201, + "weighted_rkd_loss": 0.1583926722407341 + }, + { + "epoch": 0.39860343322665115, + "flow_matching_loss": 0.06894486509263516, + "grad_norm": 0.5964990258216858, + "learning_rate": 6.845e-06, + "loss": 0.2233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0590062391012907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.036426936835050584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15443941652774812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999681007862091, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1370, + "weighted_flow_matching_loss": 0.06894486509263516, + "weighted_rkd_loss": 0.15439014732837678 + }, + { + "epoch": 0.4015129473377946, + "flow_matching_loss": 0.06370963677763938, + "grad_norm": 0.47897326946258545, + "learning_rate": 6.895e-06, + "loss": 0.2196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06004271395504475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0358419107273221, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1559273421764374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996763110160828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1380, + "weighted_flow_matching_loss": 0.06370963677763938, + "weighted_rkd_loss": 0.15587686449289323 + }, + { + "epoch": 0.404422461448938, + "flow_matching_loss": 0.06531290002167225, + "grad_norm": 0.575153648853302, + "learning_rate": 6.945e-06, + "loss": 0.225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.061217959970235825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0373310636729002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15976698249578475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996715784072876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1390, + "weighted_flow_matching_loss": 0.06531290002167225, + "weighted_rkd_loss": 0.15971450954675676 + }, + { + "epoch": 0.4073319755600815, + "flow_matching_loss": 0.0690301276743412, + "grad_norm": 0.6014179587364197, + "learning_rate": 6.995e-06, + "loss": 0.2288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.061326338723301885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.037141126580536366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15979380458593367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996668100357056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1400, + "weighted_flow_matching_loss": 0.0690301276743412, + "weighted_rkd_loss": 0.159740549325943 + }, + { + "epoch": 0.4102414896712249, + "flow_matching_loss": 0.06743581406772137, + "grad_norm": 0.5392281413078308, + "learning_rate": 7.045e-06, + "loss": 0.2254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06070746220648289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03663009591400623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1580450192093849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996620357036591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1410, + "weighted_flow_matching_loss": 0.06743581406772137, + "weighted_rkd_loss": 0.15799160301685333 + }, + { + "epoch": 0.41315100378236835, + "flow_matching_loss": 0.06288356110453605, + "grad_norm": 0.5655887126922607, + "learning_rate": 7.095e-06, + "loss": 0.218, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.059709757193923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03579328842461109, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15521280467510223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996571958065033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1420, + "weighted_flow_matching_loss": 0.06288356110453605, + "weighted_rkd_loss": 0.15515959411859512 + }, + { + "epoch": 0.4160605178935118, + "flow_matching_loss": 0.06625121831893921, + "grad_norm": 0.5906344056129456, + "learning_rate": 7.145e-06, + "loss": 0.2214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05963132493197918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03591405525803566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15517670512199402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996523320674896, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1430, + "weighted_flow_matching_loss": 0.06625121831893921, + "weighted_rkd_loss": 0.15512275993824004 + }, + { + "epoch": 0.4189700320046552, + "flow_matching_loss": 0.06508337073028088, + "grad_norm": 0.49240076541900635, + "learning_rate": 7.1950000000000006e-06, + "loss": 0.2201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05956348329782486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03593606576323509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15506303161382676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996474385261536, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1440, + "weighted_flow_matching_loss": 0.06508337073028088, + "weighted_rkd_loss": 0.15500836223363876 + }, + { + "epoch": 0.42187954611579864, + "flow_matching_loss": 0.06499363891780377, + "grad_norm": 0.4849608838558197, + "learning_rate": 7.245e-06, + "loss": 0.2128, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05692158453166485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03401860650628805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14786177426576613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996425032615661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1450, + "weighted_flow_matching_loss": 0.06499363891780377, + "weighted_rkd_loss": 0.1478089153766632 + }, + { + "epoch": 0.42478906022694213, + "flow_matching_loss": 0.06487707942724227, + "grad_norm": 0.6255431175231934, + "learning_rate": 7.2950000000000005e-06, + "loss": 0.2186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.058711574599146846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03635252602398396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15377567559480668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996375381946564, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1460, + "weighted_flow_matching_loss": 0.06487707942724227, + "weighted_rkd_loss": 0.15371993631124498 + }, + { + "epoch": 0.42769857433808556, + "flow_matching_loss": 0.06420047096908092, + "grad_norm": 0.6471863985061646, + "learning_rate": 7.345e-06, + "loss": 0.2114, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05676689967513084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03375182319432497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14728562235832215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996325373649597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1470, + "weighted_flow_matching_loss": 0.06420047096908092, + "weighted_rkd_loss": 0.14723149836063384 + }, + { + "epoch": 0.430608088449229, + "flow_matching_loss": 0.062251050025224686, + "grad_norm": 0.4966970384120941, + "learning_rate": 7.395e-06, + "loss": 0.2192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.060381852090358734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.036194222047924995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15695792585611343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996275007724762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1480, + "weighted_flow_matching_loss": 0.062251050025224686, + "weighted_rkd_loss": 0.15689945966005325 + }, + { + "epoch": 0.4335176025603724, + "flow_matching_loss": 0.06347963325679302, + "grad_norm": 0.528215765953064, + "learning_rate": 7.445000000000001e-06, + "loss": 0.2137, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05779171362519264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03465836718678474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15024179518222808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996224343776703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1490, + "weighted_flow_matching_loss": 0.06347963325679302, + "weighted_rkd_loss": 0.15018507242202758 + }, + { + "epoch": 0.43642711667151585, + "flow_matching_loss": 0.06504569835960865, + "grad_norm": 0.5837044715881348, + "learning_rate": 7.495e-06, + "loss": 0.2213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.059630103036761285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.037055587768554686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15631579607725143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996173322200775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1500, + "weighted_flow_matching_loss": 0.06504569835960865, + "weighted_rkd_loss": 0.15625598430633544 + }, + { + "epoch": 0.4393366307826593, + "flow_matching_loss": 0.06552833914756775, + "grad_norm": 0.4657436013221741, + "learning_rate": 7.545000000000001e-06, + "loss": 0.2168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.058054639399051665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0352599972859025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15136927664279937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996121942996978, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1510, + "weighted_flow_matching_loss": 0.06552833914756775, + "weighted_rkd_loss": 0.15131058394908906 + }, + { + "epoch": 0.44224614489380276, + "flow_matching_loss": 0.06798660159111022, + "grad_norm": 0.5887717008590698, + "learning_rate": 7.595e-06, + "loss": 0.2207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05896533504128456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034806281141936776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15273695141077043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996070206165314, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1520, + "weighted_flow_matching_loss": 0.06798660159111022, + "weighted_rkd_loss": 0.1526769369840622 + }, + { + "epoch": 0.4451556590049462, + "flow_matching_loss": 0.0663789976388216, + "grad_norm": 0.4982853829860687, + "learning_rate": 7.645e-06, + "loss": 0.2165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05747719556093216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03522439878433943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15017878711223603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996018171310425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1530, + "weighted_flow_matching_loss": 0.0663789976388216, + "weighted_rkd_loss": 0.15011898726224898 + }, + { + "epoch": 0.4480651731160896, + "flow_matching_loss": 0.06401664018630981, + "grad_norm": 0.5439315438270569, + "learning_rate": 7.695e-06, + "loss": 0.2123, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05722065381705761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033899225667119025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14834053367376326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995965778827667, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1540, + "weighted_flow_matching_loss": 0.06401664018630981, + "weighted_rkd_loss": 0.1482806921005249 + }, + { + "epoch": 0.45097468722723305, + "flow_matching_loss": 0.06364850029349327, + "grad_norm": 0.5566914677619934, + "learning_rate": 7.745e-06, + "loss": 0.2121, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05697323754429817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03453664742410183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1484831228852272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995913028717041, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1550, + "weighted_flow_matching_loss": 0.06364850029349327, + "weighted_rkd_loss": 0.14842244535684584 + }, + { + "epoch": 0.4538842013383765, + "flow_matching_loss": 0.062320572510361674, + "grad_norm": 0.5023523569107056, + "learning_rate": 7.795e-06, + "loss": 0.211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0570021241903305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03476249873638153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14876675009727477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999585998058319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1560, + "weighted_flow_matching_loss": 0.062320572510361674, + "weighted_rkd_loss": 0.14870516955852509 + }, + { + "epoch": 0.4567937154495199, + "flow_matching_loss": 0.06256145089864731, + "grad_norm": 0.6226035952568054, + "learning_rate": 7.845e-06, + "loss": 0.2085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05636245645582676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033286947011947635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1460118591785431, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995806515216827, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1570, + "weighted_flow_matching_loss": 0.06256145089864731, + "weighted_rkd_loss": 0.1459506243467331 + }, + { + "epoch": 0.4597032295606634, + "flow_matching_loss": 0.06032957509160042, + "grad_norm": 0.5847052931785583, + "learning_rate": 7.895e-06, + "loss": 0.2112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.058083129301667216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03481562398374081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1509818822145462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999575287103653, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1580, + "weighted_flow_matching_loss": 0.06032957509160042, + "weighted_rkd_loss": 0.15091775953769684 + }, + { + "epoch": 0.4626127436718068, + "flow_matching_loss": 0.061622760444879535, + "grad_norm": 0.5904660820960999, + "learning_rate": 7.945e-06, + "loss": 0.2106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.057418493926525115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03416650388389826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1490034893155098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995698571205139, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1590, + "weighted_flow_matching_loss": 0.061622760444879535, + "weighted_rkd_loss": 0.14893940091133118 + }, + { + "epoch": 0.46552225778295026, + "flow_matching_loss": 0.06400530822575093, + "grad_norm": 0.5718005299568176, + "learning_rate": 7.995e-06, + "loss": 0.2128, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05697246938943863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03490398712456226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1488489255309105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995644152164459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1600, + "weighted_flow_matching_loss": 0.06400530822575093, + "weighted_rkd_loss": 0.14878408759832382 + }, + { + "epoch": 0.4684317718940937, + "flow_matching_loss": 0.06543061695992947, + "grad_norm": 0.5441356301307678, + "learning_rate": 8.045e-06, + "loss": 0.2162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05781193152070045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03522603362798691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1508498966693878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999558937549591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1610, + "weighted_flow_matching_loss": 0.06543061695992947, + "weighted_rkd_loss": 0.15078335851430893 + }, + { + "epoch": 0.4713412860052371, + "flow_matching_loss": 0.06005637682974339, + "grad_norm": 0.5139184594154358, + "learning_rate": 8.095e-06, + "loss": 0.2089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05732257291674614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034305027313530446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14895017296075821, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995534181594848, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1620, + "weighted_flow_matching_loss": 0.06005637682974339, + "weighted_rkd_loss": 0.1488836482167244 + }, + { + "epoch": 0.47425080011638054, + "flow_matching_loss": 0.06192058846354485, + "grad_norm": 0.6023942232131958, + "learning_rate": 8.145e-06, + "loss": 0.2045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05489470101892948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03285214547067881, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14264154732227324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995478808879852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1630, + "weighted_flow_matching_loss": 0.06192058846354485, + "weighted_rkd_loss": 0.1425770565867424 + }, + { + "epoch": 0.47716031422752403, + "flow_matching_loss": 0.061604549363255504, + "grad_norm": 0.566170871257782, + "learning_rate": 8.195e-06, + "loss": 0.2071, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05577701404690742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03402174152433872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14557576924562454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995422899723053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1640, + "weighted_flow_matching_loss": 0.061604549363255504, + "weighted_rkd_loss": 0.14550913125276566 + }, + { + "epoch": 0.48006982833866746, + "flow_matching_loss": 0.06476300209760666, + "grad_norm": 0.513106107711792, + "learning_rate": 8.245e-06, + "loss": 0.2073, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.054993638023734095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03259650208055973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14258377850055695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995366752147674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1650, + "weighted_flow_matching_loss": 0.06476300209760666, + "weighted_rkd_loss": 0.14251771718263626 + }, + { + "epoch": 0.4829793424498109, + "flow_matching_loss": 0.05662260577082634, + "grad_norm": 0.5480000376701355, + "learning_rate": 8.295000000000001e-06, + "loss": 0.1984, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05481728985905647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03221773318946362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14185231328010559, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995310306549072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1660, + "weighted_flow_matching_loss": 0.05662260577082634, + "weighted_rkd_loss": 0.1417857900261879 + }, + { + "epoch": 0.4858888565609543, + "flow_matching_loss": 0.0615281730890274, + "grad_norm": 0.5370072722434998, + "learning_rate": 8.345e-06, + "loss": 0.2062, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05585198104381561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03305229190737009, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14475625604391099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995253384113312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1670, + "weighted_flow_matching_loss": 0.0615281730890274, + "weighted_rkd_loss": 0.14468754827976227 + }, + { + "epoch": 0.48879837067209775, + "flow_matching_loss": 0.05889873430132866, + "grad_norm": 0.5032403469085693, + "learning_rate": 8.395e-06, + "loss": 0.1989, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05408766493201256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03192695062607527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1401022791862488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995196163654327, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1680, + "weighted_flow_matching_loss": 0.05889873430132866, + "weighted_rkd_loss": 0.14003497064113618 + }, + { + "epoch": 0.4917078847832412, + "flow_matching_loss": 0.06521276757121086, + "grad_norm": 0.5473218560218811, + "learning_rate": 8.445e-06, + "loss": 0.2107, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.056061172485351564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03346403986215592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14558638483285904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995138645172119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1690, + "weighted_flow_matching_loss": 0.06521276757121086, + "weighted_rkd_loss": 0.14551560580730438 + }, + { + "epoch": 0.49461739889438466, + "flow_matching_loss": 0.06366624012589454, + "grad_norm": 0.5961484909057617, + "learning_rate": 8.495e-06, + "loss": 0.2099, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05626364685595035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03380435667932034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14633165150880814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995080828666687, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1700, + "weighted_flow_matching_loss": 0.06366624012589454, + "weighted_rkd_loss": 0.146259668469429 + }, + { + "epoch": 0.4975269130055281, + "flow_matching_loss": 0.059329664707183837, + "grad_norm": 0.4951614737510681, + "learning_rate": 8.545e-06, + "loss": 0.1936, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.051983804628252986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030382227152585983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13434983789920807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995022535324096, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1710, + "weighted_flow_matching_loss": 0.059329664707183837, + "weighted_rkd_loss": 0.13428296968340875 + }, + { + "epoch": 0.5004364271166715, + "flow_matching_loss": 0.06120773889124394, + "grad_norm": 0.5691367983818054, + "learning_rate": 8.595e-06, + "loss": 0.2028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05463713444769382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03238313887268305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14165740609169006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994963943958283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1720, + "weighted_flow_matching_loss": 0.06120773889124394, + "weighted_rkd_loss": 0.14158606827259063 + }, + { + "epoch": 0.503345941227815, + "flow_matching_loss": 0.061692988872528075, + "grad_norm": 0.4862252175807953, + "learning_rate": 8.645000000000001e-06, + "loss": 0.2048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.054997653141617774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03315154574811459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.143146850168705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99949049949646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1730, + "weighted_flow_matching_loss": 0.061692988872528075, + "weighted_rkd_loss": 0.1430739253759384 + }, + { + "epoch": 0.5062554553389584, + "flow_matching_loss": 0.05917587354779243, + "grad_norm": 0.5357176065444946, + "learning_rate": 8.695e-06, + "loss": 0.1938, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0522031057626009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030327869392931462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13473408073186874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994845986366272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1740, + "weighted_flow_matching_loss": 0.05917587354779243, + "weighted_rkd_loss": 0.13466463312506677 + }, + { + "epoch": 0.5091649694501018, + "flow_matching_loss": 0.061027367040514945, + "grad_norm": 0.5258107781410217, + "learning_rate": 8.745e-06, + "loss": 0.1966, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.052643343433737756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030359761603176594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13564644902944564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994786381721497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1750, + "weighted_flow_matching_loss": 0.061027367040514945, + "weighted_rkd_loss": 0.13557572960853576 + }, + { + "epoch": 0.5120744835612453, + "flow_matching_loss": 0.06276959516108035, + "grad_norm": 0.6022255420684814, + "learning_rate": 8.795000000000001e-06, + "loss": 0.2048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0544386375695467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03318885415792465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14206612855196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994726240634918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1760, + "weighted_flow_matching_loss": 0.06276959516108035, + "weighted_rkd_loss": 0.14199119955301284 + }, + { + "epoch": 0.5149839976723887, + "flow_matching_loss": 0.061653146892786025, + "grad_norm": 0.5503136515617371, + "learning_rate": 8.845000000000001e-06, + "loss": 0.2029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05444587543606758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03246627543121576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14135802760720254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994666039943695, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1770, + "weighted_flow_matching_loss": 0.061653146892786025, + "weighted_rkd_loss": 0.1412826284766197 + }, + { + "epoch": 0.5178935117835322, + "flow_matching_loss": 0.05826074928045273, + "grad_norm": 0.5970315337181091, + "learning_rate": 8.895e-06, + "loss": 0.2005, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.054513607546687125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.033252977579832074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1422801911830902, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994605362415314, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1780, + "weighted_flow_matching_loss": 0.05826074928045273, + "weighted_rkd_loss": 0.14220343083143233 + }, + { + "epoch": 0.5208030258946756, + "flow_matching_loss": 0.05951859802007675, + "grad_norm": 0.5569679737091064, + "learning_rate": 8.945000000000002e-06, + "loss": 0.201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05452901944518089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03253267854452133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1415907174348831, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994544386863708, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1790, + "weighted_flow_matching_loss": 0.05951859802007675, + "weighted_rkd_loss": 0.14151346310973167 + }, + { + "epoch": 0.523712540005819, + "flow_matching_loss": 0.060251862928271296, + "grad_norm": 0.5545474886894226, + "learning_rate": 8.995000000000001e-06, + "loss": 0.1934, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.051605324074625966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02996988594532013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13318053558468818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994483113288879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1800, + "weighted_flow_matching_loss": 0.060251862928271296, + "weighted_rkd_loss": 0.13310706466436387 + }, + { + "epoch": 0.5266220541169625, + "flow_matching_loss": 0.058914313837885854, + "grad_norm": 0.5739119648933411, + "learning_rate": 9.045e-06, + "loss": 0.2003, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.054234101250767705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03302047215402126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14148867428302764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994421422481536, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1810, + "weighted_flow_matching_loss": 0.058914313837885854, + "weighted_rkd_loss": 0.1414097398519516 + }, + { + "epoch": 0.5295315682281059, + "flow_matching_loss": 0.0588684119284153, + "grad_norm": 0.5241110920906067, + "learning_rate": 9.095e-06, + "loss": 0.1944, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05243082866072655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030775579623878002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13563723638653755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994359493255616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1820, + "weighted_flow_matching_loss": 0.0588684119284153, + "weighted_rkd_loss": 0.13556072115898132 + }, + { + "epoch": 0.5324410823392494, + "flow_matching_loss": 0.05969197079539299, + "grad_norm": 0.5411851406097412, + "learning_rate": 9.145000000000001e-06, + "loss": 0.1922, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05142558105289936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02972402349114418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13257518634200097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999429714679718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1830, + "weighted_flow_matching_loss": 0.05969197079539299, + "weighted_rkd_loss": 0.1324995808303356 + }, + { + "epoch": 0.5353505964503927, + "flow_matching_loss": 0.06035718433558941, + "grad_norm": 0.48660674691200256, + "learning_rate": 9.195000000000001e-06, + "loss": 0.1978, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.052686404436826706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03210034836083651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13747315853834152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994234442710876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1840, + "weighted_flow_matching_loss": 0.06035718433558941, + "weighted_rkd_loss": 0.13739390224218367 + }, + { + "epoch": 0.5382601105615362, + "flow_matching_loss": 0.05890071392059326, + "grad_norm": 0.5946103930473328, + "learning_rate": 9.244999999999999e-06, + "loss": 0.1918, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.051232308894395825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0305102264508605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13297484517097474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994171440601349, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1850, + "weighted_flow_matching_loss": 0.05890071392059326, + "weighted_rkd_loss": 0.13289734497666358 + }, + { + "epoch": 0.5411696246726797, + "flow_matching_loss": 0.06087260916829109, + "grad_norm": 0.6671967506408691, + "learning_rate": 9.295e-06, + "loss": 0.1949, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05173885002732277, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0306344797834754, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1341121807694435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994108080863953, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1860, + "weighted_flow_matching_loss": 0.06087260916829109, + "weighted_rkd_loss": 0.1340331494808197 + }, + { + "epoch": 0.5440791387838231, + "flow_matching_loss": 0.056950221210718154, + "grad_norm": 0.5491672158241272, + "learning_rate": 9.345e-06, + "loss": 0.1957, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05347903408110142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.031923739425837996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13888180553913115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994044423103332, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1870, + "weighted_flow_matching_loss": 0.056950221210718154, + "weighted_rkd_loss": 0.13879909217357636 + }, + { + "epoch": 0.5469886528949666, + "flow_matching_loss": 0.0608572032302618, + "grad_norm": 0.6039802432060242, + "learning_rate": 9.394999999999999e-06, + "loss": 0.199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05288957953453064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03244976606220007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13822892308235168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993980348110199, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1880, + "weighted_flow_matching_loss": 0.0608572032302618, + "weighted_rkd_loss": 0.13814571350812913 + }, + { + "epoch": 0.5498981670061099, + "flow_matching_loss": 0.05846393778920174, + "grad_norm": 0.6238828301429749, + "learning_rate": 9.445e-06, + "loss": 0.1979, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.053696025907993314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03209012430161238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13948217332363128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993916034698487, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1890, + "weighted_flow_matching_loss": 0.05846393778920174, + "weighted_rkd_loss": 0.1393973097205162 + }, + { + "epoch": 0.5528076811172534, + "flow_matching_loss": 0.05904997624456883, + "grad_norm": 0.6168509125709534, + "learning_rate": 9.495e-06, + "loss": 0.188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04990872107446194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02918837871402502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12900582030415536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993851244449615, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1900, + "weighted_flow_matching_loss": 0.05904997624456883, + "weighted_rkd_loss": 0.12892649471759796 + }, + { + "epoch": 0.5557171952283969, + "flow_matching_loss": 0.05922347754240036, + "grad_norm": 0.5817856788635254, + "learning_rate": 9.545e-06, + "loss": 0.1899, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05035955309867859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030055581592023373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13077468648552895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993786215782166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1910, + "weighted_flow_matching_loss": 0.05922347754240036, + "weighted_rkd_loss": 0.13069342002272605 + }, + { + "epoch": 0.5586267093395403, + "flow_matching_loss": 0.05921272784471512, + "grad_norm": 0.598483145236969, + "learning_rate": 9.595e-06, + "loss": 0.1911, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05073543041944504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030451508797705175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13192237094044684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993720948696136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1920, + "weighted_flow_matching_loss": 0.05921272784471512, + "weighted_rkd_loss": 0.13183954879641532 + }, + { + "epoch": 0.5615362234506838, + "flow_matching_loss": 0.05982213653624058, + "grad_norm": 0.6609763503074646, + "learning_rate": 9.645e-06, + "loss": 0.1958, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05230250917375088, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.031450420059263705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13605543971061707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993655025959015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1930, + "weighted_flow_matching_loss": 0.05982213653624058, + "weighted_rkd_loss": 0.13596911132335662 + }, + { + "epoch": 0.5644457375618271, + "flow_matching_loss": 0.05923366993665695, + "grad_norm": 0.5758872628211975, + "learning_rate": 9.695e-06, + "loss": 0.1898, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050689331442117694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029318523965775965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13069718703627586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993588924407959, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1940, + "weighted_flow_matching_loss": 0.05923366993665695, + "weighted_rkd_loss": 0.13061339780688286 + }, + { + "epoch": 0.5673552516729706, + "flow_matching_loss": 0.06214020811021328, + "grad_norm": 0.557820200920105, + "learning_rate": 9.744999999999999e-06, + "loss": 0.197, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05233119688928127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030239536613225936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13490193113684654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993522465229034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1950, + "weighted_flow_matching_loss": 0.06214020811021328, + "weighted_rkd_loss": 0.1348145544528961 + }, + { + "epoch": 0.570264765784114, + "flow_matching_loss": 0.06590570881962776, + "grad_norm": 0.5173530578613281, + "learning_rate": 9.795e-06, + "loss": 0.1932, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0493722926825285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028644738718867302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12738932594656943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993455708026886, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1960, + "weighted_flow_matching_loss": 0.06590570881962776, + "weighted_rkd_loss": 0.12730595096945763 + }, + { + "epoch": 0.5731742798952575, + "flow_matching_loss": 0.0608719352632761, + "grad_norm": 0.5659254193305969, + "learning_rate": 9.845e-06, + "loss": 0.1857, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0488434549421072, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02723695747554302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12492386624217033, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993388593196869, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1970, + "weighted_flow_matching_loss": 0.0608719352632761, + "weighted_rkd_loss": 0.12484127581119538 + }, + { + "epoch": 0.576083794006401, + "flow_matching_loss": 0.05688398778438568, + "grad_norm": 0.6318198442459106, + "learning_rate": 9.895e-06, + "loss": 0.1925, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.052578403428196904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03053690865635872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1356937140226364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993321120738983, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1980, + "weighted_flow_matching_loss": 0.05688398778438568, + "weighted_rkd_loss": 0.13560309037566184 + }, + { + "epoch": 0.5789933081175443, + "flow_matching_loss": 0.05519126988947391, + "grad_norm": 0.5771257281303406, + "learning_rate": 9.945e-06, + "loss": 0.1836, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04962101317942143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029266407899558543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12850843518972396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993253290653229, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1990, + "weighted_flow_matching_loss": 0.05519126988947391, + "weighted_rkd_loss": 0.12842173129320145 + }, + { + "epoch": 0.5819028222286878, + "flow_matching_loss": 0.056063612923026085, + "grad_norm": 0.6162995100021362, + "learning_rate": 9.995e-06, + "loss": 0.1839, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04966877512633801, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028579832054674625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12791738137602807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993185102939606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2000, + "weighted_flow_matching_loss": 0.056063612923026085, + "weighted_rkd_loss": 0.1278302051126957 + }, + { + "epoch": 0.5848123363398312, + "flow_matching_loss": 0.05745677463710308, + "grad_norm": 0.5318905115127563, + "learning_rate": 1.0045e-05, + "loss": 0.1861, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.049890941753983495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028964916057884694, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12874679788947105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993116676807403, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2010, + "weighted_flow_matching_loss": 0.05745677463710308, + "weighted_rkd_loss": 0.12865817546844482 + }, + { + "epoch": 0.5877218504509747, + "flow_matching_loss": 0.05727297179400921, + "grad_norm": 0.583841860294342, + "learning_rate": 1.0095e-05, + "loss": 0.186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04972337260842323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029414447955787182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1288611948490143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993047833442688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2020, + "weighted_flow_matching_loss": 0.05727297179400921, + "weighted_rkd_loss": 0.12877160981297492 + }, + { + "epoch": 0.5906313645621182, + "flow_matching_loss": 0.056251831352710724, + "grad_norm": 0.5979245901107788, + "learning_rate": 1.0145e-05, + "loss": 0.1835, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04928588457405567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02877450976520777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12734627723693848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992978632450104, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2030, + "weighted_flow_matching_loss": 0.056251831352710724, + "weighted_rkd_loss": 0.1272568643093109 + }, + { + "epoch": 0.5935408786732616, + "flow_matching_loss": 0.059769130870699884, + "grad_norm": 0.631075382232666, + "learning_rate": 1.0195e-05, + "loss": 0.1912, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05059505887329578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030324496142566205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13151461333036424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992909133434296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2040, + "weighted_flow_matching_loss": 0.059769130870699884, + "weighted_rkd_loss": 0.13142134994268417 + }, + { + "epoch": 0.596450392784405, + "flow_matching_loss": 0.05894622057676315, + "grad_norm": 0.5414192080497742, + "learning_rate": 1.0245000000000001e-05, + "loss": 0.1915, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05109623372554779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0304981030523777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13269057124853134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992839217185974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2050, + "weighted_flow_matching_loss": 0.05894622057676315, + "weighted_rkd_loss": 0.13259554877877236 + }, + { + "epoch": 0.5993599068955484, + "flow_matching_loss": 0.05588063262403011, + "grad_norm": 0.4969513714313507, + "learning_rate": 1.0295e-05, + "loss": 0.1885, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0514107171446085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02991598900407553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13273742273449898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992769062519073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2060, + "weighted_flow_matching_loss": 0.05588063262403011, + "weighted_rkd_loss": 0.13264144212007523 + }, + { + "epoch": 0.6022694210066919, + "flow_matching_loss": 0.055544904246926305, + "grad_norm": 0.5248212814331055, + "learning_rate": 1.0345e-05, + "loss": 0.1851, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050246437266469005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029180870950222017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12967374548316002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992698609828949, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2070, + "weighted_flow_matching_loss": 0.055544904246926305, + "weighted_rkd_loss": 0.1295790709555149 + }, + { + "epoch": 0.6051789351178353, + "flow_matching_loss": 0.05444401279091835, + "grad_norm": 0.5446940064430237, + "learning_rate": 1.0395e-05, + "loss": 0.1779, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04815015867352486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027213107608258725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12351342141628266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992627620697021, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2080, + "weighted_flow_matching_loss": 0.05444401279091835, + "weighted_rkd_loss": 0.12342235967516899 + }, + { + "epoch": 0.6080884492289788, + "flow_matching_loss": 0.05835124403238297, + "grad_norm": 0.5906769633293152, + "learning_rate": 1.0445e-05, + "loss": 0.1888, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050565151125192644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029426968283951284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13055727034807205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992556571960449, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2090, + "weighted_flow_matching_loss": 0.05835124403238297, + "weighted_rkd_loss": 0.13046008497476577 + }, + { + "epoch": 0.6109979633401222, + "flow_matching_loss": 0.057180216908454894, + "grad_norm": 0.5388342142105103, + "learning_rate": 1.0495e-05, + "loss": 0.1854, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04932975508272648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029656135477125645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12831564620137215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992484986782074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2100, + "weighted_flow_matching_loss": 0.057180216908454894, + "weighted_rkd_loss": 0.12821922972798347 + }, + { + "epoch": 0.6139074774512656, + "flow_matching_loss": 0.05663008652627468, + "grad_norm": 0.5769010186195374, + "learning_rate": 1.0545e-05, + "loss": 0.1855, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04967079535126686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029659465327858923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12900105565786363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999241292476654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2110, + "weighted_flow_matching_loss": 0.05663008652627468, + "weighted_rkd_loss": 0.12890318557620048 + }, + { + "epoch": 0.6168169915624091, + "flow_matching_loss": 0.057153329253196716, + "grad_norm": 0.5369510650634766, + "learning_rate": 1.0595000000000001e-05, + "loss": 0.1822, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04866100661456585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027798892371356486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1251209072768688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992340624332428, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2120, + "weighted_flow_matching_loss": 0.057153329253196716, + "weighted_rkd_loss": 0.12502507269382476 + }, + { + "epoch": 0.6197265056735525, + "flow_matching_loss": 0.05545864962041378, + "grad_norm": 0.6144338846206665, + "learning_rate": 1.0645e-05, + "loss": 0.184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.049707948416471484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029226961173117162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12864285781979562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992268085479736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2130, + "weighted_flow_matching_loss": 0.05545864962041378, + "weighted_rkd_loss": 0.12854339480400084 + }, + { + "epoch": 0.622636019784696, + "flow_matching_loss": 0.05668513663113117, + "grad_norm": 0.6078145503997803, + "learning_rate": 1.0695e-05, + "loss": 0.1817, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04873569160699844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027603003941476346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1250743880867958, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992195188999176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2140, + "weighted_flow_matching_loss": 0.05668513663113117, + "weighted_rkd_loss": 0.12497676759958268 + }, + { + "epoch": 0.6255455338958394, + "flow_matching_loss": 0.05889851786196232, + "grad_norm": 0.5989040732383728, + "learning_rate": 1.0745000000000001e-05, + "loss": 0.1877, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04977468214929104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029320691153407095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12887005731463433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992121934890748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2150, + "weighted_flow_matching_loss": 0.05889851786196232, + "weighted_rkd_loss": 0.12876852974295616 + }, + { + "epoch": 0.6284550480069828, + "flow_matching_loss": 0.05870526991784573, + "grad_norm": 0.6053840517997742, + "learning_rate": 1.0795e-05, + "loss": 0.1776, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04633651077747345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026345959305763243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11901898011565208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999204820394516, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2160, + "weighted_flow_matching_loss": 0.05870526991784573, + "weighted_rkd_loss": 0.11892433613538742 + }, + { + "epoch": 0.6313645621181263, + "flow_matching_loss": 0.059469140321016314, + "grad_norm": 0.598585844039917, + "learning_rate": 1.0845e-05, + "loss": 0.1827, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.047507689148187635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02835962623357773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12337500602006912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991974294185638, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2170, + "weighted_flow_matching_loss": 0.059469140321016314, + "weighted_rkd_loss": 0.12327598705887795 + }, + { + "epoch": 0.6342740762292697, + "flow_matching_loss": 0.05788217037916184, + "grad_norm": 0.6081219911575317, + "learning_rate": 1.0895000000000002e-05, + "loss": 0.1856, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.049328619986772536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029189983569085597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12784722223877906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991900026798248, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2180, + "weighted_flow_matching_loss": 0.05788217037916184, + "weighted_rkd_loss": 0.12774365693330764 + }, + { + "epoch": 0.6371835903404132, + "flow_matching_loss": 0.05456397570669651, + "grad_norm": 0.6606774926185608, + "learning_rate": 1.0945000000000001e-05, + "loss": 0.1767, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04731599017977715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02764740288257599, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12227938324213028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991825342178344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2190, + "weighted_flow_matching_loss": 0.05456397570669651, + "weighted_rkd_loss": 0.12217942550778389 + }, + { + "epoch": 0.6400931044515565, + "flow_matching_loss": 0.054835814610123636, + "grad_norm": 0.5084576606750488, + "learning_rate": 1.0995e-05, + "loss": 0.1764, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04748329631984234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026722407154738902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12168900072574615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991750299930573, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2200, + "weighted_flow_matching_loss": 0.054835814610123636, + "weighted_rkd_loss": 0.12158860340714454 + }, + { + "epoch": 0.6430026185627, + "flow_matching_loss": 0.056157303601503374, + "grad_norm": 0.5988796353340149, + "learning_rate": 1.1045e-05, + "loss": 0.1816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.048756206408143044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027995575778186323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12550799027085305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991675019264221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2210, + "weighted_flow_matching_loss": 0.056157303601503374, + "weighted_rkd_loss": 0.12540350183844567 + }, + { + "epoch": 0.6459121326738435, + "flow_matching_loss": 0.056031345203518865, + "grad_norm": 0.5473451018333435, + "learning_rate": 1.1095000000000001e-05, + "loss": 0.1784, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0475093137472868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02743153218179941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12245015949010848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991599321365356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2220, + "weighted_flow_matching_loss": 0.056031345203518865, + "weighted_rkd_loss": 0.12234731167554855 + }, + { + "epoch": 0.6488216467849869, + "flow_matching_loss": 0.0521981131285429, + "grad_norm": 0.5448686480522156, + "learning_rate": 1.1145000000000001e-05, + "loss": 0.1761, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04775087833404541, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028552954271435737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1240547090768814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991523265838623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2230, + "weighted_flow_matching_loss": 0.0521981131285429, + "weighted_rkd_loss": 0.12394954636693001 + }, + { + "epoch": 0.6517311608961304, + "flow_matching_loss": 0.05451536551117897, + "grad_norm": 0.5668612718582153, + "learning_rate": 1.1195e-05, + "loss": 0.1776, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04792944118380547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027375820465385912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12323470339179039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991446912288666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2240, + "weighted_flow_matching_loss": 0.05451536551117897, + "weighted_rkd_loss": 0.12312930002808571 + }, + { + "epoch": 0.6546406750072737, + "flow_matching_loss": 0.05632656775414944, + "grad_norm": 0.5689826607704163, + "learning_rate": 1.1245000000000002e-05, + "loss": 0.1792, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04797345623373985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02698497548699379, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12293188795447349, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999137020111084, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2250, + "weighted_flow_matching_loss": 0.05632656775414944, + "weighted_rkd_loss": 0.12282580882310867 + }, + { + "epoch": 0.6575501891184172, + "flow_matching_loss": 0.057144058868288994, + "grad_norm": 0.573861300945282, + "learning_rate": 1.1295000000000001e-05, + "loss": 0.1796, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04761074781417847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027337335422635077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12255883067846299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991293251514435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2260, + "weighted_flow_matching_loss": 0.057144058868288994, + "weighted_rkd_loss": 0.12245212718844414 + }, + { + "epoch": 0.6604597032295607, + "flow_matching_loss": 0.05681418478488922, + "grad_norm": 0.5692538022994995, + "learning_rate": 1.1345e-05, + "loss": 0.1829, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04877195209264755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028612189553678037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12615609392523766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991215765476227, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2270, + "weighted_flow_matching_loss": 0.05681418478488922, + "weighted_rkd_loss": 0.12604527547955513 + }, + { + "epoch": 0.6633692173407041, + "flow_matching_loss": 0.055989108979701996, + "grad_norm": 0.5985810160636902, + "learning_rate": 1.1395e-05, + "loss": 0.1788, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04795950204133988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026994878984987737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12291388139128685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991138100624084, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2280, + "weighted_flow_matching_loss": 0.055989108979701996, + "weighted_rkd_loss": 0.12280495390295983 + }, + { + "epoch": 0.6662787314518476, + "flow_matching_loss": 0.054309210553765294, + "grad_norm": 0.4650351405143738, + "learning_rate": 1.1445e-05, + "loss": 0.174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046361105516552925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02711410876363516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11983631998300552, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991059958934784, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2290, + "weighted_flow_matching_loss": 0.054309210553765294, + "weighted_rkd_loss": 0.11972918882966041 + }, + { + "epoch": 0.669188245562991, + "flow_matching_loss": 0.05747399553656578, + "grad_norm": 0.6014924645423889, + "learning_rate": 1.1495e-05, + "loss": 0.1836, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04862220622599125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02903301492333412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12627742663025857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990981578826904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2300, + "weighted_flow_matching_loss": 0.05747399553656578, + "weighted_rkd_loss": 0.126163537055254 + }, + { + "epoch": 0.6720977596741344, + "flow_matching_loss": 0.05703159756958485, + "grad_norm": 0.5159274935722351, + "learning_rate": 1.1545e-05, + "loss": 0.1775, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046785329654812816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02705116309225559, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12062182128429413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990902841091156, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2310, + "weighted_flow_matching_loss": 0.05703159756958485, + "weighted_rkd_loss": 0.1205121025443077 + }, + { + "epoch": 0.6750072737852778, + "flow_matching_loss": 0.0522497046738863, + "grad_norm": 0.5688438415527344, + "learning_rate": 1.1595e-05, + "loss": 0.1736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04698408395051956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027495323121547698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12146349027752876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999082374572754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2320, + "weighted_flow_matching_loss": 0.0522497046738863, + "weighted_rkd_loss": 0.12135203331708908 + }, + { + "epoch": 0.6779167878964213, + "flow_matching_loss": 0.05659235417842865, + "grad_norm": 0.5863134264945984, + "learning_rate": 1.1645e-05, + "loss": 0.1756, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046250739693641664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026570160686969758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11907164007425308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990744292736053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2330, + "weighted_flow_matching_loss": 0.05659235417842865, + "weighted_rkd_loss": 0.11896142661571503 + }, + { + "epoch": 0.6808263020075648, + "flow_matching_loss": 0.05418669581413269, + "grad_norm": 0.5522921681404114, + "learning_rate": 1.1694999999999999e-05, + "loss": 0.1714, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04548601433634758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02638807911425829, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11736010760068893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990664541721344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2340, + "weighted_flow_matching_loss": 0.05418669581413269, + "weighted_rkd_loss": 0.11725054159760476 + }, + { + "epoch": 0.6837358161187082, + "flow_matching_loss": 0.05828862376511097, + "grad_norm": 0.5257053971290588, + "learning_rate": 1.1745e-05, + "loss": 0.1811, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04774191752076149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02743279580026865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12291662916541099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990584373474121, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2350, + "weighted_flow_matching_loss": 0.05828862376511097, + "weighted_rkd_loss": 0.12280089631676674 + }, + { + "epoch": 0.6866453302298516, + "flow_matching_loss": 0.05645950883626938, + "grad_norm": 0.5941452383995056, + "learning_rate": 1.1795e-05, + "loss": 0.18, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.048179618641734125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027310465276241303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12366970404982566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990503907203674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2360, + "weighted_flow_matching_loss": 0.05645950883626938, + "weighted_rkd_loss": 0.123552256077528 + }, + { + "epoch": 0.689554844340995, + "flow_matching_loss": 0.056659376993775365, + "grad_norm": 0.5282540321350098, + "learning_rate": 1.1845e-05, + "loss": 0.1764, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04635180123150349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02712025847285986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11982386186718941, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990423083305359, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2370, + "weighted_flow_matching_loss": 0.056659376993775365, + "weighted_rkd_loss": 0.11970910280942917 + }, + { + "epoch": 0.6924643584521385, + "flow_matching_loss": 0.05659511908888817, + "grad_norm": 0.5657620429992676, + "learning_rate": 1.1895e-05, + "loss": 0.1751, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04615673311054706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026298289373517036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1186117559671402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999034196138382, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2380, + "weighted_flow_matching_loss": 0.05659511908888817, + "weighted_rkd_loss": 0.11849720552563667 + }, + { + "epoch": 0.695373872563282, + "flow_matching_loss": 0.05388180911540985, + "grad_norm": 0.6099326610565186, + "learning_rate": 1.1945e-05, + "loss": 0.1731, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04641719311475754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026476644538342954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11931103020906449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990260481834412, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2390, + "weighted_flow_matching_loss": 0.05388180911540985, + "weighted_rkd_loss": 0.11919481977820397 + }, + { + "epoch": 0.6982833866744254, + "flow_matching_loss": 0.054277262091636656, + "grad_norm": 0.5374425053596497, + "learning_rate": 1.1995e-05, + "loss": 0.1745, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04674565494060516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02682925071567297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12032056078314782, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990178644657135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2400, + "weighted_flow_matching_loss": 0.054277262091636656, + "weighted_rkd_loss": 0.12020239606499672 + }, + { + "epoch": 0.7011929007855688, + "flow_matching_loss": 0.05628174804151058, + "grad_norm": 0.5781452059745789, + "learning_rate": 1.2045e-05, + "loss": 0.173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.045671135559678076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02551128976047039, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11685356125235558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999009644985199, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2410, + "weighted_flow_matching_loss": 0.05628174804151058, + "weighted_rkd_loss": 0.11673782989382744 + }, + { + "epoch": 0.7041024148967122, + "flow_matching_loss": 0.05517345778644085, + "grad_norm": 0.5275495052337646, + "learning_rate": 1.2095e-05, + "loss": 0.1765, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04692261815071106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02759029846638441, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1214355356991291, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999001395702362, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2420, + "weighted_flow_matching_loss": 0.05517345778644085, + "weighted_rkd_loss": 0.12131427302956581 + }, + { + "epoch": 0.7070119290078557, + "flow_matching_loss": 0.05734231360256672, + "grad_norm": 0.5565729737281799, + "learning_rate": 1.2145e-05, + "loss": 0.1733, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0453861977905035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025329582020640375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11610197648406029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989930987358093, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2430, + "weighted_flow_matching_loss": 0.05734231360256672, + "weighted_rkd_loss": 0.11598507091403007 + }, + { + "epoch": 0.7099214431189991, + "flow_matching_loss": 0.05567450821399689, + "grad_norm": 0.5965518951416016, + "learning_rate": 1.2195e-05, + "loss": 0.1729, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.045718560367822646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025916620902717114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11735374256968498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989848017692566, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2440, + "weighted_flow_matching_loss": 0.05567450821399689, + "weighted_rkd_loss": 0.11723460629582405 + }, + { + "epoch": 0.7128309572301426, + "flow_matching_loss": 0.05312982313334942, + "grad_norm": 0.5813331604003906, + "learning_rate": 1.2245e-05, + "loss": 0.1711, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046157519519329074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025738515332341193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11805355548858643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989764213562011, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2450, + "weighted_flow_matching_loss": 0.05312982313334942, + "weighted_rkd_loss": 0.11793271079659462 + }, + { + "epoch": 0.715740471341286, + "flow_matching_loss": 0.0586465310305357, + "grad_norm": 0.5897628664970398, + "learning_rate": 1.2295e-05, + "loss": 0.1789, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0468022245913744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026747549697756768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12035199999809265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989680409431457, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2460, + "weighted_flow_matching_loss": 0.0586465310305357, + "weighted_rkd_loss": 0.12022780179977417 + }, + { + "epoch": 0.7186499854524294, + "flow_matching_loss": 0.05442985258996487, + "grad_norm": 0.6100159287452698, + "learning_rate": 1.2345e-05, + "loss": 0.1732, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046106553822755816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02666490711271763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11887801364064217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989596247673035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2470, + "weighted_flow_matching_loss": 0.05442985258996487, + "weighted_rkd_loss": 0.11875433325767518 + }, + { + "epoch": 0.7215594995635729, + "flow_matching_loss": 0.055081892758607864, + "grad_norm": 0.5411867499351501, + "learning_rate": 1.2395e-05, + "loss": 0.1697, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.044972346350550654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02484261766076088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11478731110692024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989511609077454, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2480, + "weighted_flow_matching_loss": 0.055081892758607864, + "weighted_rkd_loss": 0.11466692090034485 + }, + { + "epoch": 0.7244690136747163, + "flow_matching_loss": 0.05255715772509575, + "grad_norm": 0.6000663638114929, + "learning_rate": 1.2445e-05, + "loss": 0.1654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04425825774669647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02446709331125021, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11298360899090767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989426732063293, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2490, + "weighted_flow_matching_loss": 0.05255715772509575, + "weighted_rkd_loss": 0.11286414787173271 + }, + { + "epoch": 0.7273785277858598, + "flow_matching_loss": 0.05207918658852577, + "grad_norm": 0.5701364874839783, + "learning_rate": 1.2495e-05, + "loss": 0.1661, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04441845789551735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02531902529299259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11415594294667245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989341378211976, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2500, + "weighted_flow_matching_loss": 0.05207918658852577, + "weighted_rkd_loss": 0.11403427198529244 + }, + { + "epoch": 0.7302880418970032, + "flow_matching_loss": 0.054718827828764915, + "grad_norm": 0.5836138725280762, + "learning_rate": 1.2545000000000001e-05, + "loss": 0.1676, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04372014626860619, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025593879260122777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1130341723561287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989255905151367, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2510, + "weighted_flow_matching_loss": 0.054718827828764915, + "weighted_rkd_loss": 0.11291272938251495 + }, + { + "epoch": 0.7331975560081466, + "flow_matching_loss": 0.05314548723399639, + "grad_norm": 0.6291625499725342, + "learning_rate": 1.2595e-05, + "loss": 0.171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04583504013717175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026317853853106497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11798793375492096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989169895648956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2520, + "weighted_flow_matching_loss": 0.05314548723399639, + "weighted_rkd_loss": 0.11786014139652252 + }, + { + "epoch": 0.7361070701192901, + "flow_matching_loss": 0.0513173758983612, + "grad_norm": 0.5390762090682983, + "learning_rate": 1.2645e-05, + "loss": 0.1691, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04574361518025398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02643072810024023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11791795641183853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989083707332611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2530, + "weighted_flow_matching_loss": 0.0513173758983612, + "weighted_rkd_loss": 0.11778924018144607 + }, + { + "epoch": 0.7390165842304335, + "flow_matching_loss": 0.05565151870250702, + "grad_norm": 0.5704774260520935, + "learning_rate": 1.2695000000000001e-05, + "loss": 0.169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.044050668552517894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02539216950535774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11349350735545158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988997042179107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2540, + "weighted_flow_matching_loss": 0.05565151870250702, + "weighted_rkd_loss": 0.11336862742900848 + }, + { + "epoch": 0.741926098341577, + "flow_matching_loss": 0.0531316414475441, + "grad_norm": 0.5508947372436523, + "learning_rate": 1.2745e-05, + "loss": 0.167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0441339198499918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02576029021292925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11402812898159027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988910019397735, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2550, + "weighted_flow_matching_loss": 0.0531316414475441, + "weighted_rkd_loss": 0.11390167102217674 + }, + { + "epoch": 0.7448356124527203, + "flow_matching_loss": 0.05425988174974918, + "grad_norm": 0.7449724674224854, + "learning_rate": 1.2795e-05, + "loss": 0.1734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04590812586247921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027417468838393688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11923372074961662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988822758197784, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2560, + "weighted_flow_matching_loss": 0.05425988174974918, + "weighted_rkd_loss": 0.11910044401884079 + }, + { + "epoch": 0.7477451265638638, + "flow_matching_loss": 0.053907360136508944, + "grad_norm": 0.5694136619567871, + "learning_rate": 1.2845e-05, + "loss": 0.17, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04524196572601795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025767226703464985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11625115796923638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998873507976532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2570, + "weighted_flow_matching_loss": 0.053907360136508944, + "weighted_rkd_loss": 0.11612019687891006 + }, + { + "epoch": 0.7506546406750073, + "flow_matching_loss": 0.05246729627251625, + "grad_norm": 0.576932430267334, + "learning_rate": 1.2895000000000001e-05, + "loss": 0.1639, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043515390157699584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024530203081667423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11156098246574402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988647103309631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2580, + "weighted_flow_matching_loss": 0.05246729627251625, + "weighted_rkd_loss": 0.11143433153629304 + }, + { + "epoch": 0.7535641547861507, + "flow_matching_loss": 0.05289988480508327, + "grad_norm": 0.5577213168144226, + "learning_rate": 1.2945e-05, + "loss": 0.1651, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04355113692581654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025199761241674425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1123020350933075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988558828830719, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2590, + "weighted_flow_matching_loss": 0.05289988480508327, + "weighted_rkd_loss": 0.11217353865504265 + }, + { + "epoch": 0.7564736688972942, + "flow_matching_loss": 0.05495758652687073, + "grad_norm": 0.5407896041870117, + "learning_rate": 1.2995e-05, + "loss": 0.1673, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043958405032753944, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024529472179710864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11244628131389618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988470137119293, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2600, + "weighted_flow_matching_loss": 0.05495758652687073, + "weighted_rkd_loss": 0.11231663748621941 + }, + { + "epoch": 0.7593831830084375, + "flow_matching_loss": 0.04942953772842884, + "grad_norm": 0.5484484434127808, + "learning_rate": 1.3045000000000001e-05, + "loss": 0.1617, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043407590314745906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025586587190628052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11240176856517792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988380968570709, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2610, + "weighted_flow_matching_loss": 0.04942953772842884, + "weighted_rkd_loss": 0.11227116659283638 + }, + { + "epoch": 0.762292697119581, + "flow_matching_loss": 0.05434408113360405, + "grad_norm": 0.6276536583900452, + "learning_rate": 1.3095e-05, + "loss": 0.1659, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04331752434372902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02510126680135727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11173631623387337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988291561603546, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2620, + "weighted_flow_matching_loss": 0.05434408113360405, + "weighted_rkd_loss": 0.11160549446940422 + }, + { + "epoch": 0.7652022112307245, + "flow_matching_loss": 0.053367406129837036, + "grad_norm": 0.666516900062561, + "learning_rate": 1.3145e-05, + "loss": 0.1638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04320166856050491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024162117019295694, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11056545227766038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988202095031739, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2630, + "weighted_flow_matching_loss": 0.053367406129837036, + "weighted_rkd_loss": 0.11043500900268555 + }, + { + "epoch": 0.7681117253418679, + "flow_matching_loss": 0.0504326980561018, + "grad_norm": 0.5512955784797668, + "learning_rate": 1.3195000000000002e-05, + "loss": 0.162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043231357261538504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025269602611660957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11173231825232506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988111972808837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2640, + "weighted_flow_matching_loss": 0.0504326980561018, + "weighted_rkd_loss": 0.11159949153661727 + }, + { + "epoch": 0.7710212394530114, + "flow_matching_loss": 0.052563485130667685, + "grad_norm": 0.6003878116607666, + "learning_rate": 1.3245000000000001e-05, + "loss": 0.1623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04239666648209095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025069673918187617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10986300632357597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988021612167358, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2650, + "weighted_flow_matching_loss": 0.052563485130667685, + "weighted_rkd_loss": 0.10973140969872475 + }, + { + "epoch": 0.7739307535641547, + "flow_matching_loss": 0.05283413864672184, + "grad_norm": 0.5874362587928772, + "learning_rate": 1.3295e-05, + "loss": 0.164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043439144268631935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024413756839931012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1112920455634594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987930953502655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2660, + "weighted_flow_matching_loss": 0.05283413864672184, + "weighted_rkd_loss": 0.11115772500634194 + }, + { + "epoch": 0.7768402676752982, + "flow_matching_loss": 0.05240903533995152, + "grad_norm": 0.5988026857376099, + "learning_rate": 1.3345000000000002e-05, + "loss": 0.1622, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04305089898407459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023869572021067143, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10997136980295182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987839818000793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2670, + "weighted_flow_matching_loss": 0.05240903533995152, + "weighted_rkd_loss": 0.10983763858675957 + }, + { + "epoch": 0.7797497817864416, + "flow_matching_loss": 0.05362359099090099, + "grad_norm": 0.560093879699707, + "learning_rate": 1.3395000000000001e-05, + "loss": 0.168, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04463940560817718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02525593601167202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11453474760055542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987748384475708, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2680, + "weighted_flow_matching_loss": 0.05362359099090099, + "weighted_rkd_loss": 0.11439443677663803 + }, + { + "epoch": 0.7826592958975851, + "flow_matching_loss": 0.05123208425939083, + "grad_norm": 0.5489411950111389, + "learning_rate": 1.3445000000000001e-05, + "loss": 0.1652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.044061199948191646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025963317416608334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1140857182443142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987656652927399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2690, + "weighted_flow_matching_loss": 0.05123208425939083, + "weighted_rkd_loss": 0.11394490152597428 + }, + { + "epoch": 0.7855688100087286, + "flow_matching_loss": 0.05210009217262268, + "grad_norm": 0.5880982875823975, + "learning_rate": 1.3495e-05, + "loss": 0.159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04159187376499176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023804157227277755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10698790475726128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987564563751221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2700, + "weighted_flow_matching_loss": 0.05210009217262268, + "weighted_rkd_loss": 0.10685486048460006 + }, + { + "epoch": 0.788478324119872, + "flow_matching_loss": 0.051075038686394694, + "grad_norm": 0.49095407128334045, + "learning_rate": 1.3545e-05, + "loss": 0.1673, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04520452804863453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025959143042564393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11636819913983346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987472116947174, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2710, + "weighted_flow_matching_loss": 0.051075038686394694, + "weighted_rkd_loss": 0.1162224143743515 + }, + { + "epoch": 0.7913878382310154, + "flow_matching_loss": 0.05324024297297001, + "grad_norm": 0.5155226588249207, + "learning_rate": 1.3595e-05, + "loss": 0.1643, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04331573583185673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024521791003644468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11115326285362244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987379312515259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2720, + "weighted_flow_matching_loss": 0.05324024297297001, + "weighted_rkd_loss": 0.11101298108696937 + }, + { + "epoch": 0.7942973523421588, + "flow_matching_loss": 0.050177646800875664, + "grad_norm": 0.5193005800247192, + "learning_rate": 1.3644999999999999e-05, + "loss": 0.1638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043813129886984825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026114450581371785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1137407086789608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998728621006012, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2730, + "weighted_flow_matching_loss": 0.050177646800875664, + "weighted_rkd_loss": 0.11359611228108406 + }, + { + "epoch": 0.7972068664533023, + "flow_matching_loss": 0.054588022828102115, + "grad_norm": 0.5588687062263489, + "learning_rate": 1.3695e-05, + "loss": 0.1672, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04399482011795044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024756692349910736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1127463310956955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987192749977112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2740, + "weighted_flow_matching_loss": 0.054588022828102115, + "weighted_rkd_loss": 0.11260192468762398 + }, + { + "epoch": 0.8001163805644458, + "flow_matching_loss": 0.05295138694345951, + "grad_norm": 0.49491164088249207, + "learning_rate": 1.3745e-05, + "loss": 0.1577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04075958132743836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023342208191752433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10486137121915817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998709899187088, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2750, + "weighted_flow_matching_loss": 0.05295138694345951, + "weighted_rkd_loss": 0.10472609624266624 + }, + { + "epoch": 0.8030258946755892, + "flow_matching_loss": 0.055555862560868266, + "grad_norm": 0.6048070788383484, + "learning_rate": 1.3795e-05, + "loss": 0.1684, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043970583379268645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02500629648566246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11294746473431587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998700475692749, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2760, + "weighted_flow_matching_loss": 0.055555862560868266, + "weighted_rkd_loss": 0.11280068829655647 + }, + { + "epoch": 0.8059354087867326, + "flow_matching_loss": 0.05340900458395481, + "grad_norm": 0.4826850891113281, + "learning_rate": 1.3845e-05, + "loss": 0.1613, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04216344840824604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02369567658752203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10802257284522057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986910343170166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2770, + "weighted_flow_matching_loss": 0.05340900458395481, + "weighted_rkd_loss": 0.10788118839263916 + }, + { + "epoch": 0.808844922897876, + "flow_matching_loss": 0.05391257219016552, + "grad_norm": 0.6263678073883057, + "learning_rate": 1.3895e-05, + "loss": 0.1596, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04108302630484104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023686864227056504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10585291758179664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986815452575684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2780, + "weighted_flow_matching_loss": 0.05391257219016552, + "weighted_rkd_loss": 0.10571335703134536 + }, + { + "epoch": 0.8117544370090195, + "flow_matching_loss": 0.05090631544589996, + "grad_norm": 0.5610925555229187, + "learning_rate": 1.3945e-05, + "loss": 0.1609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0430491179227829, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024030387029051782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11012862399220466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986720085144043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2790, + "weighted_flow_matching_loss": 0.05090631544589996, + "weighted_rkd_loss": 0.10998236089944839 + }, + { + "epoch": 0.814663951120163, + "flow_matching_loss": 0.05159410461783409, + "grad_norm": 0.5696685314178467, + "learning_rate": 1.3995e-05, + "loss": 0.1631, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043393485620617865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024900772795081138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11168774440884591, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986624717712402, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2800, + "weighted_flow_matching_loss": 0.05159410461783409, + "weighted_rkd_loss": 0.11153836026787758 + }, + { + "epoch": 0.8175734652313064, + "flow_matching_loss": 0.04860458113253117, + "grad_norm": 0.5256023406982422, + "learning_rate": 1.4045e-05, + "loss": 0.1534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04088545143604279, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023178218491375448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10494912043213844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986528992652893, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2810, + "weighted_flow_matching_loss": 0.04860458113253117, + "weighted_rkd_loss": 0.1048077516257763 + }, + { + "epoch": 0.8204829793424498, + "flow_matching_loss": 0.05062551833689213, + "grad_norm": 0.5607681274414062, + "learning_rate": 1.4095e-05, + "loss": 0.1609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.042934360727667806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024560067802667618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1104287900030613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986432790756226, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2820, + "weighted_flow_matching_loss": 0.05062551833689213, + "weighted_rkd_loss": 0.11027897298336028 + }, + { + "epoch": 0.8233924934535932, + "flow_matching_loss": 0.052995895966887474, + "grad_norm": 0.5666263699531555, + "learning_rate": 1.4145e-05, + "loss": 0.1604, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04185772426426411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02385660782456398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10757205709815025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998633623123169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2830, + "weighted_flow_matching_loss": 0.052995895966887474, + "weighted_rkd_loss": 0.1074250727891922 + }, + { + "epoch": 0.8263020075647367, + "flow_matching_loss": 0.052753083780407904, + "grad_norm": 0.5312226414680481, + "learning_rate": 1.4195e-05, + "loss": 0.1619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04291415587067604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023431144841015338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1092594563961029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986239314079285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2840, + "weighted_flow_matching_loss": 0.052753083780407904, + "weighted_rkd_loss": 0.10910912305116653 + }, + { + "epoch": 0.8292115216758801, + "flow_matching_loss": 0.05327401794493199, + "grad_norm": 0.49518048763275146, + "learning_rate": 1.4245e-05, + "loss": 0.1604, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041857875883579254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023538301326334478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10725405290722848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986142218112946, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2850, + "weighted_flow_matching_loss": 0.05327401794493199, + "weighted_rkd_loss": 0.1071054294705391 + }, + { + "epoch": 0.8321210357870236, + "flow_matching_loss": 0.05146398805081844, + "grad_norm": 0.5571969747543335, + "learning_rate": 1.4295e-05, + "loss": 0.1664, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04463683553040028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025798938050866128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11507260873913765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986044585704803, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2860, + "weighted_flow_matching_loss": 0.05146398805081844, + "weighted_rkd_loss": 0.1149120382964611 + }, + { + "epoch": 0.835030549898167, + "flow_matching_loss": 0.05408451668918133, + "grad_norm": 0.519325852394104, + "learning_rate": 1.4345e-05, + "loss": 0.1646, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04288436807692051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024892499670386314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1106612354516983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985946655273438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2870, + "weighted_flow_matching_loss": 0.05408451668918133, + "weighted_rkd_loss": 0.11050572246313095 + }, + { + "epoch": 0.8379400640093104, + "flow_matching_loss": 0.0491247083991766, + "grad_norm": 0.54548579454422, + "learning_rate": 1.4395e-05, + "loss": 0.1531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.040758980065584184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02263985313475132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10415781363844871, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985848367214203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2880, + "weighted_flow_matching_loss": 0.0491247083991766, + "weighted_rkd_loss": 0.10401041805744171 + }, + { + "epoch": 0.8408495781204539, + "flow_matching_loss": 0.05026371255517006, + "grad_norm": 0.5572834014892578, + "learning_rate": 1.4445e-05, + "loss": 0.1562, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04154955632984638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022981259413063525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10608037188649178, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985749900341034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2890, + "weighted_flow_matching_loss": 0.05026371255517006, + "weighted_rkd_loss": 0.10592920929193497 + }, + { + "epoch": 0.8437590922315973, + "flow_matching_loss": 0.05046714022755623, + "grad_norm": 0.6165193915367126, + "learning_rate": 1.4495000000000001e-05, + "loss": 0.1567, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041202500090003015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023986118845641613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10639111995697022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985650897026062, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2900, + "weighted_flow_matching_loss": 0.05046714022755623, + "weighted_rkd_loss": 0.10623844414949417 + }, + { + "epoch": 0.8466686063427408, + "flow_matching_loss": 0.04894876703619957, + "grad_norm": 0.6088021397590637, + "learning_rate": 1.4545e-05, + "loss": 0.1609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043784454092383386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024515938013792038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11208484694361687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998555165529251, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2910, + "weighted_flow_matching_loss": 0.04894876703619957, + "weighted_rkd_loss": 0.11192289739847183 + }, + { + "epoch": 0.8495781204538843, + "flow_matching_loss": 0.05146616101264954, + "grad_norm": 0.7471389174461365, + "learning_rate": 1.4595e-05, + "loss": 0.158, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04154739938676357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023590316995978355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10668511539697648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985451996326447, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2920, + "weighted_flow_matching_loss": 0.05146616101264954, + "weighted_rkd_loss": 0.10652989968657493 + }, + { + "epoch": 0.8524876345650276, + "flow_matching_loss": 0.05200444906949997, + "grad_norm": 0.5661550760269165, + "learning_rate": 1.4645000000000001e-05, + "loss": 0.1599, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041895124688744546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024282733164727687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10807298347353936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985352039337159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2930, + "weighted_flow_matching_loss": 0.05200444906949997, + "weighted_rkd_loss": 0.10791468396782875 + }, + { + "epoch": 0.8553971486761711, + "flow_matching_loss": 0.05357785820960999, + "grad_norm": 0.554816484451294, + "learning_rate": 1.4695e-05, + "loss": 0.1631, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.042803014442324636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024084057472646237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10969008654356002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985251784324646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2940, + "weighted_flow_matching_loss": 0.05357785820960999, + "weighted_rkd_loss": 0.10952831134200096 + }, + { + "epoch": 0.8583066627873145, + "flow_matching_loss": 0.05019809529185295, + "grad_norm": 0.5117571353912354, + "learning_rate": 1.4745e-05, + "loss": 0.1555, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041104314476251604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023227435164153577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10543606355786324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998515111207962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2950, + "weighted_flow_matching_loss": 0.05019809529185295, + "weighted_rkd_loss": 0.10527950003743172 + }, + { + "epoch": 0.861216176898458, + "flow_matching_loss": 0.050238396227359775, + "grad_norm": 0.5365942120552063, + "learning_rate": 1.4795e-05, + "loss": 0.1568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041618795692920686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023483802378177644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1067213922739029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998505026102066, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2960, + "weighted_flow_matching_loss": 0.050238396227359775, + "weighted_rkd_loss": 0.106561841070652 + }, + { + "epoch": 0.8641256910096013, + "flow_matching_loss": 0.05111735016107559, + "grad_norm": 0.5043098330497742, + "learning_rate": 1.4845000000000001e-05, + "loss": 0.155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04073482677340508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022542841732501984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10401249527931214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984948933124542, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2970, + "weighted_flow_matching_loss": 0.05111735016107559, + "weighted_rkd_loss": 0.1038559541106224 + }, + { + "epoch": 0.8670352051207448, + "flow_matching_loss": 0.05079076886177063, + "grad_norm": 0.6516866683959961, + "learning_rate": 1.4895e-05, + "loss": 0.1562, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04147014617919922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022621745243668556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10556203797459603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984847187995911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2980, + "weighted_flow_matching_loss": 0.05079076886177063, + "weighted_rkd_loss": 0.10540207251906394 + }, + { + "epoch": 0.8699447192318883, + "flow_matching_loss": 0.048941899463534354, + "grad_norm": 0.4899444282054901, + "learning_rate": 1.4945e-05, + "loss": 0.1471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03845195695757866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021375772170722486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09827968552708626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99847452044487, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2990, + "weighted_flow_matching_loss": 0.048941899463534354, + "weighted_rkd_loss": 0.09812976941466331 + }, + { + "epoch": 0.8728542333430317, + "flow_matching_loss": 0.05082063153386116, + "grad_norm": 0.5132097601890564, + "learning_rate": 1.4995000000000001e-05, + "loss": 0.1536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04046923145651817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0219767726957798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10291523560881614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984642863273621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3000, + "weighted_flow_matching_loss": 0.05082063153386116, + "weighted_rkd_loss": 0.10275718942284584 + }, + { + "epoch": 0.8757637474541752, + "flow_matching_loss": 0.05413910299539566, + "grad_norm": 0.45190688967704773, + "learning_rate": 1.5044999999999999e-05, + "loss": 0.1561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03961932510137558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02285395786166191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10209260731935502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984540164470672, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3010, + "weighted_flow_matching_loss": 0.05413910299539566, + "weighted_rkd_loss": 0.10193478167057038 + }, + { + "epoch": 0.8786732615653186, + "flow_matching_loss": 0.04831705987453461, + "grad_norm": 0.57741779088974, + "learning_rate": 1.5095e-05, + "loss": 0.1556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04200905784964561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023469840362668037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1074879564344883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984437108039856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3020, + "weighted_flow_matching_loss": 0.04831705987453461, + "weighted_rkd_loss": 0.10732067078351974 + }, + { + "epoch": 0.881582775676462, + "flow_matching_loss": 0.050993964076042175, + "grad_norm": 0.541658341884613, + "learning_rate": 1.5145000000000002e-05, + "loss": 0.1509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039277638867497446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02146726492792368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10002254247665406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984333753585816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3030, + "weighted_flow_matching_loss": 0.050993964076042175, + "weighted_rkd_loss": 0.09986585453152656 + }, + { + "epoch": 0.8844922897876055, + "flow_matching_loss": 0.050450551137328145, + "grad_norm": 0.4737790524959564, + "learning_rate": 1.5195e-05, + "loss": 0.1564, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04126228615641594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02357132900506258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10609590262174606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984230041503906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3040, + "weighted_flow_matching_loss": 0.050450551137328145, + "weighted_rkd_loss": 0.1059285931289196 + }, + { + "epoch": 0.8874018038987489, + "flow_matching_loss": 0.04742561839520931, + "grad_norm": 0.5675654411315918, + "learning_rate": 1.5245e-05, + "loss": 0.1513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04051504619419575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023031195998191832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10406128764152527, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984126031398773, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3050, + "weighted_flow_matching_loss": 0.04742561839520931, + "weighted_rkd_loss": 0.10389609336853027 + }, + { + "epoch": 0.8903113180098924, + "flow_matching_loss": 0.04939896836876869, + "grad_norm": 0.5063665509223938, + "learning_rate": 1.5295000000000002e-05, + "loss": 0.15, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03950835652649402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021783088520169257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10079980120062829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984021604061126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3060, + "weighted_flow_matching_loss": 0.04939896836876869, + "weighted_rkd_loss": 0.10063874199986458 + }, + { + "epoch": 0.8932208321210358, + "flow_matching_loss": 0.049150871485471724, + "grad_norm": 0.6538882851600647, + "learning_rate": 1.5345e-05, + "loss": 0.1509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03993519097566604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022052730433642864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10192311182618141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983916878700256, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3070, + "weighted_flow_matching_loss": 0.049150871485471724, + "weighted_rkd_loss": 0.10175918191671371 + }, + { + "epoch": 0.8961303462321792, + "flow_matching_loss": 0.05002770014107227, + "grad_norm": 0.5958120822906494, + "learning_rate": 1.5395e-05, + "loss": 0.1582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04238907136023044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02354912906885147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10832727178931237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983811795711517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3080, + "weighted_flow_matching_loss": 0.05002770014107227, + "weighted_rkd_loss": 0.1081518903374672 + }, + { + "epoch": 0.8990398603433226, + "flow_matching_loss": 0.04923038110136986, + "grad_norm": 0.555134117603302, + "learning_rate": 1.5445000000000002e-05, + "loss": 0.1535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04075954966247082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022893472760915756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10441257357597351, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998370635509491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3090, + "weighted_flow_matching_loss": 0.04923038110136986, + "weighted_rkd_loss": 0.10424245521426201 + }, + { + "epoch": 0.9019493744544661, + "flow_matching_loss": 0.04671185426414013, + "grad_norm": 0.5755724906921387, + "learning_rate": 1.5495e-05, + "loss": 0.1477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039581763371825215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022033459134399892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10119698643684387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983600556850434, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3100, + "weighted_flow_matching_loss": 0.04671185426414013, + "weighted_rkd_loss": 0.10103103145956993 + }, + { + "epoch": 0.9048588885656096, + "flow_matching_loss": 0.05024966634809971, + "grad_norm": 0.5461774468421936, + "learning_rate": 1.5545e-05, + "loss": 0.155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04090358540415764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02316421438008547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10497138425707817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983494520187378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3110, + "weighted_flow_matching_loss": 0.05024966634809971, + "weighted_rkd_loss": 0.10479813143610954 + }, + { + "epoch": 0.907768402676753, + "flow_matching_loss": 0.04726914465427399, + "grad_norm": 0.5606062412261963, + "learning_rate": 1.5595000000000002e-05, + "loss": 0.1465, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039006733894348146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021399946324527262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09941341429948806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983388006687164, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3120, + "weighted_flow_matching_loss": 0.04726914465427399, + "weighted_rkd_loss": 0.09924827367067338 + }, + { + "epoch": 0.9106779167878964, + "flow_matching_loss": 0.04918776489794254, + "grad_norm": 0.5514481067657471, + "learning_rate": 1.5645e-05, + "loss": 0.1522, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04019602090120315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02280464731156826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1031966894865036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983281314373016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3130, + "weighted_flow_matching_loss": 0.04918776489794254, + "weighted_rkd_loss": 0.10302415266633033 + }, + { + "epoch": 0.9135874308990398, + "flow_matching_loss": 0.04940441511571407, + "grad_norm": 0.4973818063735962, + "learning_rate": 1.5695e-05, + "loss": 0.1466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.038430973142385486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02051886059343815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09738080650568008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983174204826355, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3140, + "weighted_flow_matching_loss": 0.04940441511571407, + "weighted_rkd_loss": 0.09721695855259896 + }, + { + "epoch": 0.9164969450101833, + "flow_matching_loss": 0.04956630654633045, + "grad_norm": 0.5687256455421448, + "learning_rate": 1.5745000000000003e-05, + "loss": 0.1503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03920447342097759, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022499096021056174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1009080432355404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998306667804718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3150, + "weighted_flow_matching_loss": 0.04956630654633045, + "weighted_rkd_loss": 0.1007371723651886 + }, + { + "epoch": 0.9194064591213268, + "flow_matching_loss": 0.049648822471499446, + "grad_norm": 0.5037240982055664, + "learning_rate": 1.5795e-05, + "loss": 0.1505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03987342827022076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02125598695129156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10100284293293953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982958853244781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3160, + "weighted_flow_matching_loss": 0.049648822471499446, + "weighted_rkd_loss": 0.10083073601126671 + }, + { + "epoch": 0.9223159732324702, + "flow_matching_loss": 0.052635323628783225, + "grad_norm": 0.6181201338768005, + "learning_rate": 1.5845e-05, + "loss": 0.1564, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.040855918824672696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022270618565380575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10398245677351951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982850730419159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3170, + "weighted_flow_matching_loss": 0.052635323628783225, + "weighted_rkd_loss": 0.1038041166961193 + }, + { + "epoch": 0.9252254873436137, + "flow_matching_loss": 0.04906488731503487, + "grad_norm": 0.580618143081665, + "learning_rate": 1.5895000000000003e-05, + "loss": 0.1479, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03891323953866958, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021142952889204026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09896943271160126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982742190361023, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3180, + "weighted_flow_matching_loss": 0.04906488731503487, + "weighted_rkd_loss": 0.0987986333668232 + }, + { + "epoch": 0.928135001454757, + "flow_matching_loss": 0.04783103503286838, + "grad_norm": 0.5062934160232544, + "learning_rate": 1.5945e-05, + "loss": 0.1456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.038452576100826266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02102830186486244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09793345481157303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982633352279663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3190, + "weighted_flow_matching_loss": 0.04783103503286838, + "weighted_rkd_loss": 0.09776338115334511 + }, + { + "epoch": 0.9310445155659005, + "flow_matching_loss": 0.04882286004722118, + "grad_norm": 0.5346640944480896, + "learning_rate": 1.5995000000000002e-05, + "loss": 0.1497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03944065012037754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0221378643065691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10101916491985322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982524216175079, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3200, + "weighted_flow_matching_loss": 0.04882286004722118, + "weighted_rkd_loss": 0.10084262043237686 + }, + { + "epoch": 0.9339540296770439, + "flow_matching_loss": 0.0479841835796833, + "grad_norm": 0.467324823141098, + "learning_rate": 1.6045000000000003e-05, + "loss": 0.1452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03810444250702858, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021134102903306484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09734298884868622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982414662837982, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3210, + "weighted_flow_matching_loss": 0.0479841835796833, + "weighted_rkd_loss": 0.09717181324958801 + }, + { + "epoch": 0.9368635437881874, + "flow_matching_loss": 0.052227739989757535, + "grad_norm": 0.5385797023773193, + "learning_rate": 1.6095e-05, + "loss": 0.1519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0387599840760231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022289449349045753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09980941861867905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982304811477661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3220, + "weighted_flow_matching_loss": 0.052227739989757535, + "weighted_rkd_loss": 0.09963281899690628 + }, + { + "epoch": 0.9397730578993309, + "flow_matching_loss": 0.04841237217187881, + "grad_norm": 0.5279974937438965, + "learning_rate": 1.6145000000000002e-05, + "loss": 0.1475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03901519663631916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021233214437961577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0992636077105999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982194602489471, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3230, + "weighted_flow_matching_loss": 0.04841237217187881, + "weighted_rkd_loss": 0.09908687323331833 + }, + { + "epoch": 0.9426825720104742, + "flow_matching_loss": 0.047147220373153685, + "grad_norm": 0.5168206095695496, + "learning_rate": 1.6195000000000003e-05, + "loss": 0.1451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03850598856806755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021121947467327117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09813392385840417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982084035873413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3240, + "weighted_flow_matching_loss": 0.047147220373153685, + "weighted_rkd_loss": 0.09795810878276826 + }, + { + "epoch": 0.9455920861216177, + "flow_matching_loss": 0.048709763959050176, + "grad_norm": 0.5351173281669617, + "learning_rate": 1.6245e-05, + "loss": 0.1417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03678491860628128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01954884473234415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09311868250370026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981973171234131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3250, + "weighted_flow_matching_loss": 0.048709763959050176, + "weighted_rkd_loss": 0.09295082613825797 + }, + { + "epoch": 0.9485016002327611, + "flow_matching_loss": 0.046630115061998365, + "grad_norm": 0.5212579369544983, + "learning_rate": 1.6295000000000002e-05, + "loss": 0.1392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03626081831753254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02023659981787205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09275823608040809, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981861889362336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3260, + "weighted_flow_matching_loss": 0.046630115061998365, + "weighted_rkd_loss": 0.09258999899029732 + }, + { + "epoch": 0.9514111143439046, + "flow_matching_loss": 0.04973732344806194, + "grad_norm": 0.574344277381897, + "learning_rate": 1.6345e-05, + "loss": 0.1438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03710375912487507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020012914203107357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09422043263912201, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981750428676606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3270, + "weighted_flow_matching_loss": 0.04973732344806194, + "weighted_rkd_loss": 0.09404849112033845 + }, + { + "epoch": 0.9543206284550481, + "flow_matching_loss": 0.051079580560326576, + "grad_norm": 0.5668206810951233, + "learning_rate": 1.6395e-05, + "loss": 0.1475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03795112371444702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02065947651863098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09656172394752502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981638431549072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3280, + "weighted_flow_matching_loss": 0.051079580560326576, + "weighted_rkd_loss": 0.09638442918658256 + }, + { + "epoch": 0.9572301425661914, + "flow_matching_loss": 0.05097241848707199, + "grad_norm": 0.5217329859733582, + "learning_rate": 1.6445000000000003e-05, + "loss": 0.1527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03962727002799511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02266002781689167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10191456750035285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998152619600296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3290, + "weighted_flow_matching_loss": 0.05097241848707199, + "weighted_rkd_loss": 0.10172627866268158 + }, + { + "epoch": 0.9601396566773349, + "flow_matching_loss": 0.0498815231025219, + "grad_norm": 0.61789870262146, + "learning_rate": 1.6495e-05, + "loss": 0.1515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03968808650970459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02241020742803812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10178638100624085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981413662433625, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3300, + "weighted_flow_matching_loss": 0.0498815231025219, + "weighted_rkd_loss": 0.10159719213843346 + }, + { + "epoch": 0.9630491707884783, + "flow_matching_loss": 0.04579632692039013, + "grad_norm": 0.6435328125953674, + "learning_rate": 1.6545e-05, + "loss": 0.145, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03916786275804043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021058996580541132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09939472302794457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981300532817841, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3310, + "weighted_flow_matching_loss": 0.04579632692039013, + "weighted_rkd_loss": 0.09920885860919952 + }, + { + "epoch": 0.9659586848996218, + "flow_matching_loss": 0.047137299180030824, + "grad_norm": 0.5568784475326538, + "learning_rate": 1.6595000000000003e-05, + "loss": 0.1404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03659901171922684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02027126457542181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09346928820014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981187283992767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3320, + "weighted_flow_matching_loss": 0.047137299180030824, + "weighted_rkd_loss": 0.0932934433221817 + }, + { + "epoch": 0.9688681990107652, + "flow_matching_loss": 0.05090763345360756, + "grad_norm": 0.5088266730308533, + "learning_rate": 1.6645e-05, + "loss": 0.152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03981305696070194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021682454273104668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10130856856703759, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981073796749115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3330, + "weighted_flow_matching_loss": 0.05090763345360756, + "weighted_rkd_loss": 0.10111684501171112 + }, + { + "epoch": 0.9717777131219086, + "flow_matching_loss": 0.050063781440258026, + "grad_norm": 0.5261766910552979, + "learning_rate": 1.6695000000000002e-05, + "loss": 0.1484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03854995034635067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021445206180214883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09854510575532913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980959892272949, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3340, + "weighted_flow_matching_loss": 0.050063781440258026, + "weighted_rkd_loss": 0.09835746586322784 + }, + { + "epoch": 0.9746872272330521, + "flow_matching_loss": 0.051562833786010745, + "grad_norm": 0.5216694474220276, + "learning_rate": 1.6745e-05, + "loss": 0.1483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03767938800156116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021594740636646747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09695351496338844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998084557056427, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3350, + "weighted_flow_matching_loss": 0.051562833786010745, + "weighted_rkd_loss": 0.09676780998706817 + }, + { + "epoch": 0.9775967413441955, + "flow_matching_loss": 0.05240785963833332, + "grad_norm": 0.49402281641960144, + "learning_rate": 1.6794999999999998e-05, + "loss": 0.1484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03765035644173622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020924682915210723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09622539728879928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980730831623077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3360, + "weighted_flow_matching_loss": 0.05240785963833332, + "weighted_rkd_loss": 0.0960399828851223 + }, + { + "epoch": 0.980506255455339, + "flow_matching_loss": 0.049746543541550635, + "grad_norm": 0.580884575843811, + "learning_rate": 1.6845e-05, + "loss": 0.1476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.038443407788872716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02116964440792799, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09805645942687988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980615973472595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3370, + "weighted_flow_matching_loss": 0.049746543541550635, + "weighted_rkd_loss": 0.09786637723445893 + }, + { + "epoch": 0.9834157695664824, + "flow_matching_loss": 0.0491918470710516, + "grad_norm": 0.5440691113471985, + "learning_rate": 1.6895e-05, + "loss": 0.1463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03820734359323978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020893362164497376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09730804935097695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980500638484955, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3380, + "weighted_flow_matching_loss": 0.0491918470710516, + "weighted_rkd_loss": 0.09711830392479896 + }, + { + "epoch": 0.9863252836776258, + "flow_matching_loss": 0.04681455940008163, + "grad_norm": 0.5895993709564209, + "learning_rate": 1.6944999999999998e-05, + "loss": 0.1473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039529497548937795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021639873646199702, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10069886893033982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980384945869446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3390, + "weighted_flow_matching_loss": 0.04681455940008163, + "weighted_rkd_loss": 0.10050134807825088 + }, + { + "epoch": 0.9892347977887693, + "flow_matching_loss": 0.04898383468389511, + "grad_norm": 0.5569920539855957, + "learning_rate": 1.6995e-05, + "loss": 0.1438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.037326335161924365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020320919808000326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09497358947992325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980268955230713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3400, + "weighted_flow_matching_loss": 0.04898383468389511, + "weighted_rkd_loss": 0.09478620439767838 + }, + { + "epoch": 0.9921443118999127, + "flow_matching_loss": 0.05060780495405197, + "grad_norm": 0.5744966268539429, + "learning_rate": 1.7045e-05, + "loss": 0.1516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03922816887497902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022739792801439764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10119613111019135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980152547359467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3410, + "weighted_flow_matching_loss": 0.05060780495405197, + "weighted_rkd_loss": 0.10099527165293694 + }, + { + "epoch": 0.9950538260110562, + "flow_matching_loss": 0.051025420799851416, + "grad_norm": 0.5384489297866821, + "learning_rate": 1.7094999999999998e-05, + "loss": 0.1476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03778270781040192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02117924727499485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09674466252326966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980035901069642, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3420, + "weighted_flow_matching_loss": 0.051025420799851416, + "weighted_rkd_loss": 0.09655151516199112 + }, + { + "epoch": 0.9979633401221996, + "flow_matching_loss": 0.04895262196660042, + "grad_norm": 0.6332145929336548, + "learning_rate": 1.7145e-05, + "loss": 0.1494, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03918495625257492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02232430689036846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10069421902298928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979918897151947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3430, + "weighted_flow_matching_loss": 0.04895262196660042, + "weighted_rkd_loss": 0.10049201548099518 + }, + { + "epoch": 1.000872854233343, + "flow_matching_loss": 0.04635028690099716, + "grad_norm": 0.6135357022285461, + "learning_rate": 1.7195e-05, + "loss": 0.144, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03850857987999916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020840805768966675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09785796627402306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997980147600174, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3440, + "weighted_flow_matching_loss": 0.04635028690099716, + "weighted_rkd_loss": 0.09766031056642532 + }, + { + "epoch": 1.0037823683444864, + "flow_matching_loss": 0.048681368306279185, + "grad_norm": 0.5198493003845215, + "learning_rate": 1.7245e-05, + "loss": 0.1462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.038204680755734445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021299827843904495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09770918861031533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979683756828308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3450, + "weighted_flow_matching_loss": 0.048681368306279185, + "weighted_rkd_loss": 0.0975106731057167 + }, + { + "epoch": 1.00669188245563, + "flow_matching_loss": 0.04650473482906818, + "grad_norm": 0.5461238026618958, + "learning_rate": 1.7295e-05, + "loss": 0.1432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.037917041406035426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021015508845448495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09684959053993225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979565739631653, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3460, + "weighted_flow_matching_loss": 0.04650473482906818, + "weighted_rkd_loss": 0.09665169194340706 + }, + { + "epoch": 1.0096013965667734, + "flow_matching_loss": 0.047082680836319925, + "grad_norm": 0.5413183569908142, + "learning_rate": 1.7345e-05, + "loss": 0.1476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03943704329431057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021867111697793006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10074119791388511, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979447364807129, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3470, + "weighted_flow_matching_loss": 0.047082680836319925, + "weighted_rkd_loss": 0.10053414553403854 + }, + { + "epoch": 1.0125109106779169, + "flow_matching_loss": 0.04683869145810604, + "grad_norm": 0.5256470441818237, + "learning_rate": 1.7395e-05, + "loss": 0.142, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03761216625571251, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020162288285791873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09538662135601043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979328513145447, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3480, + "weighted_flow_matching_loss": 0.04683869145810604, + "weighted_rkd_loss": 0.0951894536614418 + }, + { + "epoch": 1.0154204247890601, + "flow_matching_loss": 0.04561420418322086, + "grad_norm": 0.5749679207801819, + "learning_rate": 1.7445e-05, + "loss": 0.1384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03656471110880375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019881667755544185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09301108941435814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979209303855896, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3490, + "weighted_flow_matching_loss": 0.04561420418322086, + "weighted_rkd_loss": 0.09281772151589393 + }, + { + "epoch": 1.0183299389002036, + "flow_matching_loss": 0.04857480637729168, + "grad_norm": 0.5218759179115295, + "learning_rate": 1.7495e-05, + "loss": 0.1488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03893794640898705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022568684257566928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10044457837939262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979090094566345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3500, + "weighted_flow_matching_loss": 0.04857480637729168, + "weighted_rkd_loss": 0.10023457035422326 + }, + { + "epoch": 1.021239453011347, + "flow_matching_loss": 0.046617753058671954, + "grad_norm": 0.592261552810669, + "learning_rate": 1.7545e-05, + "loss": 0.1438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03855832554399967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020227570086717606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09734422117471694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978970348834991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3510, + "weighted_flow_matching_loss": 0.046617753058671954, + "weighted_rkd_loss": 0.09713949784636497 + }, + { + "epoch": 1.0241489671224906, + "flow_matching_loss": 0.04782446064054966, + "grad_norm": 0.5218790769577026, + "learning_rate": 1.7595e-05, + "loss": 0.1432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03750624023377895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020578769594430925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09559125006198883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978850245475769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3520, + "weighted_flow_matching_loss": 0.04782446064054966, + "weighted_rkd_loss": 0.09538906067609787 + }, + { + "epoch": 1.027058481233634, + "flow_matching_loss": 0.05038895942270756, + "grad_norm": 0.5853084921836853, + "learning_rate": 1.7644999999999998e-05, + "loss": 0.1448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.037011326476931575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02054757662117481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09457023069262505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978729784488678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3530, + "weighted_flow_matching_loss": 0.05038895942270756, + "weighted_rkd_loss": 0.09436907172203064 + }, + { + "epoch": 1.0299679953447773, + "flow_matching_loss": 0.0455540869385004, + "grad_norm": 0.5311583280563354, + "learning_rate": 1.7695e-05, + "loss": 0.1424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03794744051992893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021145936287939548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09704081863164901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978609025478363, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3540, + "weighted_flow_matching_loss": 0.0455540869385004, + "weighted_rkd_loss": 0.0968332476913929 + }, + { + "epoch": 1.0328775094559208, + "flow_matching_loss": 0.04797373376786709, + "grad_norm": 0.5668309926986694, + "learning_rate": 1.7745e-05, + "loss": 0.1403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03630850538611412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019883918575942516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09250092953443527, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978487908840179, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3550, + "weighted_flow_matching_loss": 0.04797373376786709, + "weighted_rkd_loss": 0.09230194017291068 + }, + { + "epoch": 1.0357870235670643, + "flow_matching_loss": 0.04802851341664791, + "grad_norm": 0.6214016079902649, + "learning_rate": 1.7794999999999998e-05, + "loss": 0.148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03924030251801014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02170492596924305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10018553212285042, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978366434574127, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3560, + "weighted_flow_matching_loss": 0.04802851341664791, + "weighted_rkd_loss": 0.0999687857925892 + }, + { + "epoch": 1.0386965376782078, + "flow_matching_loss": 0.045998860523104666, + "grad_norm": 0.5309196710586548, + "learning_rate": 1.7845e-05, + "loss": 0.1415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03745119646191597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02081234771758318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0957147404551506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978244602680206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3570, + "weighted_flow_matching_loss": 0.045998860523104666, + "weighted_rkd_loss": 0.09550652503967286 + }, + { + "epoch": 1.0416060517893513, + "flow_matching_loss": 0.046863388270139694, + "grad_norm": 0.5831785798072815, + "learning_rate": 1.7895e-05, + "loss": 0.1392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03633980602025986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019816631078720094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0924962431192398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978122413158417, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3580, + "weighted_flow_matching_loss": 0.046863388270139694, + "weighted_rkd_loss": 0.09229387417435646 + }, + { + "epoch": 1.0445155659004945, + "flow_matching_loss": 0.046553775668144226, + "grad_norm": 0.49537816643714905, + "learning_rate": 1.7945e-05, + "loss": 0.139, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03636344000697136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01992076151072979, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09264764115214348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977999985218048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3590, + "weighted_flow_matching_loss": 0.046553775668144226, + "weighted_rkd_loss": 0.0924438290297985 + }, + { + "epoch": 1.047425080011638, + "flow_matching_loss": 0.04788044057786465, + "grad_norm": 0.5689855813980103, + "learning_rate": 1.7995e-05, + "loss": 0.14, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036016037315130235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020265855453908445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09229793027043343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977877140045166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3600, + "weighted_flow_matching_loss": 0.04788044057786465, + "weighted_rkd_loss": 0.09209374114871025 + }, + { + "epoch": 1.0503345941227815, + "flow_matching_loss": 0.047172756120562556, + "grad_norm": 0.5386833548545837, + "learning_rate": 1.8045e-05, + "loss": 0.1448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03819792792201042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021445644460618495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09784149900078773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977753937244416, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3610, + "weighted_flow_matching_loss": 0.047172756120562556, + "weighted_rkd_loss": 0.09762382358312607 + }, + { + "epoch": 1.053244108233925, + "flow_matching_loss": 0.047509682923555376, + "grad_norm": 0.584926426410675, + "learning_rate": 1.8095e-05, + "loss": 0.1415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03676427230238914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020708163268864155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09423670768737794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997763043642044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3620, + "weighted_flow_matching_loss": 0.047509682923555376, + "weighted_rkd_loss": 0.09402590915560723 + }, + { + "epoch": 1.0561536223450685, + "flow_matching_loss": 0.048878346011042594, + "grad_norm": 0.5383336544036865, + "learning_rate": 1.8145e-05, + "loss": 0.1422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036851595342159274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01982964426279068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09353283494710922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977506577968598, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3630, + "weighted_flow_matching_loss": 0.048878346011042594, + "weighted_rkd_loss": 0.09332245588302612 + }, + { + "epoch": 1.0590631364562118, + "flow_matching_loss": 0.0469945851713419, + "grad_norm": 0.5971601605415344, + "learning_rate": 1.8195e-05, + "loss": 0.1358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03515190687030554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018718286976218225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08902210220694542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997738242149353, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3640, + "weighted_flow_matching_loss": 0.0469945851713419, + "weighted_rkd_loss": 0.08882074430584908 + }, + { + "epoch": 1.0619726505673552, + "flow_matching_loss": 0.05051624178886414, + "grad_norm": 0.47399288415908813, + "learning_rate": 1.8245e-05, + "loss": 0.1408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03534868899732828, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019775092229247095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09047247022390366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977257907390594, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3650, + "weighted_flow_matching_loss": 0.05051624178886414, + "weighted_rkd_loss": 0.09026672765612602 + }, + { + "epoch": 1.0648821646784987, + "flow_matching_loss": 0.04763525202870369, + "grad_norm": 0.5309594869613647, + "learning_rate": 1.8295e-05, + "loss": 0.1379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03554010167717934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01939570214599371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09047590643167495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977132976055145, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3660, + "weighted_flow_matching_loss": 0.04763525202870369, + "weighted_rkd_loss": 0.09026902914047241 + }, + { + "epoch": 1.0677916787896422, + "flow_matching_loss": 0.04726281352341175, + "grad_norm": 0.6670515537261963, + "learning_rate": 1.8345e-05, + "loss": 0.1453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03837408870458603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021529753878712653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09827793166041374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977007806301117, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3670, + "weighted_flow_matching_loss": 0.04726281352341175, + "weighted_rkd_loss": 0.09805194810032844 + }, + { + "epoch": 1.0707011929007855, + "flow_matching_loss": 0.04842272624373436, + "grad_norm": 0.589271068572998, + "learning_rate": 1.8395e-05, + "loss": 0.1413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03654457628726959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019983070716261864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09307222217321395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976882219314576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3680, + "weighted_flow_matching_loss": 0.04842272624373436, + "weighted_rkd_loss": 0.09285705611109733 + }, + { + "epoch": 1.073610707011929, + "flow_matching_loss": 0.047484827041625974, + "grad_norm": 0.5190726518630981, + "learning_rate": 1.8445e-05, + "loss": 0.1377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03559718206524849, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01918454738333821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09037891179323196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976756334304809, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3690, + "weighted_flow_matching_loss": 0.047484827041625974, + "weighted_rkd_loss": 0.09016884416341782 + }, + { + "epoch": 1.0765202211230724, + "flow_matching_loss": 0.04862516708672047, + "grad_norm": 0.6141235828399658, + "learning_rate": 1.8495e-05, + "loss": 0.1418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03648029454052448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020476182363927363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09343677014112473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976630091667176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3700, + "weighted_flow_matching_loss": 0.04862516708672047, + "weighted_rkd_loss": 0.09321839064359665 + }, + { + "epoch": 1.079429735234216, + "flow_matching_loss": 0.043015826120972635, + "grad_norm": 0.5100975036621094, + "learning_rate": 1.8545e-05, + "loss": 0.1332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0356826014816761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018981745839118956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09034694880247116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976503551006317, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3710, + "weighted_flow_matching_loss": 0.043015826120972635, + "weighted_rkd_loss": 0.09013466835021973 + }, + { + "epoch": 1.0823392493453594, + "flow_matching_loss": 0.04383485205471516, + "grad_norm": 0.5592804551124573, + "learning_rate": 1.8595e-05, + "loss": 0.1383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.037003351002931596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02071569338440895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0947223961353302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976376473903656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3720, + "weighted_flow_matching_loss": 0.04383485205471516, + "weighted_rkd_loss": 0.0944986216723919 + }, + { + "epoch": 1.0852487634565027, + "flow_matching_loss": 0.04698893018066883, + "grad_norm": 0.5696859359741211, + "learning_rate": 1.8645000000000002e-05, + "loss": 0.1397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036507651954889295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019898262061178684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09291356578469276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976249277591706, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3730, + "weighted_flow_matching_loss": 0.04698893018066883, + "weighted_rkd_loss": 0.09269289895892144 + }, + { + "epoch": 1.0881582775676462, + "flow_matching_loss": 0.044391053915023806, + "grad_norm": 0.5400590896606445, + "learning_rate": 1.8695e-05, + "loss": 0.1349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0354529544711113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01978205181658268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09068796038627625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976121664047242, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3740, + "weighted_flow_matching_loss": 0.044391053915023806, + "weighted_rkd_loss": 0.09047139883041382 + }, + { + "epoch": 1.0910677916787896, + "flow_matching_loss": 0.049801741540431974, + "grad_norm": 0.49584001302719116, + "learning_rate": 1.8745e-05, + "loss": 0.1401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0353713259100914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01976099982857704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09050365164875984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975993633270264, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3750, + "weighted_flow_matching_loss": 0.049801741540431974, + "weighted_rkd_loss": 0.0902864046394825 + }, + { + "epoch": 1.0939773057899331, + "flow_matching_loss": 0.046001844480633734, + "grad_norm": 0.5113233923912048, + "learning_rate": 1.8795e-05, + "loss": 0.1397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036537958309054375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02082803100347519, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09390394687652588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975865423679352, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3760, + "weighted_flow_matching_loss": 0.046001844480633734, + "weighted_rkd_loss": 0.0936773143708706 + }, + { + "epoch": 1.0968868199010766, + "flow_matching_loss": 0.047763596102595327, + "grad_norm": 0.5411504507064819, + "learning_rate": 1.8845e-05, + "loss": 0.1392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035936289280653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01982693113386631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09169950932264329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975736796855926, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3770, + "weighted_flow_matching_loss": 0.047763596102595327, + "weighted_rkd_loss": 0.09147702753543854 + }, + { + "epoch": 1.0997963340122199, + "flow_matching_loss": 0.04747206792235374, + "grad_norm": 0.5041340589523315, + "learning_rate": 1.8895e-05, + "loss": 0.1387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03596528135240078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019509437680244445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.091440000385046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975607872009278, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3780, + "weighted_flow_matching_loss": 0.04747206792235374, + "weighted_rkd_loss": 0.09121696725487709 + }, + { + "epoch": 1.1027058481233634, + "flow_matching_loss": 0.047649850696325304, + "grad_norm": 0.48188531398773193, + "learning_rate": 1.8945e-05, + "loss": 0.1379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03559200614690781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01929164556786418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09047565758228301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997547858953476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3790, + "weighted_flow_matching_loss": 0.047649850696325304, + "weighted_rkd_loss": 0.0902538150548935 + }, + { + "epoch": 1.1056153622345068, + "flow_matching_loss": 0.04890488646924496, + "grad_norm": 0.4702511131763458, + "learning_rate": 1.8995e-05, + "loss": 0.1415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036646289937198165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01957522388547659, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09286780431866645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975348889827729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3800, + "weighted_flow_matching_loss": 0.04890488646924496, + "weighted_rkd_loss": 0.09263888373970985 + }, + { + "epoch": 1.1085248763456503, + "flow_matching_loss": 0.04269614890217781, + "grad_norm": 0.4449506998062134, + "learning_rate": 1.9045e-05, + "loss": 0.135, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03648071866482496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01959128975868225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09255272746086121, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975218892097473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3810, + "weighted_flow_matching_loss": 0.04269614890217781, + "weighted_rkd_loss": 0.09232338219881057 + }, + { + "epoch": 1.1114343904567938, + "flow_matching_loss": 0.04709891602396965, + "grad_norm": 0.4699045419692993, + "learning_rate": 1.9095e-05, + "loss": 0.1375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035909492149949075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018807675689458847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.090626659989357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975088536739349, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3820, + "weighted_flow_matching_loss": 0.04709891602396965, + "weighted_rkd_loss": 0.09040089920163155 + }, + { + "epoch": 1.114343904567937, + "flow_matching_loss": 0.04517010301351547, + "grad_norm": 0.5942275524139404, + "learning_rate": 1.9145e-05, + "loss": 0.1355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035425349697470666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019699859991669656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09055055901408196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974957704544067, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3830, + "weighted_flow_matching_loss": 0.04517010301351547, + "weighted_rkd_loss": 0.09032381027936935 + }, + { + "epoch": 1.1172534186790806, + "flow_matching_loss": 0.04429149366915226, + "grad_norm": 0.5141463875770569, + "learning_rate": 1.9195000000000002e-05, + "loss": 0.1288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0335472509264946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01761364471167326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08470814675092697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974826693534851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3840, + "weighted_flow_matching_loss": 0.04429149366915226, + "weighted_rkd_loss": 0.08449491187930107 + }, + { + "epoch": 1.120162932790224, + "flow_matching_loss": 0.04778568185865879, + "grad_norm": 0.5133700370788574, + "learning_rate": 1.9245e-05, + "loss": 0.1348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03441853020340204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018410392664372922, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0872474528849125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974695444107056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3850, + "weighted_flow_matching_loss": 0.04778568185865879, + "weighted_rkd_loss": 0.08702667355537415 + }, + { + "epoch": 1.1230724469013675, + "flow_matching_loss": 0.045320615917444226, + "grad_norm": 0.540767252445221, + "learning_rate": 1.9295e-05, + "loss": 0.1348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03500394560396671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019734440371394157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08974233120679856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974563837051391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3860, + "weighted_flow_matching_loss": 0.045320615917444226, + "weighted_rkd_loss": 0.08951406553387642 + }, + { + "epoch": 1.1259819610125108, + "flow_matching_loss": 0.0443276971578598, + "grad_norm": 0.47711730003356934, + "learning_rate": 1.9345000000000002e-05, + "loss": 0.1312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03418560307472944, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01869363524019718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08706484213471413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974431753158569, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3870, + "weighted_flow_matching_loss": 0.0443276971578598, + "weighted_rkd_loss": 0.08684225007891655 + }, + { + "epoch": 1.1288914751236543, + "flow_matching_loss": 0.04547134712338448, + "grad_norm": 0.49943771958351135, + "learning_rate": 1.9395e-05, + "loss": 0.1348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03515714965760708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019277076050639153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08959137499332429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974299371242523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3880, + "weighted_flow_matching_loss": 0.04547134712338448, + "weighted_rkd_loss": 0.08936112001538277 + }, + { + "epoch": 1.1318009892347978, + "flow_matching_loss": 0.04735165387392044, + "grad_norm": 0.5890047550201416, + "learning_rate": 1.9445e-05, + "loss": 0.1406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0367443535476923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019990790449082852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09347949773073197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974166631698609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3890, + "weighted_flow_matching_loss": 0.04735165387392044, + "weighted_rkd_loss": 0.0932380236685276 + }, + { + "epoch": 1.1347105033459413, + "flow_matching_loss": 0.044112324342131616, + "grad_norm": 0.5424422025680542, + "learning_rate": 1.9495000000000002e-05, + "loss": 0.1303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0338917002081871, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018639175221323967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08642257526516914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974033534526825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3900, + "weighted_flow_matching_loss": 0.044112324342131616, + "weighted_rkd_loss": 0.08619817793369293 + }, + { + "epoch": 1.1376200174570847, + "flow_matching_loss": 0.04721888042986393, + "grad_norm": 0.5274159908294678, + "learning_rate": 1.9545e-05, + "loss": 0.1349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03451374433934688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018887140229344367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08791462928056717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973900198936463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3910, + "weighted_flow_matching_loss": 0.04721888042986393, + "weighted_rkd_loss": 0.08768518492579461 + }, + { + "epoch": 1.140529531568228, + "flow_matching_loss": 0.047169451415538785, + "grad_norm": 0.5117771625518799, + "learning_rate": 1.9595e-05, + "loss": 0.1368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03497645296156406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019912156648933886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0898650623857975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973766446113587, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3920, + "weighted_flow_matching_loss": 0.047169451415538785, + "weighted_rkd_loss": 0.08962932825088502 + }, + { + "epoch": 1.1434390456793715, + "flow_matching_loss": 0.0453702412545681, + "grad_norm": 0.5541881918907166, + "learning_rate": 1.9645000000000002e-05, + "loss": 0.1354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03571005593985319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01886914363130927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09028925597667695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973632395267487, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3930, + "weighted_flow_matching_loss": 0.0453702412545681, + "weighted_rkd_loss": 0.09005118682980537 + }, + { + "epoch": 1.146348559790515, + "flow_matching_loss": 0.0476836483925581, + "grad_norm": 0.5959421992301941, + "learning_rate": 1.9695e-05, + "loss": 0.1367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035389356501400474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018463181797415018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08924189433455468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973497867584229, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3940, + "weighted_flow_matching_loss": 0.0476836483925581, + "weighted_rkd_loss": 0.08900535702705384 + }, + { + "epoch": 1.1492580739016585, + "flow_matching_loss": 0.0491762213408947, + "grad_norm": 0.49935659766197205, + "learning_rate": 1.9745e-05, + "loss": 0.1349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034077657572925094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01781783252954483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08597314804792404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973363220691681, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3950, + "weighted_flow_matching_loss": 0.0491762213408947, + "weighted_rkd_loss": 0.08574413582682609 + }, + { + "epoch": 1.152167588012802, + "flow_matching_loss": 0.04614810459315777, + "grad_norm": 0.5538299083709717, + "learning_rate": 1.9795000000000003e-05, + "loss": 0.1322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03384405486285687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01857427442446351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08626238331198692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997322803735733, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3960, + "weighted_flow_matching_loss": 0.04614810459315777, + "weighted_rkd_loss": 0.08603144064545631 + }, + { + "epoch": 1.1550771021239452, + "flow_matching_loss": 0.044516392052173615, + "grad_norm": 0.5291272401809692, + "learning_rate": 1.9845e-05, + "loss": 0.131, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03401238154619932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018727291282266377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08675205409526825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973092615604401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3970, + "weighted_flow_matching_loss": 0.044516392052173615, + "weighted_rkd_loss": 0.08651860877871513 + }, + { + "epoch": 1.1579866162350887, + "flow_matching_loss": 0.04540497325360775, + "grad_norm": 0.6106089353561401, + "learning_rate": 1.9895000000000002e-05, + "loss": 0.1381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03674643374979496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0194937976077199, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0929866649210453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9972956776618958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3980, + "weighted_flow_matching_loss": 0.04540497325360775, + "weighted_rkd_loss": 0.09273517951369285 + }, + { + "epoch": 1.1608961303462322, + "flow_matching_loss": 0.04574172794818878, + "grad_norm": 0.6354818344116211, + "learning_rate": 1.9945000000000003e-05, + "loss": 0.1352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035411165468394755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01886875070631504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08969108164310455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997282075881958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3990, + "weighted_flow_matching_loss": 0.04574172794818878, + "weighted_rkd_loss": 0.08944730684161187 + }, + { + "epoch": 1.1638056444573757, + "flow_matching_loss": 0.04606879837810993, + "grad_norm": 0.49878135323524475, + "learning_rate": 1.9995e-05, + "loss": 0.1337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03450005110353231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01886850707232952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08786860853433609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9972684144973755, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4000, + "weighted_flow_matching_loss": 0.04606879837810993, + "weighted_rkd_loss": 0.08762860000133514 + }, + { + "epoch": 1.1667151585685192, + "flow_matching_loss": 0.046422158181667325, + "grad_norm": 0.5227116942405701, + "learning_rate": 2.0045000000000002e-05, + "loss": 0.1333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03440186735242605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018303010798990726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0871067464351654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9972547590732574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4010, + "weighted_flow_matching_loss": 0.046422158181667325, + "weighted_rkd_loss": 0.08686762377619743 + }, + { + "epoch": 1.1696246726796624, + "flow_matching_loss": 0.04577297233045101, + "grad_norm": 0.5584054589271545, + "learning_rate": 2.0095e-05, + "loss": 0.1324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03397071920335293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01891871988773346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08686015829443931, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9972410321235656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4020, + "weighted_flow_matching_loss": 0.04577297233045101, + "weighted_rkd_loss": 0.0866205096244812 + }, + { + "epoch": 1.172534186790806, + "flow_matching_loss": 0.04652638658881188, + "grad_norm": 0.6073059439659119, + "learning_rate": 2.0145e-05, + "loss": 0.1332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03416723217815161, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01861414946615696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08694861307740212, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9972272753715515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4030, + "weighted_flow_matching_loss": 0.04652638658881188, + "weighted_rkd_loss": 0.08670751824975013 + }, + { + "epoch": 1.1754437009019494, + "flow_matching_loss": 0.0455147311091423, + "grad_norm": 0.5859513282775879, + "learning_rate": 2.0195000000000002e-05, + "loss": 0.1274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0323388684540987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01745364498347044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08213138282299041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9972134888172149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4040, + "weighted_flow_matching_loss": 0.0455147311091423, + "weighted_rkd_loss": 0.08190253525972366 + }, + { + "epoch": 1.1783532150130929, + "flow_matching_loss": 0.045943711325526236, + "grad_norm": 0.5226259231567383, + "learning_rate": 2.0245e-05, + "loss": 0.1386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03658322654664516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019741949066519736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09290840253233909, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971996784210205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4050, + "weighted_flow_matching_loss": 0.045943711325526236, + "weighted_rkd_loss": 0.09264824092388153 + }, + { + "epoch": 1.1812627291242364, + "flow_matching_loss": 0.04842827580869198, + "grad_norm": 0.4880910813808441, + "learning_rate": 2.0295e-05, + "loss": 0.1352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034344461746513844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018290190119296314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08697911351919174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971858203411103, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4060, + "weighted_flow_matching_loss": 0.04842827580869198, + "weighted_rkd_loss": 0.08673435226082801 + }, + { + "epoch": 1.1841722432353796, + "flow_matching_loss": 0.04447956047952175, + "grad_norm": 0.511455774307251, + "learning_rate": 2.0345000000000003e-05, + "loss": 0.1314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03443551566451788, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01834205398336053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08721308559179305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997171938419342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4070, + "weighted_flow_matching_loss": 0.04447956047952175, + "weighted_rkd_loss": 0.08696644082665443 + }, + { + "epoch": 1.187081757346523, + "flow_matching_loss": 0.04279467351734638, + "grad_norm": 0.5368521213531494, + "learning_rate": 2.0395e-05, + "loss": 0.1275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0334963096305728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01793208848685026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08492470905184746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997158020734787, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4080, + "weighted_flow_matching_loss": 0.04279467351734638, + "weighted_rkd_loss": 0.08468335047364235 + }, + { + "epoch": 1.1899912714576666, + "flow_matching_loss": 0.0462889302521944, + "grad_norm": 0.524768054485321, + "learning_rate": 2.0445e-05, + "loss": 0.1338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03433028254657984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019069364108145238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08772992938756943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971440672874451, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4090, + "weighted_flow_matching_loss": 0.0462889302521944, + "weighted_rkd_loss": 0.08747938349843025 + }, + { + "epoch": 1.19290078556881, + "flow_matching_loss": 0.04453776441514492, + "grad_norm": 0.5693824887275696, + "learning_rate": 2.0495000000000003e-05, + "loss": 0.1294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03343633655458689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018253167252987625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08512584120035172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971300661563873, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4100, + "weighted_flow_matching_loss": 0.04453776441514492, + "weighted_rkd_loss": 0.08488154336810112 + }, + { + "epoch": 1.1958102996799536, + "flow_matching_loss": 0.04673537611961365, + "grad_norm": 0.5785642862319946, + "learning_rate": 2.0545e-05, + "loss": 0.1319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03372513931244612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01799989612773061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08545017391443252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971160531044007, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4110, + "weighted_flow_matching_loss": 0.04673537611961365, + "weighted_rkd_loss": 0.08520374521613121 + }, + { + "epoch": 1.1987198137910968, + "flow_matching_loss": 0.04475206919014454, + "grad_norm": 0.5887666940689087, + "learning_rate": 2.0595000000000002e-05, + "loss": 0.1319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03424893245100975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018895838037133218, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08739370331168175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971019983291626, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4120, + "weighted_flow_matching_loss": 0.04475206919014454, + "weighted_rkd_loss": 0.08714044019579888 + }, + { + "epoch": 1.2016293279022403, + "flow_matching_loss": 0.047510042786598206, + "grad_norm": 0.5780134201049805, + "learning_rate": 2.0645000000000003e-05, + "loss": 0.1368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03511148206889629, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019312041718512772, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08953500613570213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970879077911377, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4130, + "weighted_flow_matching_loss": 0.047510042786598206, + "weighted_rkd_loss": 0.08927427902817726 + }, + { + "epoch": 1.2045388420133838, + "flow_matching_loss": 0.043177996203303336, + "grad_norm": 0.5451218485832214, + "learning_rate": 2.0695e-05, + "loss": 0.1299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03406156077980995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018805553298443557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08692867308855057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970737814903259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4140, + "weighted_flow_matching_loss": 0.043177996203303336, + "weighted_rkd_loss": 0.08667428568005561 + }, + { + "epoch": 1.2074483561245273, + "flow_matching_loss": 0.04422332793474197, + "grad_norm": 0.5602923631668091, + "learning_rate": 2.0745000000000002e-05, + "loss": 0.1307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03418207056820392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0183753109537065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08673945218324661, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970596253871917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4150, + "weighted_flow_matching_loss": 0.04422332793474197, + "weighted_rkd_loss": 0.08648440539836884 + }, + { + "epoch": 1.2103578702356708, + "flow_matching_loss": 0.04532556906342507, + "grad_norm": 0.5940003395080566, + "learning_rate": 2.0795000000000003e-05, + "loss": 0.1291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03300146348774433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017985593527555466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08398852050304413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970454335212707, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4160, + "weighted_flow_matching_loss": 0.04532556906342507, + "weighted_rkd_loss": 0.08374036699533463 + }, + { + "epoch": 1.213267384346814, + "flow_matching_loss": 0.044788339734077455, + "grad_norm": 0.47334611415863037, + "learning_rate": 2.0845e-05, + "loss": 0.1262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03226299230009318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01714313803240657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08166912198066711, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970312058925629, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4170, + "weighted_flow_matching_loss": 0.044788339734077455, + "weighted_rkd_loss": 0.08142666555941105 + }, + { + "epoch": 1.2161768984579575, + "flow_matching_loss": 0.04464582279324532, + "grad_norm": 0.5550275444984436, + "learning_rate": 2.0895000000000002e-05, + "loss": 0.1327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03469216376543045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018939447402954102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08832377567887306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970169544219971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4180, + "weighted_flow_matching_loss": 0.04464582279324532, + "weighted_rkd_loss": 0.0880602940917015 + }, + { + "epoch": 1.219086412569101, + "flow_matching_loss": 0.04746285080909729, + "grad_norm": 0.5568026900291443, + "learning_rate": 2.0945000000000004e-05, + "loss": 0.1305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03285074774175882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017630233988165857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08333172872662545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970026612281799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4190, + "weighted_flow_matching_loss": 0.04746285080909729, + "weighted_rkd_loss": 0.08308194577693939 + }, + { + "epoch": 1.2219959266802445, + "flow_matching_loss": 0.04514145664870739, + "grad_norm": 0.5149775743484497, + "learning_rate": 2.0995e-05, + "loss": 0.1325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03435428850352764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018924994207918645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08763357028365135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996988320350647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4200, + "weighted_flow_matching_loss": 0.04514145664870739, + "weighted_rkd_loss": 0.08736964389681816 + }, + { + "epoch": 1.2249054407913877, + "flow_matching_loss": 0.04462982602417469, + "grad_norm": 0.5299723148345947, + "learning_rate": 2.1045e-05, + "loss": 0.1304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03366926796734333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018715318012982606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08605385422706605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969739615917206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4210, + "weighted_flow_matching_loss": 0.04462982602417469, + "weighted_rkd_loss": 0.08579346612095833 + }, + { + "epoch": 1.2278149549025312, + "flow_matching_loss": 0.04671806022524834, + "grad_norm": 0.5045397877693176, + "learning_rate": 2.1095e-05, + "loss": 0.1264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03169089388102293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016529479529708624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07991126775741578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969595611095429, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4220, + "weighted_flow_matching_loss": 0.04671806022524834, + "weighted_rkd_loss": 0.07966830134391785 + }, + { + "epoch": 1.2307244690136747, + "flow_matching_loss": 0.04284069538116455, + "grad_norm": 0.4970933496952057, + "learning_rate": 2.1144999999999998e-05, + "loss": 0.1233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03173822090029717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017216861620545386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08069330230355262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969451308250428, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4230, + "weighted_flow_matching_loss": 0.04284069538116455, + "weighted_rkd_loss": 0.08044680058956147 + }, + { + "epoch": 1.2336339831248182, + "flow_matching_loss": 0.04504731670022011, + "grad_norm": 0.5473188757896423, + "learning_rate": 2.1195e-05, + "loss": 0.1272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03254187498241663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017329671420156956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0824134223163128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969306647777557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4240, + "weighted_flow_matching_loss": 0.04504731670022011, + "weighted_rkd_loss": 0.08216047510504723 + }, + { + "epoch": 1.2365434972359617, + "flow_matching_loss": 0.04640094116330147, + "grad_norm": 0.47598326206207275, + "learning_rate": 2.1245e-05, + "loss": 0.1313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.033627843111753465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017946867272257805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08520255386829376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969161629676819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4250, + "weighted_flow_matching_loss": 0.04640094116330147, + "weighted_rkd_loss": 0.08493981063365937 + }, + { + "epoch": 1.239453011347105, + "flow_matching_loss": 0.04315512590110302, + "grad_norm": 0.5190625190734863, + "learning_rate": 2.1295e-05, + "loss": 0.1245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032062133960425857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01751935128122568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08164361864328384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969016313552856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4260, + "weighted_flow_matching_loss": 0.04315512590110302, + "weighted_rkd_loss": 0.08139063864946365 + }, + { + "epoch": 1.2423625254582484, + "flow_matching_loss": 0.0456058531999588, + "grad_norm": 0.5081694722175598, + "learning_rate": 2.1345e-05, + "loss": 0.1211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02978556100279093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016208013519644737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07577913478016854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9968870639801025, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4270, + "weighted_flow_matching_loss": 0.0456058531999588, + "weighted_rkd_loss": 0.07554324641823769 + }, + { + "epoch": 1.245272039569392, + "flow_matching_loss": 0.04442383199930191, + "grad_norm": 0.4875968098640442, + "learning_rate": 2.1394999999999998e-05, + "loss": 0.1277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03287611231207847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017799000442028045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.083551225066185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9968724608421325, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4280, + "weighted_flow_matching_loss": 0.04442383199930191, + "weighted_rkd_loss": 0.08328990936279297 + }, + { + "epoch": 1.2481815536805354, + "flow_matching_loss": 0.04580344744026661, + "grad_norm": 0.4681280851364136, + "learning_rate": 2.1445e-05, + "loss": 0.1267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031965444423258306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017259451281279325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08119034096598625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9968578279018402, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4290, + "weighted_flow_matching_loss": 0.04580344744026661, + "weighted_rkd_loss": 0.08093524128198623 + }, + { + "epoch": 1.2510910677916787, + "flow_matching_loss": 0.04442611820995808, + "grad_norm": 0.549009382724762, + "learning_rate": 2.1495e-05, + "loss": 0.1257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032134470716118815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017272565886378287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08154150769114495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996843159198761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4300, + "weighted_flow_matching_loss": 0.04442611820995808, + "weighted_rkd_loss": 0.08128410205245018 + }, + { + "epoch": 1.2540005819028222, + "flow_matching_loss": 0.0430817823857069, + "grad_norm": 0.48123425245285034, + "learning_rate": 2.1544999999999998e-05, + "loss": 0.1246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03252163678407669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016772860381752252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0818161353468895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9968284547328949, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4310, + "weighted_flow_matching_loss": 0.0430817823857069, + "weighted_rkd_loss": 0.0815566562116146 + }, + { + "epoch": 1.2569100960139656, + "flow_matching_loss": 0.04156135693192482, + "grad_norm": 0.5506204962730408, + "learning_rate": 2.1595e-05, + "loss": 0.1222, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032078872062265874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016708337422460317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08086608126759529, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9968137204647064, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4320, + "weighted_flow_matching_loss": 0.04156135693192482, + "weighted_rkd_loss": 0.0806084156036377 + }, + { + "epoch": 1.2598196101251091, + "flow_matching_loss": 0.04462959840893745, + "grad_norm": 0.4797137379646301, + "learning_rate": 2.1645e-05, + "loss": 0.1231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031169520877301692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01637398637831211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07871302738785743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967989504337311, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4330, + "weighted_flow_matching_loss": 0.04462959840893745, + "weighted_rkd_loss": 0.0784610703587532 + }, + { + "epoch": 1.2627291242362526, + "flow_matching_loss": 0.043302136659622195, + "grad_norm": 0.49250248074531555, + "learning_rate": 2.1694999999999998e-05, + "loss": 0.1193, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030077224038541316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016055668331682683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07621011584997177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967841446399689, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4340, + "weighted_flow_matching_loss": 0.043302136659622195, + "weighted_rkd_loss": 0.07596505731344223 + }, + { + "epoch": 1.2656386383473959, + "flow_matching_loss": 0.04422338120639324, + "grad_norm": 0.5789178013801575, + "learning_rate": 2.1745e-05, + "loss": 0.1257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031977207958698274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01775688985362649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08171130493283271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967693030834198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4350, + "weighted_flow_matching_loss": 0.04422338120639324, + "weighted_rkd_loss": 0.08144730851054191 + }, + { + "epoch": 1.2685481524585394, + "flow_matching_loss": 0.045841571688652036, + "grad_norm": 0.46816468238830566, + "learning_rate": 2.1795e-05, + "loss": 0.1269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03193739615380764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017483258992433547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08135805130004883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967544376850128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4360, + "weighted_flow_matching_loss": 0.045841571688652036, + "weighted_rkd_loss": 0.08109401315450668 + }, + { + "epoch": 1.2714576665696828, + "flow_matching_loss": 0.04188045188784599, + "grad_norm": 0.5458246469497681, + "learning_rate": 2.1845e-05, + "loss": 0.125, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032687753438949585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018016592599451542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0833920992910862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996739536523819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4370, + "weighted_flow_matching_loss": 0.04188045188784599, + "weighted_rkd_loss": 0.0831201732158661 + }, + { + "epoch": 1.2743671806808263, + "flow_matching_loss": 0.043874204903841016, + "grad_norm": 0.48017939925193787, + "learning_rate": 2.1895e-05, + "loss": 0.1267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032876462861895564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017353336419910194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08310626149177551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967245757579803, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4380, + "weighted_flow_matching_loss": 0.043874204903841016, + "weighted_rkd_loss": 0.08283406421542168 + }, + { + "epoch": 1.2772766947919698, + "flow_matching_loss": 0.046855796873569486, + "grad_norm": 0.5345659852027893, + "learning_rate": 2.1945e-05, + "loss": 0.1288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032365872897207736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017464876547455786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08219662308692932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967096030712128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4390, + "weighted_flow_matching_loss": 0.046855796873569486, + "weighted_rkd_loss": 0.0819261610507965 + }, + { + "epoch": 1.280186208903113, + "flow_matching_loss": 0.04541050866246223, + "grad_norm": 0.5144512057304382, + "learning_rate": 2.1995e-05, + "loss": 0.1277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03250450883060694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01753091923892498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08253993615508079, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966946005821228, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4400, + "weighted_flow_matching_loss": 0.04541050866246223, + "weighted_rkd_loss": 0.08226711377501487 + }, + { + "epoch": 1.2830957230142566, + "flow_matching_loss": 0.04566192515194416, + "grad_norm": 0.5084413290023804, + "learning_rate": 2.2045e-05, + "loss": 0.1277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03231615368276834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017665164731442928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08229747265577317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966795563697814, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4410, + "weighted_flow_matching_loss": 0.04566192515194416, + "weighted_rkd_loss": 0.08202420994639396 + }, + { + "epoch": 1.2860052371254, + "flow_matching_loss": 0.045105167850852015, + "grad_norm": 0.5148775577545166, + "learning_rate": 2.2095e-05, + "loss": 0.123, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03073482420295477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016690705344080924, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07816035374999046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966644763946533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4420, + "weighted_flow_matching_loss": 0.045105167850852015, + "weighted_rkd_loss": 0.07789964899420738 + }, + { + "epoch": 1.2889147512365435, + "flow_matching_loss": 0.046415095776319505, + "grad_norm": 0.6108140349388123, + "learning_rate": 2.2145e-05, + "loss": 0.1302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03298190385103226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01812657341361046, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08409038111567498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966493606567383, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4430, + "weighted_flow_matching_loss": 0.046415095776319505, + "weighted_rkd_loss": 0.08380859419703483 + }, + { + "epoch": 1.291824265347687, + "flow_matching_loss": 0.046719455346465114, + "grad_norm": 0.5858809351921082, + "learning_rate": 2.2195e-05, + "loss": 0.1343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03480630051344633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01829100027680397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08790360018610954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966342151165009, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4440, + "weighted_flow_matching_loss": 0.046719455346465114, + "weighted_rkd_loss": 0.08760773316025734 + }, + { + "epoch": 1.2947337794588303, + "flow_matching_loss": 0.04676872156560421, + "grad_norm": 0.45874878764152527, + "learning_rate": 2.2245e-05, + "loss": 0.1306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03325203526765108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017609613947570324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08411368429660797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966190457344055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4450, + "weighted_flow_matching_loss": 0.04676872156560421, + "weighted_rkd_loss": 0.08382930532097817 + }, + { + "epoch": 1.2976432935699738, + "flow_matching_loss": 0.0439389705657959, + "grad_norm": 0.49442803859710693, + "learning_rate": 2.2295e-05, + "loss": 0.1274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032913419976830484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01789995636790991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08372679725289345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966038227081299, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4460, + "weighted_flow_matching_loss": 0.0439389705657959, + "weighted_rkd_loss": 0.08344245478510856 + }, + { + "epoch": 1.3005528076811173, + "flow_matching_loss": 0.04484090097248554, + "grad_norm": 0.48015791177749634, + "learning_rate": 2.2345e-05, + "loss": 0.1241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031247580610215663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017023864574730395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0795190267264843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9965885758399964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4470, + "weighted_flow_matching_loss": 0.04484090097248554, + "weighted_rkd_loss": 0.07924777492880822 + }, + { + "epoch": 1.3034623217922607, + "flow_matching_loss": 0.04789797216653824, + "grad_norm": 0.4691483676433563, + "learning_rate": 2.2395e-05, + "loss": 0.1332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0339116457849741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017744648829102518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08556793928146363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9965732991695404, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4480, + "weighted_flow_matching_loss": 0.04789797216653824, + "weighted_rkd_loss": 0.08527472913265229 + }, + { + "epoch": 1.3063718359034042, + "flow_matching_loss": 0.04208698309957981, + "grad_norm": 0.5351406931877136, + "learning_rate": 2.2445e-05, + "loss": 0.1186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03026898354291916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016243044938892125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0767810121178627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9965579748153687, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4490, + "weighted_flow_matching_loss": 0.04208698309957981, + "weighted_rkd_loss": 0.07651673927903176 + }, + { + "epoch": 1.3092813500145475, + "flow_matching_loss": 0.04263514019548893, + "grad_norm": 0.4784594178199768, + "learning_rate": 2.2495e-05, + "loss": 0.1216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031139439903199673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016986145172268152, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07926502525806427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996542626619339, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4500, + "weighted_flow_matching_loss": 0.04263514019548893, + "weighted_rkd_loss": 0.07899098470807076 + }, + { + "epoch": 1.312190864125691, + "flow_matching_loss": 0.04402074851095676, + "grad_norm": 0.5312668681144714, + "learning_rate": 2.2545e-05, + "loss": 0.1232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031462348997592926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016535590030252935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07946028932929039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9965272367000579, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4510, + "weighted_flow_matching_loss": 0.04402074851095676, + "weighted_rkd_loss": 0.07918432950973511 + }, + { + "epoch": 1.3151003782368345, + "flow_matching_loss": 0.04429629556834698, + "grad_norm": 0.552133321762085, + "learning_rate": 2.2595e-05, + "loss": 0.1264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03255357872694731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017297784425318243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08240494057536125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996511822938919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4520, + "weighted_flow_matching_loss": 0.04429629556834698, + "weighted_rkd_loss": 0.08211750984191894 + }, + { + "epoch": 1.318009892347978, + "flow_matching_loss": 0.04626792259514332, + "grad_norm": 0.5334192514419556, + "learning_rate": 2.2645e-05, + "loss": 0.1306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03323903940618038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018149752356112003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08462783247232437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964963793754578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4530, + "weighted_flow_matching_loss": 0.04626792259514332, + "weighted_rkd_loss": 0.08433130905032157 + }, + { + "epoch": 1.3209194064591214, + "flow_matching_loss": 0.04199973344802856, + "grad_norm": 0.5270112752914429, + "learning_rate": 2.2695e-05, + "loss": 0.12, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030973743088543414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016295740194618703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07824322581291199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964808821678162, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4540, + "weighted_flow_matching_loss": 0.04199973344802856, + "weighted_rkd_loss": 0.07796786725521088 + }, + { + "epoch": 1.3238289205702647, + "flow_matching_loss": 0.04367719888687134, + "grad_norm": 0.5058850646018982, + "learning_rate": 2.2745e-05, + "loss": 0.122, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03100354764610529, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01658547492697835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07859257087111474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964653730392456, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4550, + "weighted_flow_matching_loss": 0.04367719888687134, + "weighted_rkd_loss": 0.07831477969884873 + }, + { + "epoch": 1.3267384346814082, + "flow_matching_loss": 0.04417466521263123, + "grad_norm": 0.5197693109512329, + "learning_rate": 2.2795e-05, + "loss": 0.1232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03140227440744638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01648282902315259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07928737848997117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964498102664947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4560, + "weighted_flow_matching_loss": 0.04417466521263123, + "weighted_rkd_loss": 0.07900588810443879 + }, + { + "epoch": 1.3296479487925517, + "flow_matching_loss": 0.04482066966593266, + "grad_norm": 0.6070793867111206, + "learning_rate": 2.2845e-05, + "loss": 0.1217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030297493562102318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016512976679950952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07710796371102333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964342236518859, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4570, + "weighted_flow_matching_loss": 0.04482066966593266, + "weighted_rkd_loss": 0.07683300673961639 + }, + { + "epoch": 1.3325574629036951, + "flow_matching_loss": 0.044448211416602136, + "grad_norm": 0.5513048768043518, + "learning_rate": 2.2895e-05, + "loss": 0.1224, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03096422478556633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016349021531641483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07827747017145156, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964185953140259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4580, + "weighted_flow_matching_loss": 0.044448211416602136, + "weighted_rkd_loss": 0.07799711972475051 + }, + { + "epoch": 1.3354669770148386, + "flow_matching_loss": 0.043344876170158385, + "grad_norm": 0.5266980528831482, + "learning_rate": 2.2945e-05, + "loss": 0.1214, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03088665660470724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016577486880123615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07835079953074456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964029431343079, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4590, + "weighted_flow_matching_loss": 0.043344876170158385, + "weighted_rkd_loss": 0.07806896269321442 + }, + { + "epoch": 1.338376491125982, + "flow_matching_loss": 0.04482905901968479, + "grad_norm": 0.4991006851196289, + "learning_rate": 2.2995e-05, + "loss": 0.1237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031141203455626966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01688284669071436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0791652537882328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963872551918029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4600, + "weighted_flow_matching_loss": 0.04482905901968479, + "weighted_rkd_loss": 0.0788792572915554 + }, + { + "epoch": 1.3412860052371254, + "flow_matching_loss": 0.04490094967186451, + "grad_norm": 0.479941189289093, + "learning_rate": 2.3045e-05, + "loss": 0.1208, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030187912099063397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015830876491963863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07620670050382614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963715255260468, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4610, + "weighted_flow_matching_loss": 0.04490094967186451, + "weighted_rkd_loss": 0.07593020722270012 + }, + { + "epoch": 1.3441955193482689, + "flow_matching_loss": 0.04484557211399078, + "grad_norm": 0.4484858512878418, + "learning_rate": 2.3095e-05, + "loss": 0.127, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03238124717026949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017703716829419135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08246621191501617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963557720184326, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4620, + "weighted_flow_matching_loss": 0.04484557211399078, + "weighted_rkd_loss": 0.08216569796204568 + }, + { + "epoch": 1.3471050334594123, + "flow_matching_loss": 0.0411180205643177, + "grad_norm": 0.4733911454677582, + "learning_rate": 2.3145e-05, + "loss": 0.1166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029899370670318604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01595134073868394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07575008124113083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963399767875671, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4630, + "weighted_flow_matching_loss": 0.0411180205643177, + "weighted_rkd_loss": 0.07547283619642257 + }, + { + "epoch": 1.3500145475705558, + "flow_matching_loss": 0.045737313106656075, + "grad_norm": 0.4847317039966583, + "learning_rate": 2.3195e-05, + "loss": 0.1242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031125559471547604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016523010656237604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0787741295993328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963241517543793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4640, + "weighted_flow_matching_loss": 0.045737313106656075, + "weighted_rkd_loss": 0.07848456129431725 + }, + { + "epoch": 1.352924061681699, + "flow_matching_loss": 0.04171308763325214, + "grad_norm": 0.4771668314933777, + "learning_rate": 2.3245000000000002e-05, + "loss": 0.1196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030802962370216846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0165635596960783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07816948443651199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963082849979401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4650, + "weighted_flow_matching_loss": 0.04171308763325214, + "weighted_rkd_loss": 0.07788091376423836 + }, + { + "epoch": 1.3558335757928426, + "flow_matching_loss": 0.04184253849089146, + "grad_norm": 0.4842948615550995, + "learning_rate": 2.3295e-05, + "loss": 0.1199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03082246221601963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016694412846118212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07833933718502521, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962923943996429, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4660, + "weighted_flow_matching_loss": 0.04184253849089146, + "weighted_rkd_loss": 0.07804889231920242 + }, + { + "epoch": 1.358743089903986, + "flow_matching_loss": 0.044346661493182185, + "grad_norm": 0.5117174386978149, + "learning_rate": 2.3345e-05, + "loss": 0.1243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03146872725337744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017326223850250243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08026367872953415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996276468038559, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4670, + "weighted_flow_matching_loss": 0.044346661493182185, + "weighted_rkd_loss": 0.07996480837464333 + }, + { + "epoch": 1.3616526040151296, + "flow_matching_loss": 0.042052822560071944, + "grad_norm": 0.5014215111732483, + "learning_rate": 2.3395000000000002e-05, + "loss": 0.124, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03241212218999863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017472876515239478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08229712098836899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962605059146881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4680, + "weighted_flow_matching_loss": 0.042052822560071944, + "weighted_rkd_loss": 0.08198937475681305 + }, + { + "epoch": 1.364562118126273, + "flow_matching_loss": 0.04128942675888538, + "grad_norm": 0.5005990862846375, + "learning_rate": 2.3445e-05, + "loss": 0.1169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030026891455054282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01581104677170515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07586482986807823, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962445020675659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4690, + "weighted_flow_matching_loss": 0.04128942675888538, + "weighted_rkd_loss": 0.07557989843189716 + }, + { + "epoch": 1.3674716322374163, + "flow_matching_loss": 0.04679095894098282, + "grad_norm": 0.5060570240020752, + "learning_rate": 2.3495e-05, + "loss": 0.1249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031028494797647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01633110847324133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07838809788227082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962284803390503, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4700, + "weighted_flow_matching_loss": 0.04679095894098282, + "weighted_rkd_loss": 0.0780924528837204 + }, + { + "epoch": 1.3703811463485598, + "flow_matching_loss": 0.041962244920432566, + "grad_norm": 0.4737093150615692, + "learning_rate": 2.3545000000000002e-05, + "loss": 0.1184, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03024671394377947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016217229887843132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07671065852046013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962123930454254, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4710, + "weighted_flow_matching_loss": 0.041962244920432566, + "weighted_rkd_loss": 0.07642010860145092 + }, + { + "epoch": 1.3732906604597033, + "flow_matching_loss": 0.04150658324360847, + "grad_norm": 0.48893263936042786, + "learning_rate": 2.3595e-05, + "loss": 0.1216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03182344827800989, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01673230081796646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08037919625639915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961962997913361, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4720, + "weighted_flow_matching_loss": 0.04150658324360847, + "weighted_rkd_loss": 0.08007345646619797 + }, + { + "epoch": 1.3762001745708465, + "flow_matching_loss": 0.0430319607257843, + "grad_norm": 0.4598107933998108, + "learning_rate": 2.3645e-05, + "loss": 0.1198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03062412664294243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015832900907844304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07708115428686142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961801826953888, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4730, + "weighted_flow_matching_loss": 0.0430319607257843, + "weighted_rkd_loss": 0.07678673043847084 + }, + { + "epoch": 1.3791096886819902, + "flow_matching_loss": 0.04357755333185196, + "grad_norm": 0.4516480267047882, + "learning_rate": 2.3695000000000003e-05, + "loss": 0.1207, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030591201968491077, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016218949109315872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07740135192871093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961640179157257, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4740, + "weighted_flow_matching_loss": 0.04357755333185196, + "weighted_rkd_loss": 0.07710444405674935 + }, + { + "epoch": 1.3820192027931335, + "flow_matching_loss": 0.04131851941347122, + "grad_norm": 0.43265029788017273, + "learning_rate": 2.3745e-05, + "loss": 0.113, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028701019659638404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01458593299612403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07198797166347504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961478114128113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4750, + "weighted_flow_matching_loss": 0.04131851941347122, + "weighted_rkd_loss": 0.07171067260205746 + }, + { + "epoch": 1.384928716904277, + "flow_matching_loss": 0.04429641850292683, + "grad_norm": 0.5126135349273682, + "learning_rate": 2.3795000000000002e-05, + "loss": 0.1181, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02942121960222721, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015246367175132036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07408880665898324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961315810680389, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4760, + "weighted_flow_matching_loss": 0.04429641850292683, + "weighted_rkd_loss": 0.07380218394100665 + }, + { + "epoch": 1.3878382310154205, + "flow_matching_loss": 0.04209242425858974, + "grad_norm": 0.5658693313598633, + "learning_rate": 2.3845e-05, + "loss": 0.1171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029692774824798106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01590500697493553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07529055625200272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961153090000152, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4770, + "weighted_flow_matching_loss": 0.04209242425858974, + "weighted_rkd_loss": 0.0749980803579092 + }, + { + "epoch": 1.3907477451265637, + "flow_matching_loss": 0.04314549192786217, + "grad_norm": 0.5069836378097534, + "learning_rate": 2.3895e-05, + "loss": 0.1226, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03175110425800085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01624085698276758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07974306493997574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960990071296691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4780, + "weighted_flow_matching_loss": 0.04314549192786217, + "weighted_rkd_loss": 0.07943197190761567 + }, + { + "epoch": 1.3936572592377072, + "flow_matching_loss": 0.04404900372028351, + "grad_norm": 0.5494664907455444, + "learning_rate": 2.3945000000000002e-05, + "loss": 0.1186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029437338188290595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016011341847479342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07488601729273796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960826754570007, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4790, + "weighted_flow_matching_loss": 0.04404900372028351, + "weighted_rkd_loss": 0.07459267899394036 + }, + { + "epoch": 1.3965667733488507, + "flow_matching_loss": 0.041549882292747496, + "grad_norm": 0.49289852380752563, + "learning_rate": 2.3995e-05, + "loss": 0.1139, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028821900859475134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01500080768018961, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07264460884034633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960663020610809, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4800, + "weighted_flow_matching_loss": 0.041549882292747496, + "weighted_rkd_loss": 0.07235885635018349 + }, + { + "epoch": 1.3994762874599942, + "flow_matching_loss": 0.04184562861919403, + "grad_norm": 0.5413471460342407, + "learning_rate": 2.4045e-05, + "loss": 0.1183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030213256552815437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016296792309731245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07672330439090728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960498988628388, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4810, + "weighted_flow_matching_loss": 0.04184562861919403, + "weighted_rkd_loss": 0.07642023488879204 + }, + { + "epoch": 1.4023858015711377, + "flow_matching_loss": 0.04238753728568554, + "grad_norm": 0.5425558686256409, + "learning_rate": 2.4095000000000002e-05, + "loss": 0.1204, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030931010469794275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016428069211542606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0782900907099247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960334599018097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4820, + "weighted_flow_matching_loss": 0.04238753728568554, + "weighted_rkd_loss": 0.07797956988215446 + }, + { + "epoch": 1.405295315682281, + "flow_matching_loss": 0.040809236094355585, + "grad_norm": 0.52039635181427, + "learning_rate": 2.4145e-05, + "loss": 0.1169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03042989708483219, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015551321394741534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07641111612319947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960169911384582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4830, + "weighted_flow_matching_loss": 0.040809236094355585, + "weighted_rkd_loss": 0.07610676288604737 + }, + { + "epoch": 1.4082048297934244, + "flow_matching_loss": 0.045377876982092856, + "grad_norm": 0.5849554538726807, + "learning_rate": 2.4195e-05, + "loss": 0.1235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030871140025556087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01669047074392438, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07843274995684624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996000474691391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4840, + "weighted_flow_matching_loss": 0.045377876982092856, + "weighted_rkd_loss": 0.07811905667185784 + }, + { + "epoch": 1.411114343904568, + "flow_matching_loss": 0.04192312844097614, + "grad_norm": 0.476043701171875, + "learning_rate": 2.4245000000000002e-05, + "loss": 0.1203, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031170235201716422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016304184310138227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07864465489983559, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959839403629303, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4850, + "weighted_flow_matching_loss": 0.04192312844097614, + "weighted_rkd_loss": 0.0783288098871708 + }, + { + "epoch": 1.4140238580157114, + "flow_matching_loss": 0.04590509682893753, + "grad_norm": 0.510171115398407, + "learning_rate": 2.4295e-05, + "loss": 0.1199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029435120336711407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01544699603691697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0743172362446785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959673643112182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4860, + "weighted_flow_matching_loss": 0.04590509682893753, + "weighted_rkd_loss": 0.07401754111051559 + }, + { + "epoch": 1.4169333721268549, + "flow_matching_loss": 0.03991883099079132, + "grad_norm": 0.48003068566322327, + "learning_rate": 2.4345e-05, + "loss": 0.1139, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02930215746164322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015637150686234235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07424146570265293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959507524967194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4870, + "weighted_flow_matching_loss": 0.03991883099079132, + "weighted_rkd_loss": 0.0739408291876316 + }, + { + "epoch": 1.4198428862379981, + "flow_matching_loss": 0.04165617898106575, + "grad_norm": 0.5421159267425537, + "learning_rate": 2.4395000000000003e-05, + "loss": 0.1171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03009853810071945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015597395598888397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0757944718003273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.995934122800827, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4880, + "weighted_flow_matching_loss": 0.04165617898106575, + "weighted_rkd_loss": 0.07548632100224495 + }, + { + "epoch": 1.4227524003491416, + "flow_matching_loss": 0.04291961491107941, + "grad_norm": 0.5582727193832397, + "learning_rate": 2.4445e-05, + "loss": 0.1187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030098510161042213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015939982142299414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0761370025575161, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959174394607544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4890, + "weighted_flow_matching_loss": 0.04291961491107941, + "weighted_rkd_loss": 0.07582618296146393 + }, + { + "epoch": 1.4256619144602851, + "flow_matching_loss": 0.044600742682814595, + "grad_norm": 0.5179396271705627, + "learning_rate": 2.4495000000000002e-05, + "loss": 0.1255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031955149956047535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01735329916700721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0812635987997055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959007322788238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4900, + "weighted_flow_matching_loss": 0.044600742682814595, + "weighted_rkd_loss": 0.08093045949935913 + }, + { + "epoch": 1.4285714285714286, + "flow_matching_loss": 0.04413844496011734, + "grad_norm": 0.5267637372016907, + "learning_rate": 2.4545000000000003e-05, + "loss": 0.1211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030752195045351982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015787849854677915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07729224041104317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958839774131775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4910, + "weighted_flow_matching_loss": 0.04413844496011734, + "weighted_rkd_loss": 0.07697409652173519 + }, + { + "epoch": 1.431480942682572, + "flow_matching_loss": 0.04128686040639877, + "grad_norm": 0.5341330170631409, + "learning_rate": 2.4595e-05, + "loss": 0.1199, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030978860333561897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016957191191613673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07891491129994392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958672046661377, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4920, + "weighted_flow_matching_loss": 0.04128686040639877, + "weighted_rkd_loss": 0.07858878001570702 + }, + { + "epoch": 1.4343904567937154, + "flow_matching_loss": 0.0449594285339117, + "grad_norm": 0.44644981622695923, + "learning_rate": 2.4645000000000002e-05, + "loss": 0.1186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02921724896878004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015490975137799979, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07392547354102134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958503901958465, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4930, + "weighted_flow_matching_loss": 0.0449594285339117, + "weighted_rkd_loss": 0.07361870482563973 + }, + { + "epoch": 1.4372999709048588, + "flow_matching_loss": 0.04152275621891022, + "grad_norm": 0.47468605637550354, + "learning_rate": 2.4695000000000003e-05, + "loss": 0.1192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031021990440785886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015977317560464145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07802129834890366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958335399627686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4940, + "weighted_flow_matching_loss": 0.04152275621891022, + "weighted_rkd_loss": 0.07769623324275017 + }, + { + "epoch": 1.4402094850160023, + "flow_matching_loss": 0.04069807566702366, + "grad_norm": 0.5551746487617493, + "learning_rate": 2.4745e-05, + "loss": 0.1171, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030014824867248536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016665180772542955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07669482976198197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958166599273681, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4950, + "weighted_flow_matching_loss": 0.04069807566702366, + "weighted_rkd_loss": 0.0763739988207817 + }, + { + "epoch": 1.4431189991271458, + "flow_matching_loss": 0.042391576990485194, + "grad_norm": 0.4531959891319275, + "learning_rate": 2.4795000000000002e-05, + "loss": 0.1147, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028611348755657674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015388564299792051, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0726112611591816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957997441291809, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4960, + "weighted_flow_matching_loss": 0.042391576990485194, + "weighted_rkd_loss": 0.07230627834796906 + }, + { + "epoch": 1.4460285132382893, + "flow_matching_loss": 0.04030802436172962, + "grad_norm": 0.5969329476356506, + "learning_rate": 2.4845000000000004e-05, + "loss": 0.1165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030391155183315276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01571120535954833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07649351619184017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957827925682068, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4970, + "weighted_flow_matching_loss": 0.04030802436172962, + "weighted_rkd_loss": 0.07617089375853539 + }, + { + "epoch": 1.4489380273494326, + "flow_matching_loss": 0.04224388860166073, + "grad_norm": 0.5282535552978516, + "learning_rate": 2.4895e-05, + "loss": 0.1153, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02891715578734875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01551342075690627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07334773316979408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957658171653747, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4980, + "weighted_flow_matching_loss": 0.04224388860166073, + "weighted_rkd_loss": 0.07303716391324996 + }, + { + "epoch": 1.451847541460576, + "flow_matching_loss": 0.04217173010110855, + "grad_norm": 0.5318191647529602, + "learning_rate": 2.4945000000000003e-05, + "loss": 0.1157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029305890947580338, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015278606116771698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07389038875699043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957488000392913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4990, + "weighted_flow_matching_loss": 0.04217173010110855, + "weighted_rkd_loss": 0.07357626780867577 + }, + { + "epoch": 1.4547570555717195, + "flow_matching_loss": 0.03993020262569189, + "grad_norm": 0.5331554412841797, + "learning_rate": 2.4995000000000004e-05, + "loss": 0.1133, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029206221178174018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01527288630604744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07368532828986644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957317471504211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5000, + "weighted_flow_matching_loss": 0.03993020262569189, + "weighted_rkd_loss": 0.07337081208825111 + }, + { + "epoch": 1.457666569682863, + "flow_matching_loss": 0.04358153380453587, + "grad_norm": 0.5150550007820129, + "learning_rate": 2.5045e-05, + "loss": 0.1133, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027756115049123765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014493719860911369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07000594958662987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957146644592285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5010, + "weighted_flow_matching_loss": 0.04358153380453587, + "weighted_rkd_loss": 0.069705930352211 + }, + { + "epoch": 1.4605760837940065, + "flow_matching_loss": 0.041769911721348765, + "grad_norm": 0.5586439967155457, + "learning_rate": 2.5095000000000003e-05, + "loss": 0.1211, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03139507211744785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01684995470568538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07964009866118431, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956975400447845, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5020, + "weighted_flow_matching_loss": 0.041769911721348765, + "weighted_rkd_loss": 0.07929743379354477 + }, + { + "epoch": 1.4634855979051498, + "flow_matching_loss": 0.040325346402823925, + "grad_norm": 0.4919849932193756, + "learning_rate": 2.5145e-05, + "loss": 0.1122, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02858587261289358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014967425260692835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07213917076587677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956803917884827, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5030, + "weighted_flow_matching_loss": 0.040325346402823925, + "weighted_rkd_loss": 0.07182755693793297 + }, + { + "epoch": 1.4663951120162932, + "flow_matching_loss": 0.041921428963541986, + "grad_norm": 0.5575131773948669, + "learning_rate": 2.5195000000000002e-05, + "loss": 0.1216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03182960413396359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01634123232215643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08000044077634812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956632137298584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5040, + "weighted_flow_matching_loss": 0.041921428963541986, + "weighted_rkd_loss": 0.07965350300073623 + }, + { + "epoch": 1.4693046261274367, + "flow_matching_loss": 0.0427427526563406, + "grad_norm": 0.5561245679855347, + "learning_rate": 2.5245000000000003e-05, + "loss": 0.1175, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02947978172451258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01610988872125745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07506945356726646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956459879875184, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5050, + "weighted_flow_matching_loss": 0.0427427526563406, + "weighted_rkd_loss": 0.07474259063601493 + }, + { + "epoch": 1.4722141402385802, + "flow_matching_loss": 0.041811144724488256, + "grad_norm": 0.5413380265235901, + "learning_rate": 2.5295e-05, + "loss": 0.1198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030638173036277295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017021388467401265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07829773426055908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956287443637848, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5060, + "weighted_flow_matching_loss": 0.041811144724488256, + "weighted_rkd_loss": 0.07795545533299446 + }, + { + "epoch": 1.4751236543497237, + "flow_matching_loss": 0.04108384326100349, + "grad_norm": 0.5505955219268799, + "learning_rate": 2.5345e-05, + "loss": 0.1161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029633889347314833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01606691423803568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0753346934914589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956114590167999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5070, + "weighted_flow_matching_loss": 0.04108384326100349, + "weighted_rkd_loss": 0.0750040590763092 + }, + { + "epoch": 1.478033168460867, + "flow_matching_loss": 0.04285131879150868, + "grad_norm": 0.5893809795379639, + "learning_rate": 2.5395e-05, + "loss": 0.1196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03041365183889866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016263923700898887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07709122747182846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9955941200256347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5080, + "weighted_flow_matching_loss": 0.04285131879150868, + "weighted_rkd_loss": 0.07675156593322754 + }, + { + "epoch": 1.4809426825720104, + "flow_matching_loss": 0.04084861241281033, + "grad_norm": 0.49257293343544006, + "learning_rate": 2.5444999999999998e-05, + "loss": 0.1154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0295830674469471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015738281793892382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07490441650152206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9955767631530762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5090, + "weighted_flow_matching_loss": 0.04084861241281033, + "weighted_rkd_loss": 0.07457310780882835 + }, + { + "epoch": 1.483852196683154, + "flow_matching_loss": 0.04336781203746796, + "grad_norm": 0.5392817854881287, + "learning_rate": 2.5495e-05, + "loss": 0.118, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029234857484698295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01645757146179676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07492728605866432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9955593764781951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5100, + "weighted_flow_matching_loss": 0.04336781203746796, + "weighted_rkd_loss": 0.07459455877542495 + }, + { + "epoch": 1.4867617107942974, + "flow_matching_loss": 0.04113542214035988, + "grad_norm": 0.5207589864730835, + "learning_rate": 2.5545e-05, + "loss": 0.1132, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02835085056722164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015653869975358246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0723555713891983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9955419600009918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5110, + "weighted_flow_matching_loss": 0.04113542214035988, + "weighted_rkd_loss": 0.07203300967812538 + }, + { + "epoch": 1.489671224905441, + "flow_matching_loss": 0.04303363375365734, + "grad_norm": 0.46696770191192627, + "learning_rate": 2.5594999999999998e-05, + "loss": 0.1179, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029871758073568344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015494148992002011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07523766458034516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9955244958400726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5120, + "weighted_flow_matching_loss": 0.04303363375365734, + "weighted_rkd_loss": 0.07490093037486076 + }, + { + "epoch": 1.4925807390165842, + "flow_matching_loss": 0.04012807607650757, + "grad_norm": 0.4532116651535034, + "learning_rate": 2.5645e-05, + "loss": 0.1144, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02934359461069107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01594381928443909, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07463100925087929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.995507001876831, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5130, + "weighted_flow_matching_loss": 0.04012807607650757, + "weighted_rkd_loss": 0.07429568395018578 + }, + { + "epoch": 1.4954902531277277, + "flow_matching_loss": 0.04173567071557045, + "grad_norm": 0.4925368130207062, + "learning_rate": 2.5695e-05, + "loss": 0.113, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028201545216143133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015200645849108695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07160373479127884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954894781112671, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5140, + "weighted_flow_matching_loss": 0.04173567071557045, + "weighted_rkd_loss": 0.07128075435757637 + }, + { + "epoch": 1.4983997672388711, + "flow_matching_loss": 0.041132301837205884, + "grad_norm": 0.4465537369251251, + "learning_rate": 2.5744999999999998e-05, + "loss": 0.116, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029877835884690285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015447455272078515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07520312741398812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954719245433807, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5150, + "weighted_flow_matching_loss": 0.041132301837205884, + "weighted_rkd_loss": 0.07486260607838631 + }, + { + "epoch": 1.5013092813500144, + "flow_matching_loss": 0.04042027145624161, + "grad_norm": 0.5536547303199768, + "learning_rate": 2.5795e-05, + "loss": 0.1133, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028902390226721763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015427786577492952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07323256731033326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954543232917785, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5160, + "weighted_flow_matching_loss": 0.04042027145624161, + "weighted_rkd_loss": 0.07289966046810151 + }, + { + "epoch": 1.504218795461158, + "flow_matching_loss": 0.042050937190651894, + "grad_norm": 0.5209099650382996, + "learning_rate": 2.5845e-05, + "loss": 0.1163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029361690394580364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015817377902567386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07454075813293456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.995436692237854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5170, + "weighted_flow_matching_loss": 0.042050937190651894, + "weighted_rkd_loss": 0.07420062348246574 + }, + { + "epoch": 1.5071283095723014, + "flow_matching_loss": 0.04020294994115829, + "grad_norm": 0.4794981777667999, + "learning_rate": 2.5895e-05, + "loss": 0.1161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030229418911039828, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01577084781602025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07622968479990959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954190373420715, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5180, + "weighted_flow_matching_loss": 0.04020294994115829, + "weighted_rkd_loss": 0.07588048726320266 + }, + { + "epoch": 1.5100378236834449, + "flow_matching_loss": 0.03978218846023083, + "grad_norm": 0.47442826628685, + "learning_rate": 2.5945e-05, + "loss": 0.1136, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02923715189099312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015666454471647738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07414075881242752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954013347625732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5190, + "weighted_flow_matching_loss": 0.03978218846023083, + "weighted_rkd_loss": 0.07379978895187378 + }, + { + "epoch": 1.5129473377945883, + "flow_matching_loss": 0.04058701954782009, + "grad_norm": 0.5672383904457092, + "learning_rate": 2.5995e-05, + "loss": 0.1102, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0276757188141346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014575479924678803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.069926917552948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9953836083412171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5200, + "weighted_flow_matching_loss": 0.04058701954782009, + "weighted_rkd_loss": 0.0696041002869606 + }, + { + "epoch": 1.5158568519057316, + "flow_matching_loss": 0.043448731675744055, + "grad_norm": 0.4833974838256836, + "learning_rate": 2.6045e-05, + "loss": 0.1201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030139721930027008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01668962175026536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07696906551718712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9953658401966095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5210, + "weighted_flow_matching_loss": 0.043448731675744055, + "weighted_rkd_loss": 0.07661238089203834 + }, + { + "epoch": 1.5187663660168753, + "flow_matching_loss": 0.04386789426207542, + "grad_norm": 0.4392998516559601, + "learning_rate": 2.6095e-05, + "loss": 0.1176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029428223334252834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015201866254210472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0740583136677742, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.995348048210144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5220, + "weighted_flow_matching_loss": 0.04386789426207542, + "weighted_rkd_loss": 0.07371380031108857 + }, + { + "epoch": 1.5216758801280186, + "flow_matching_loss": 0.04098653718829155, + "grad_norm": 0.49525928497314453, + "learning_rate": 2.6145e-05, + "loss": 0.114, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029000826366245746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01539333313703537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0733949862420559, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9953302145004272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5230, + "weighted_flow_matching_loss": 0.04098653718829155, + "weighted_rkd_loss": 0.07305225506424903 + }, + { + "epoch": 1.524585394239162, + "flow_matching_loss": 0.04158078245818615, + "grad_norm": 0.4991057813167572, + "learning_rate": 2.6195e-05, + "loss": 0.1148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02912954967468977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015272990707308054, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07353208847343921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9953123331069946, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5240, + "weighted_flow_matching_loss": 0.04158078245818615, + "weighted_rkd_loss": 0.07318738698959351 + }, + { + "epoch": 1.5274949083503055, + "flow_matching_loss": 0.042361154407262805, + "grad_norm": 0.45189088582992554, + "learning_rate": 2.6245e-05, + "loss": 0.1166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029684115201234818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015192167460918426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07456039823591709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.995294451713562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5250, + "weighted_flow_matching_loss": 0.042361154407262805, + "weighted_rkd_loss": 0.07420957274734974 + }, + { + "epoch": 1.5304044224614488, + "flow_matching_loss": 0.04153250381350517, + "grad_norm": 0.5006435513496399, + "learning_rate": 2.6294999999999998e-05, + "loss": 0.1162, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029696076922118664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015644286572933198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07503644078969955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9952765166759491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5260, + "weighted_flow_matching_loss": 0.04153250381350517, + "weighted_rkd_loss": 0.07468201518058777 + }, + { + "epoch": 1.5333139365725925, + "flow_matching_loss": 0.041182616539299485, + "grad_norm": 0.5160431265830994, + "learning_rate": 2.6345e-05, + "loss": 0.1114, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027848699316382407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014872459508478642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07056985720992089, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9952585458755493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5270, + "weighted_flow_matching_loss": 0.041182616539299485, + "weighted_rkd_loss": 0.07023521438241005 + }, + { + "epoch": 1.5362234506837358, + "flow_matching_loss": 0.042605215683579445, + "grad_norm": 0.5002658367156982, + "learning_rate": 2.6395e-05, + "loss": 0.1112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027213257737457752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014450577087700367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06887709237635135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9952405512332916, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5280, + "weighted_flow_matching_loss": 0.042605215683579445, + "weighted_rkd_loss": 0.06854928880929947 + }, + { + "epoch": 1.5391329647948793, + "flow_matching_loss": 0.04318515658378601, + "grad_norm": 0.45914679765701294, + "learning_rate": 2.6445e-05, + "loss": 0.1119, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02745679821819067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014155286271125079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06906888261437416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9952225089073181, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5290, + "weighted_flow_matching_loss": 0.04318515658378601, + "weighted_rkd_loss": 0.0687388963997364 + }, + { + "epoch": 1.5420424789060228, + "flow_matching_loss": 0.0396482739597559, + "grad_norm": 0.45377543568611145, + "learning_rate": 2.6495e-05, + "loss": 0.1102, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02814194094389677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014640911109745502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07092479206621646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9952044367790223, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5300, + "weighted_flow_matching_loss": 0.0396482739597559, + "weighted_rkd_loss": 0.07058469019830227 + }, + { + "epoch": 1.544951993017166, + "flow_matching_loss": 0.03964730873703957, + "grad_norm": 0.49224773049354553, + "learning_rate": 2.6545e-05, + "loss": 0.1093, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027589914202690125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014855949487537146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07003577761352062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951863408088684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5310, + "weighted_flow_matching_loss": 0.03964730873703957, + "weighted_rkd_loss": 0.06969864070415496 + }, + { + "epoch": 1.5478615071283097, + "flow_matching_loss": 0.0411466795951128, + "grad_norm": 0.3907390832901001, + "learning_rate": 2.6595e-05, + "loss": 0.1079, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026600496284663676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013872578740119934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06707357130944729, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951682031154633, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5320, + "weighted_flow_matching_loss": 0.0411466795951128, + "weighted_rkd_loss": 0.06674948967993259 + }, + { + "epoch": 1.550771021239453, + "flow_matching_loss": 0.04218398258090019, + "grad_norm": 0.525648832321167, + "learning_rate": 2.6645e-05, + "loss": 0.112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02789376899600029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014391845278441905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07017938382923603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951500296592712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5330, + "weighted_flow_matching_loss": 0.04218398258090019, + "weighted_rkd_loss": 0.06983901523053646 + }, + { + "epoch": 1.5536805353505965, + "flow_matching_loss": 0.04164282642304897, + "grad_norm": 0.4639146327972412, + "learning_rate": 2.6695e-05, + "loss": 0.1077, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026569329760968687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013259463012218475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06639812290668487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951318264007568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5340, + "weighted_flow_matching_loss": 0.04164282642304897, + "weighted_rkd_loss": 0.06607489809393882 + }, + { + "epoch": 1.55659004946174, + "flow_matching_loss": 0.04187535084784031, + "grad_norm": 0.48225370049476624, + "learning_rate": 2.6745e-05, + "loss": 0.109, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0268386647105217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013764330185949802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06744165942072869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951135873794555, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5350, + "weighted_flow_matching_loss": 0.04187535084784031, + "weighted_rkd_loss": 0.06711210981011391 + }, + { + "epoch": 1.5594995635728832, + "flow_matching_loss": 0.03909413665533066, + "grad_norm": 0.5812745690345764, + "learning_rate": 2.6795e-05, + "loss": 0.1091, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027953347936272622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014480488561093807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07038718350231647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9950953125953674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5360, + "weighted_flow_matching_loss": 0.03909413665533066, + "weighted_rkd_loss": 0.07004193216562271 + }, + { + "epoch": 1.562409077684027, + "flow_matching_loss": 0.04196959361433983, + "grad_norm": 0.502668559551239, + "learning_rate": 2.6845e-05, + "loss": 0.111, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027429329045116902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01452661557123065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06938527263700962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9950770080089569, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5370, + "weighted_flow_matching_loss": 0.04196959361433983, + "weighted_rkd_loss": 0.06904369331896305 + }, + { + "epoch": 1.5653185917951702, + "flow_matching_loss": 0.0396241519600153, + "grad_norm": 0.5130542516708374, + "learning_rate": 2.6895e-05, + "loss": 0.1106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02817479334771633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014969048276543617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07131863385438919, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9950586676597595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5380, + "weighted_flow_matching_loss": 0.0396241519600153, + "weighted_rkd_loss": 0.0709662303328514 + }, + { + "epoch": 1.5682281059063137, + "flow_matching_loss": 0.04094395488500595, + "grad_norm": 0.46721041202545166, + "learning_rate": 2.6945e-05, + "loss": 0.1088, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026816859282553195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014525292348116637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06815901175141334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9950402975082397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5390, + "weighted_flow_matching_loss": 0.04094395488500595, + "weighted_rkd_loss": 0.06782097592949868 + }, + { + "epoch": 1.5711376200174572, + "flow_matching_loss": 0.04114324264228344, + "grad_norm": 0.5048618316650391, + "learning_rate": 2.6995e-05, + "loss": 0.1087, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026916772499680518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014067908376455307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06790145412087441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9950218856334686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5400, + "weighted_flow_matching_loss": 0.04114324264228344, + "weighted_rkd_loss": 0.06756342276930809 + }, + { + "epoch": 1.5740471341286004, + "flow_matching_loss": 0.04030226469039917, + "grad_norm": 0.41758909821510315, + "learning_rate": 2.7045e-05, + "loss": 0.108, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02705778870731592, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013893797900527716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06800937466323376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9950034320354462, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5410, + "weighted_flow_matching_loss": 0.04030226469039917, + "weighted_rkd_loss": 0.06766956858336926 + }, + { + "epoch": 1.576956648239744, + "flow_matching_loss": 0.039034884236752984, + "grad_norm": 0.48951759934425354, + "learning_rate": 2.7095e-05, + "loss": 0.1063, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027044008113443853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01355922743678093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06764724478125572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9949849545955658, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5420, + "weighted_flow_matching_loss": 0.039034884236752984, + "weighted_rkd_loss": 0.06730801165103913 + }, + { + "epoch": 1.5798661623508874, + "flow_matching_loss": 0.04063068851828575, + "grad_norm": 0.45939987897872925, + "learning_rate": 2.7145000000000002e-05, + "loss": 0.1074, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026521454192698003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014051882736384868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06709479168057442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994966459274292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5430, + "weighted_flow_matching_loss": 0.04063068851828575, + "weighted_rkd_loss": 0.06675705388188362 + }, + { + "epoch": 1.5827756764620309, + "flow_matching_loss": 0.040053740888834, + "grad_norm": 0.42719388008117676, + "learning_rate": 2.7195e-05, + "loss": 0.1115, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028518005087971688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014759140368551017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07179515063762665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9949479103088379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5440, + "weighted_flow_matching_loss": 0.040053740888834, + "weighted_rkd_loss": 0.07143247425556183 + }, + { + "epoch": 1.5856851905731744, + "flow_matching_loss": 0.04285055622458458, + "grad_norm": 0.47188708186149597, + "learning_rate": 2.7245e-05, + "loss": 0.1148, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02887026034295559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014585093595087528, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07232561483979225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9949293315410614, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5450, + "weighted_flow_matching_loss": 0.04285055622458458, + "weighted_rkd_loss": 0.07195885442197322 + }, + { + "epoch": 1.5885947046843176, + "flow_matching_loss": 0.04131327345967293, + "grad_norm": 0.4980974793434143, + "learning_rate": 2.7295000000000002e-05, + "loss": 0.1106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027716648764908314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0142349012196064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06966819763183593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994910717010498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5460, + "weighted_flow_matching_loss": 0.04131327345967293, + "weighted_rkd_loss": 0.06931363418698311 + }, + { + "epoch": 1.5915042187954611, + "flow_matching_loss": 0.04052703604102135, + "grad_norm": 0.4827839434146881, + "learning_rate": 2.7345e-05, + "loss": 0.1076, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026807987317442893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013805016316473484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0674209900200367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948920726776123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5470, + "weighted_flow_matching_loss": 0.04052703604102135, + "weighted_rkd_loss": 0.06707661934196948 + }, + { + "epoch": 1.5944137329066046, + "flow_matching_loss": 0.04018960632383824, + "grad_norm": 0.5465809106826782, + "learning_rate": 2.7395e-05, + "loss": 0.1106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028076101467013358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014637916255742312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07079011797904969, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948733985424042, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5480, + "weighted_flow_matching_loss": 0.04018960632383824, + "weighted_rkd_loss": 0.07042717635631561 + }, + { + "epoch": 1.597323247017748, + "flow_matching_loss": 0.04106347374618054, + "grad_norm": 0.5319583415985107, + "learning_rate": 2.7445000000000002e-05, + "loss": 0.1076, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026511434651911258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013818510994315147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06684137992560864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948546767234803, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5490, + "weighted_flow_matching_loss": 0.04106347374618054, + "weighted_rkd_loss": 0.06649745404720306 + }, + { + "epoch": 1.6002327611288916, + "flow_matching_loss": 0.039453381672501564, + "grad_norm": 0.42422905564308167, + "learning_rate": 2.7495e-05, + "loss": 0.1071, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02679452523589134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014359744545072318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06794879473745823, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948359370231629, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5500, + "weighted_flow_matching_loss": 0.039453381672501564, + "weighted_rkd_loss": 0.06759789697825909 + }, + { + "epoch": 1.6031422752400348, + "flow_matching_loss": 0.04149070270359516, + "grad_norm": 0.46598997712135315, + "learning_rate": 2.7545e-05, + "loss": 0.1125, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02822013720870018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014892091043293476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07133236564695836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948171496391296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5510, + "weighted_flow_matching_loss": 0.04149070270359516, + "weighted_rkd_loss": 0.07096268944442272 + }, + { + "epoch": 1.6060517893511783, + "flow_matching_loss": 0.040398200042545795, + "grad_norm": 0.5003460049629211, + "learning_rate": 2.7595e-05, + "loss": 0.112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028590204194188118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014782637637108565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07196304649114608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9947983384132385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5520, + "weighted_flow_matching_loss": 0.040398200042545795, + "weighted_rkd_loss": 0.07158870548009873 + }, + { + "epoch": 1.6089613034623218, + "flow_matching_loss": 0.04374849572777748, + "grad_norm": 0.43169477581977844, + "learning_rate": 2.7645e-05, + "loss": 0.1151, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028338635712862013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015055139176547527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07173241041600704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9947794795036315, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5530, + "weighted_flow_matching_loss": 0.04374849572777748, + "weighted_rkd_loss": 0.07135795690119266 + }, + { + "epoch": 1.6118708175734653, + "flow_matching_loss": 0.03824809528887272, + "grad_norm": 0.43386685848236084, + "learning_rate": 2.7695e-05, + "loss": 0.1084, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02793598212301731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014644645992666483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07051661014556884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9947605967521668, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5540, + "weighted_flow_matching_loss": 0.03824809528887272, + "weighted_rkd_loss": 0.07014715187251568 + }, + { + "epoch": 1.6147803316846088, + "flow_matching_loss": 0.040653522685170175, + "grad_norm": 0.5400391221046448, + "learning_rate": 2.7745e-05, + "loss": 0.1064, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026137337274849414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013777087256312371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06605176217854022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9947416841983795, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5550, + "weighted_flow_matching_loss": 0.040653522685170175, + "weighted_rkd_loss": 0.06570442132651806 + }, + { + "epoch": 1.617689845795752, + "flow_matching_loss": 0.04077931270003319, + "grad_norm": 0.4879228472709656, + "learning_rate": 2.7795e-05, + "loss": 0.1054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026049969531595708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012880561128258704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0649804998189211, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9947227358818054, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5560, + "weighted_flow_matching_loss": 0.04077931270003319, + "weighted_rkd_loss": 0.06463758051395416 + }, + { + "epoch": 1.6205993599068955, + "flow_matching_loss": 0.04111233167350292, + "grad_norm": 0.4383086860179901, + "learning_rate": 2.7845000000000002e-05, + "loss": 0.1102, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02720225788652897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015028162114322185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06943267732858657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99470374584198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5570, + "weighted_flow_matching_loss": 0.04111233167350292, + "weighted_rkd_loss": 0.06906493753194809 + }, + { + "epoch": 1.623508874018039, + "flow_matching_loss": 0.04031649995595217, + "grad_norm": 0.4812328815460205, + "learning_rate": 2.7895e-05, + "loss": 0.1084, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027126256190240384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01416535433381796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06841786690056324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946847319602966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5580, + "weighted_flow_matching_loss": 0.04031649995595217, + "weighted_rkd_loss": 0.06805421262979508 + }, + { + "epoch": 1.6264183881291823, + "flow_matching_loss": 0.04050728715956211, + "grad_norm": 0.4456148147583008, + "learning_rate": 2.7945e-05, + "loss": 0.1085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02722302172333002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013908369839191437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06835441291332245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946656823158264, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5590, + "weighted_flow_matching_loss": 0.04050728715956211, + "weighted_rkd_loss": 0.06798980198800564 + }, + { + "epoch": 1.629327902240326, + "flow_matching_loss": 0.03936343491077423, + "grad_norm": 0.5222947001457214, + "learning_rate": 2.7995000000000002e-05, + "loss": 0.1088, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027537000924348833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014723751693964005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06979775354266167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946465969085694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5600, + "weighted_flow_matching_loss": 0.03936343491077423, + "weighted_rkd_loss": 0.06942408978939056 + }, + { + "epoch": 1.6322374163514692, + "flow_matching_loss": 0.03771541640162468, + "grad_norm": 0.48763397336006165, + "learning_rate": 2.8045e-05, + "loss": 0.1054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026945194602012633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014141700603067875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06803208962082863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946274638175965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5610, + "weighted_flow_matching_loss": 0.03771541640162468, + "weighted_rkd_loss": 0.06766656003892421 + }, + { + "epoch": 1.6351469304626127, + "flow_matching_loss": 0.039929192140698434, + "grad_norm": 0.4568731188774109, + "learning_rate": 2.8095e-05, + "loss": 0.1057, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026272975280880927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013570292573422193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06611624360084534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946083128452301, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5620, + "weighted_flow_matching_loss": 0.039929192140698434, + "weighted_rkd_loss": 0.06575978063046932 + }, + { + "epoch": 1.6380564445737562, + "flow_matching_loss": 0.03896003216505051, + "grad_norm": 0.4362732470035553, + "learning_rate": 2.8145000000000002e-05, + "loss": 0.1058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026762381568551062, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013671079464256764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06719584316015244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9945891261100769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5630, + "weighted_flow_matching_loss": 0.03896003216505051, + "weighted_rkd_loss": 0.06683225631713867 + }, + { + "epoch": 1.6409659586848995, + "flow_matching_loss": 0.041521957516670226, + "grad_norm": 0.5091007947921753, + "learning_rate": 2.8195e-05, + "loss": 0.1098, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027261172793805598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014100476074963808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06862282231450081, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9945698976516724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5640, + "weighted_flow_matching_loss": 0.041521957516670226, + "weighted_rkd_loss": 0.06825016923248768 + }, + { + "epoch": 1.6438754727960432, + "flow_matching_loss": 0.04030283764004707, + "grad_norm": 0.4678085148334503, + "learning_rate": 2.8245e-05, + "loss": 0.1066, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026455946266651154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01370058637112379, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06661247871816159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9945506453514099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5650, + "weighted_flow_matching_loss": 0.04030283764004707, + "weighted_rkd_loss": 0.06624946892261505 + }, + { + "epoch": 1.6467849869071864, + "flow_matching_loss": 0.0408238559961319, + "grad_norm": 0.45132213830947876, + "learning_rate": 2.8295000000000003e-05, + "loss": 0.1061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026465552859008312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012733016815036535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06566412150859832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9945313572883606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5660, + "weighted_flow_matching_loss": 0.0408238559961319, + "weighted_rkd_loss": 0.06530504748225212 + }, + { + "epoch": 1.64969450101833, + "flow_matching_loss": 0.041032562777400014, + "grad_norm": 0.4667578339576721, + "learning_rate": 2.8345e-05, + "loss": 0.1045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025216492637991906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013384862244129181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06381784826517105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9945120275020599, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5670, + "weighted_flow_matching_loss": 0.041032562777400014, + "weighted_rkd_loss": 0.06346761174499989 + }, + { + "epoch": 1.6526040151294734, + "flow_matching_loss": 0.040352129563689235, + "grad_norm": 0.49776580929756165, + "learning_rate": 2.8395000000000002e-05, + "loss": 0.1076, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026823949255049228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013950799964368344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06759869940578937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9944926798343658, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5680, + "weighted_flow_matching_loss": 0.040352129563689235, + "weighted_rkd_loss": 0.06722639799118042 + }, + { + "epoch": 1.6555135292406167, + "flow_matching_loss": 0.03884877488017082, + "grad_norm": 0.4620320200920105, + "learning_rate": 2.8445000000000003e-05, + "loss": 0.106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026907638646662235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013710259459912777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06752553693950177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9944732785224915, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5690, + "weighted_flow_matching_loss": 0.03884877488017082, + "weighted_rkd_loss": 0.06715232953429222 + }, + { + "epoch": 1.6584230433517604, + "flow_matching_loss": 0.039375171437859534, + "grad_norm": 0.46438372135162354, + "learning_rate": 2.8495e-05, + "loss": 0.1049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026195803843438625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013513282779604197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06590489037334919, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9944538533687591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5700, + "weighted_flow_matching_loss": 0.039375171437859534, + "weighted_rkd_loss": 0.06553936637938022 + }, + { + "epoch": 1.6613325574629036, + "flow_matching_loss": 0.03947191536426544, + "grad_norm": 0.526942789554596, + "learning_rate": 2.8545000000000002e-05, + "loss": 0.1078, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027362713217735292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01397915231063962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06870457902550697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9944343984127044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5710, + "weighted_flow_matching_loss": 0.03947191536426544, + "weighted_rkd_loss": 0.06832219287753105 + }, + { + "epoch": 1.6642420715740471, + "flow_matching_loss": 0.03991708718240261, + "grad_norm": 0.4749137759208679, + "learning_rate": 2.8595000000000003e-05, + "loss": 0.1039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02553188782185316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013236972037702799, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0643007479608059, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9944149017333984, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5720, + "weighted_flow_matching_loss": 0.03991708718240261, + "weighted_rkd_loss": 0.06394161507487298 + }, + { + "epoch": 1.6671515856851906, + "flow_matching_loss": 0.04383313022553921, + "grad_norm": 0.47167345881462097, + "learning_rate": 2.8645e-05, + "loss": 0.1085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02560295406728983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013821639493107796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06502754800021648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9943953692913056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5730, + "weighted_flow_matching_loss": 0.04383313022553921, + "weighted_rkd_loss": 0.06466309018433095 + }, + { + "epoch": 1.6700610997963339, + "flow_matching_loss": 0.042277519777417186, + "grad_norm": 0.49574756622314453, + "learning_rate": 2.8695000000000002e-05, + "loss": 0.1106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027200651355087757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01427710885182023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06867841146886348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9943758189678192, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5740, + "weighted_flow_matching_loss": 0.042277519777417186, + "weighted_rkd_loss": 0.06829213798046112 + }, + { + "epoch": 1.6729706139074776, + "flow_matching_loss": 0.03996553383767605, + "grad_norm": 0.5029524564743042, + "learning_rate": 2.8745000000000003e-05, + "loss": 0.1061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02636704295873642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013756102789193392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06649018973112106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9943562150001526, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5750, + "weighted_flow_matching_loss": 0.03996553383767605, + "weighted_rkd_loss": 0.06611493788659573 + }, + { + "epoch": 1.6758801280186209, + "flow_matching_loss": 0.04436424933373928, + "grad_norm": 0.4702453017234802, + "learning_rate": 2.8795e-05, + "loss": 0.1135, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027297872118651867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01491205682978034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06950780116021633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9943365752696991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5760, + "weighted_flow_matching_loss": 0.04436424933373928, + "weighted_rkd_loss": 0.06911415196955203 + }, + { + "epoch": 1.6787896421297643, + "flow_matching_loss": 0.041709637828171255, + "grad_norm": 0.4473077952861786, + "learning_rate": 2.8845000000000003e-05, + "loss": 0.1083, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026605736277997494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013793881610035896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06700535528361798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9943169057369232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5770, + "weighted_flow_matching_loss": 0.041709637828171255, + "weighted_rkd_loss": 0.06662454232573509 + }, + { + "epoch": 1.6816991562409078, + "flow_matching_loss": 0.04340220093727112, + "grad_norm": 0.4786553680896759, + "learning_rate": 2.8895e-05, + "loss": 0.1106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02701451871544123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013547690864652396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06757672876119614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9942972302436829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5780, + "weighted_flow_matching_loss": 0.04340220093727112, + "weighted_rkd_loss": 0.06719132959842682 + }, + { + "epoch": 1.684608670352051, + "flow_matching_loss": 0.040213387832045554, + "grad_norm": 0.4425288438796997, + "learning_rate": 2.8945e-05, + "loss": 0.1079, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026929754205048083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01418915381655097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06804866269230843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9942774891853332, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5790, + "weighted_flow_matching_loss": 0.040213387832045554, + "weighted_rkd_loss": 0.06765925474464893 + }, + { + "epoch": 1.6875181844631948, + "flow_matching_loss": 0.0380432054400444, + "grad_norm": 0.5096471905708313, + "learning_rate": 2.8995000000000003e-05, + "loss": 0.1026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0255573283880949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013812456652522087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06492711380124092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9942577362060547, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5800, + "weighted_flow_matching_loss": 0.0380432054400444, + "weighted_rkd_loss": 0.06455427296459675 + }, + { + "epoch": 1.690427698574338, + "flow_matching_loss": 0.04037588723003864, + "grad_norm": 0.46856629848480225, + "learning_rate": 2.9045e-05, + "loss": 0.1088, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027254436910152436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01436185259371996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06887072734534741, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9942379295825958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5810, + "weighted_flow_matching_loss": 0.04037588723003864, + "weighted_rkd_loss": 0.06847390532493591 + }, + { + "epoch": 1.6933372126854815, + "flow_matching_loss": 0.040415994450449945, + "grad_norm": 0.5526347160339355, + "learning_rate": 2.9095000000000002e-05, + "loss": 0.1102, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027718987315893173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01474534198641777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07018331587314605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9942180991172791, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5820, + "weighted_flow_matching_loss": 0.040415994450449945, + "weighted_rkd_loss": 0.06977750808000564 + }, + { + "epoch": 1.696246726796625, + "flow_matching_loss": 0.03782545384019613, + "grad_norm": 0.470735639333725, + "learning_rate": 2.9145000000000003e-05, + "loss": 0.1039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0264679791405797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013563562650233507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06649952009320259, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9941982328891754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5830, + "weighted_flow_matching_loss": 0.03782545384019613, + "weighted_rkd_loss": 0.06611370295286179 + }, + { + "epoch": 1.6991562409077683, + "flow_matching_loss": 0.0407441608607769, + "grad_norm": 0.45182231068611145, + "learning_rate": 2.9195e-05, + "loss": 0.1074, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026615353114902974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013829907402396202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06706061512231827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9941783368587493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5840, + "weighted_flow_matching_loss": 0.0407441608607769, + "weighted_rkd_loss": 0.06667021960020066 + }, + { + "epoch": 1.702065755018912, + "flow_matching_loss": 0.04025957435369491, + "grad_norm": 0.45508140325546265, + "learning_rate": 2.9245000000000002e-05, + "loss": 0.1034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025221443548798562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013033841736614704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06347672902047634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994158399105072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5850, + "weighted_flow_matching_loss": 0.04025957435369491, + "weighted_rkd_loss": 0.06310592144727707 + }, + { + "epoch": 1.7049752691300553, + "flow_matching_loss": 0.03854752667248249, + "grad_norm": 0.43691810965538025, + "learning_rate": 2.9295000000000003e-05, + "loss": 0.1096, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028489620611071586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014454633183777332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07143387347459793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9941384315490722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5860, + "weighted_flow_matching_loss": 0.03854752667248249, + "weighted_rkd_loss": 0.07101517170667648 + }, + { + "epoch": 1.7078847832411987, + "flow_matching_loss": 0.041477135941386224, + "grad_norm": 0.4218618869781494, + "learning_rate": 2.9345e-05, + "loss": 0.1081, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026834535412490368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013331872131675481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06700094267725945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9941184282302856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5870, + "weighted_flow_matching_loss": 0.041477135941386224, + "weighted_rkd_loss": 0.06660687774419785 + }, + { + "epoch": 1.7107942973523422, + "flow_matching_loss": 0.041637174785137177, + "grad_norm": 0.5619790554046631, + "learning_rate": 2.9395000000000002e-05, + "loss": 0.1061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025878177769482136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013088714610785247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06484507024288177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9940983951091766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5880, + "weighted_flow_matching_loss": 0.041637174785137177, + "weighted_rkd_loss": 0.06446235552430153 + }, + { + "epoch": 1.7137038114634855, + "flow_matching_loss": 0.04194956570863724, + "grad_norm": 0.49991023540496826, + "learning_rate": 2.9445000000000004e-05, + "loss": 0.1042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02484878171235323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012884426116943359, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0625819906592369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9940783262252808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5890, + "weighted_flow_matching_loss": 0.04194956570863724, + "weighted_rkd_loss": 0.06221141144633293 + }, + { + "epoch": 1.7166133255746292, + "flow_matching_loss": 0.041554175689816474, + "grad_norm": 0.5017691254615784, + "learning_rate": 2.9495e-05, + "loss": 0.1061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025829938240349293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013306155428290367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06496603190898895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994058221578598, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5900, + "weighted_flow_matching_loss": 0.041554175689816474, + "weighted_rkd_loss": 0.0645800456404686 + }, + { + "epoch": 1.7195228396857725, + "flow_matching_loss": 0.04087883569300175, + "grad_norm": 0.4792731702327728, + "learning_rate": 2.9545000000000003e-05, + "loss": 0.1065, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026077096350491047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013872701209038496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06602689437568188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9940380811691284, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5910, + "weighted_flow_matching_loss": 0.04087883569300175, + "weighted_rkd_loss": 0.06563325300812721 + }, + { + "epoch": 1.722432353796916, + "flow_matching_loss": 0.04063737392425537, + "grad_norm": 0.46532851457595825, + "learning_rate": 2.9595e-05, + "loss": 0.1064, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026409338228404522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013298335298895837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06611701212823391, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994017916917801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5920, + "weighted_flow_matching_loss": 0.04063737392425537, + "weighted_rkd_loss": 0.06572148874402046 + }, + { + "epoch": 1.7253418679080594, + "flow_matching_loss": 0.04264799989759922, + "grad_norm": 0.47779080271720886, + "learning_rate": 2.9644999999999998e-05, + "loss": 0.1066, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02539948411285877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013578211702406407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06437717974185944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9939976990222931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5930, + "weighted_flow_matching_loss": 0.04264799989759922, + "weighted_rkd_loss": 0.06399075202643871 + }, + { + "epoch": 1.7282513820192027, + "flow_matching_loss": 0.042229699715971944, + "grad_norm": 0.518882155418396, + "learning_rate": 2.9695e-05, + "loss": 0.104, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024948061257600785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012220094818621874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.062116217240691184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9939774870872498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5940, + "weighted_flow_matching_loss": 0.042229699715971944, + "weighted_rkd_loss": 0.061742125824093816 + }, + { + "epoch": 1.7311608961303462, + "flow_matching_loss": 0.040656183287501334, + "grad_norm": 0.424621045589447, + "learning_rate": 2.9745e-05, + "loss": 0.1004, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023801893182098865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012461141310632229, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060064927116036414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9939572036266326, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5950, + "weighted_flow_matching_loss": 0.040656183287501334, + "weighted_rkd_loss": 0.05970196016132832 + }, + { + "epoch": 1.7340704102414897, + "flow_matching_loss": 0.04062037169933319, + "grad_norm": 0.46885645389556885, + "learning_rate": 2.9795e-05, + "loss": 0.1065, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026340895891189577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013553119357675314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06623491160571575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9939368903636933, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5960, + "weighted_flow_matching_loss": 0.04062037169933319, + "weighted_rkd_loss": 0.06583332791924476 + }, + { + "epoch": 1.7369799243526332, + "flow_matching_loss": 0.04245202019810677, + "grad_norm": 0.43663474917411804, + "learning_rate": 2.9845e-05, + "loss": 0.1071, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025762397982180118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01349018830806017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06501498483121396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9939165472984314, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5970, + "weighted_flow_matching_loss": 0.04245202019810677, + "weighted_rkd_loss": 0.06461947709321976 + }, + { + "epoch": 1.7398894384637766, + "flow_matching_loss": 0.03909266609698534, + "grad_norm": 0.45264652371406555, + "learning_rate": 2.9895e-05, + "loss": 0.1057, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02656238228082657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013914895243942738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06703966148197651, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9938961744308472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5980, + "weighted_flow_matching_loss": 0.03909266609698534, + "weighted_rkd_loss": 0.06663048155605793 + }, + { + "epoch": 1.74279895257492, + "flow_matching_loss": 0.03967259619385004, + "grad_norm": 0.45394426584243774, + "learning_rate": 2.9945e-05, + "loss": 0.1029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025154765509068967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013263756036758423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06357328668236732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9938757658004761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5990, + "weighted_flow_matching_loss": 0.03967259619385004, + "weighted_rkd_loss": 0.06318395622074605 + }, + { + "epoch": 1.7457084666860634, + "flow_matching_loss": 0.03863349631428718, + "grad_norm": 0.44977232813835144, + "learning_rate": 2.9995e-05, + "loss": 0.103, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02571549843996763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013300218153744936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06473121531307698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9938553273677826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6000, + "weighted_flow_matching_loss": 0.03863349631428718, + "weighted_rkd_loss": 0.0643334586173296 + }, + { + "epoch": 1.7486179807972069, + "flow_matching_loss": 0.03950294852256775, + "grad_norm": 0.4413333833217621, + "learning_rate": 2.999999953864384e-05, + "loss": 0.104, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025561483949422835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013778280280530453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0649012491106987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9938348531723022, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6010, + "weighted_flow_matching_loss": 0.03950294852256775, + "weighted_rkd_loss": 0.06450114361941814 + }, + { + "epoch": 1.7515274949083504, + "flow_matching_loss": 0.04095322340726852, + "grad_norm": 0.44586947560310364, + "learning_rate": 2.9999997943832465e-05, + "loss": 0.1023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024482170678675176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012771477829664946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061735819280147555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9938143372535706, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6020, + "weighted_flow_matching_loss": 0.04095322340726852, + "weighted_rkd_loss": 0.061353910341858865 + }, + { + "epoch": 1.7544370090194938, + "flow_matching_loss": 0.03815806694328785, + "grad_norm": 0.44891318678855896, + "learning_rate": 2.9999995209870235e-05, + "loss": 0.1019, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02569618560373783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01277390606701374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0641662772744894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993793785572052, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6030, + "weighted_flow_matching_loss": 0.03815806694328785, + "weighted_rkd_loss": 0.06376807428896428 + }, + { + "epoch": 1.757346523130637, + "flow_matching_loss": 0.03926766142249107, + "grad_norm": 0.4191233515739441, + "learning_rate": 2.9999991336757358e-05, + "loss": 0.1084, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027698160149157047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014124190714210273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06952050998806954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99377321600914, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6040, + "weighted_flow_matching_loss": 0.03926766142249107, + "weighted_rkd_loss": 0.06908762790262699 + }, + { + "epoch": 1.7602560372417806, + "flow_matching_loss": 0.042339741438627246, + "grad_norm": 0.5369196534156799, + "learning_rate": 2.9999986324494133e-05, + "loss": 0.1072, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02588445246219635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01354572493582964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06531463041901589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9937526047229767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6050, + "weighted_flow_matching_loss": 0.042339741438627246, + "weighted_rkd_loss": 0.06490657478570938 + }, + { + "epoch": 1.763165551352924, + "flow_matching_loss": 0.03760665860027075, + "grad_norm": 0.4513804614543915, + "learning_rate": 2.9999980173080936e-05, + "loss": 0.0982, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024241629615426065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012517157010734081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061000417172908786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9937319576740264, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6060, + "weighted_flow_matching_loss": 0.03760665860027075, + "weighted_rkd_loss": 0.06061805859208107 + }, + { + "epoch": 1.7660750654640673, + "flow_matching_loss": 0.04336174540221691, + "grad_norm": 0.48404815793037415, + "learning_rate": 2.999997288251824e-05, + "loss": 0.1089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02632270324975252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013311832677572966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0659572396427393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993711280822754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6070, + "weighted_flow_matching_loss": 0.04336174540221691, + "weighted_rkd_loss": 0.06554246172308922 + }, + { + "epoch": 1.768984579575211, + "flow_matching_loss": 0.04077918715775013, + "grad_norm": 0.4448257386684418, + "learning_rate": 2.999996445280659e-05, + "loss": 0.1042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025530119985342027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012781684473156928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06384192369878292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993690574169159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6080, + "weighted_flow_matching_loss": 0.04077918715775013, + "weighted_rkd_loss": 0.06343912221491337 + }, + { + "epoch": 1.7718940936863543, + "flow_matching_loss": 0.03968315795063972, + "grad_norm": 0.4569772183895111, + "learning_rate": 2.9999954883946635e-05, + "loss": 0.1047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025896035507321356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013677684590220451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0654697559773922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9936698198318481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6090, + "weighted_flow_matching_loss": 0.03968315795063972, + "weighted_rkd_loss": 0.06505532190203667 + }, + { + "epoch": 1.7748036077974978, + "flow_matching_loss": 0.04079181626439095, + "grad_norm": 0.4769936203956604, + "learning_rate": 2.9999944175939098e-05, + "loss": 0.106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026040188781917096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013564947061240673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06564532406628132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9936490416526794, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6100, + "weighted_flow_matching_loss": 0.04079181626439095, + "weighted_rkd_loss": 0.06522841565310955 + }, + { + "epoch": 1.7777131219086413, + "flow_matching_loss": 0.03715069908648729, + "grad_norm": 0.5566121339797974, + "learning_rate": 2.9999932328784788e-05, + "loss": 0.1042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026835948415100574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013795451074838639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06746734902262688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9936282336711884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6110, + "weighted_flow_matching_loss": 0.03715069908648729, + "weighted_rkd_loss": 0.06703744754195214 + }, + { + "epoch": 1.7806226360197845, + "flow_matching_loss": 0.040339451655745505, + "grad_norm": 0.4844239056110382, + "learning_rate": 2.9999919342484612e-05, + "loss": 0.1027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024980498664081096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012751814629882573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06271281130611897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993607372045517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6120, + "weighted_flow_matching_loss": 0.040339451655745505, + "weighted_rkd_loss": 0.06231190040707588 + }, + { + "epoch": 1.7835321501309283, + "flow_matching_loss": 0.03781597018241882, + "grad_norm": 0.4486694037914276, + "learning_rate": 2.9999905217039546e-05, + "loss": 0.1003, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024877018481492996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013167391251772641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06292142793536186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9935865044593811, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6130, + "weighted_flow_matching_loss": 0.03781597018241882, + "weighted_rkd_loss": 0.06251787580549717 + }, + { + "epoch": 1.7864416642420715, + "flow_matching_loss": 0.042795583605766296, + "grad_norm": 0.4765702486038208, + "learning_rate": 2.9999889952450677e-05, + "loss": 0.1072, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025734212063252927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013353019673377275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06482144296169282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9935655772686005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6140, + "weighted_flow_matching_loss": 0.042795583605766296, + "weighted_rkd_loss": 0.06440438032150268 + }, + { + "epoch": 1.789351178353215, + "flow_matching_loss": 0.03840897716581822, + "grad_norm": 0.41334664821624756, + "learning_rate": 2.9999873548719155e-05, + "loss": 0.1, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02443597838282585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013079178612679243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06195113509893417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9935446381568909, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6150, + "weighted_flow_matching_loss": 0.03840897716581822, + "weighted_rkd_loss": 0.06155120804905891 + }, + { + "epoch": 1.7922606924643585, + "flow_matching_loss": 0.03762277215719223, + "grad_norm": 0.4255625605583191, + "learning_rate": 2.999985600584623e-05, + "loss": 0.0988, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024465718492865562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012662813626229763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06159425005316734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9935236513614655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6160, + "weighted_flow_matching_loss": 0.03762277215719223, + "weighted_rkd_loss": 0.061195332929491995 + }, + { + "epoch": 1.7951702065755017, + "flow_matching_loss": 0.03684165142476559, + "grad_norm": 0.44022995233535767, + "learning_rate": 2.9999837323833227e-05, + "loss": 0.0946, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023231475986540316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011622891388833522, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.058085843548178674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9935026288032531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6170, + "weighted_flow_matching_loss": 0.03684165142476559, + "weighted_rkd_loss": 0.057708435133099555 + }, + { + "epoch": 1.7980797206866455, + "flow_matching_loss": 0.03930737506598234, + "grad_norm": 0.47530990839004517, + "learning_rate": 2.999981750268157e-05, + "loss": 0.1, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024256017245352268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012595776189118623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061107811704277995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9934815764427185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6180, + "weighted_flow_matching_loss": 0.03930737506598234, + "weighted_rkd_loss": 0.060709493979811666 + }, + { + "epoch": 1.8009892347977887, + "flow_matching_loss": 0.03686355464160442, + "grad_norm": 0.38552916049957275, + "learning_rate": 2.9999796542392766e-05, + "loss": 0.0987, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02493463847786188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012398557364940643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06226783394813538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9934604942798615, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6190, + "weighted_flow_matching_loss": 0.03686355464160442, + "weighted_rkd_loss": 0.06186064891517162 + }, + { + "epoch": 1.8038987489089322, + "flow_matching_loss": 0.03909970968961716, + "grad_norm": 0.46365803480148315, + "learning_rate": 2.99997744429684e-05, + "loss": 0.1005, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024723799526691438, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01239138888195157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061838988214731216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9934393763542175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6200, + "weighted_flow_matching_loss": 0.03909970968961716, + "weighted_rkd_loss": 0.06143326871097088 + }, + { + "epoch": 1.8068082630200757, + "flow_matching_loss": 0.037398047558963296, + "grad_norm": 0.4582616090774536, + "learning_rate": 2.9999751204410165e-05, + "loss": 0.098, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02459098305553198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011833081860095263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061015046760439874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9934182167053223, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6210, + "weighted_flow_matching_loss": 0.037398047558963296, + "weighted_rkd_loss": 0.06061345376074314 + }, + { + "epoch": 1.809717777131219, + "flow_matching_loss": 0.0395655382424593, + "grad_norm": 0.46919840574264526, + "learning_rate": 2.9999726826719808e-05, + "loss": 0.101, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0245030777528882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012857572361826896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06186372824013233, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9933970391750335, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6220, + "weighted_flow_matching_loss": 0.0395655382424593, + "weighted_rkd_loss": 0.061455241218209264 + }, + { + "epoch": 1.8126272912423627, + "flow_matching_loss": 0.03741294890642166, + "grad_norm": 0.4719553291797638, + "learning_rate": 2.9999701309899187e-05, + "loss": 0.1022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026018952205777168, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013197520934045314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0652354259043932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9933758199214935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6230, + "weighted_flow_matching_loss": 0.03741294890642166, + "weighted_rkd_loss": 0.06480330154299736 + }, + { + "epoch": 1.815536805353506, + "flow_matching_loss": 0.03832711689174175, + "grad_norm": 0.3915034234523773, + "learning_rate": 2.9999674653950248e-05, + "loss": 0.1002, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02480060402303934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012646529916673898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06224773824214935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9933545649051666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6240, + "weighted_flow_matching_loss": 0.03832711689174175, + "weighted_rkd_loss": 0.06183407343924045 + }, + { + "epoch": 1.8184463194646494, + "flow_matching_loss": 0.03855853490531445, + "grad_norm": 0.43505507707595825, + "learning_rate": 2.9999646858875002e-05, + "loss": 0.0996, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02456670217216015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01235075257718563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061484156921505925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9933332800865173, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6250, + "weighted_flow_matching_loss": 0.03855853490531445, + "weighted_rkd_loss": 0.06107423864305019 + }, + { + "epoch": 1.821355833575793, + "flow_matching_loss": 0.038575905561447146, + "grad_norm": 0.45986729860305786, + "learning_rate": 2.9999617924675568e-05, + "loss": 0.1032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025839832797646522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013389159366488457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06506882570683956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9933119595050812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6260, + "weighted_flow_matching_loss": 0.038575905561447146, + "weighted_rkd_loss": 0.06463364697992802 + }, + { + "epoch": 1.8242653476869362, + "flow_matching_loss": 0.0366479204967618, + "grad_norm": 0.4338770806789398, + "learning_rate": 2.999958785135414e-05, + "loss": 0.1001, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02533418908715248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013255346659570932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06392372362315654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9932905972003937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6270, + "weighted_flow_matching_loss": 0.0366479204967618, + "weighted_rkd_loss": 0.06349483095109462 + }, + { + "epoch": 1.8271748617980799, + "flow_matching_loss": 0.03952050507068634, + "grad_norm": 0.4560433626174927, + "learning_rate": 2.9999556638913004e-05, + "loss": 0.0988, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023892238922417163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011935286968946458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.059719765186309816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9932692050933838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6280, + "weighted_flow_matching_loss": 0.03952050507068634, + "weighted_rkd_loss": 0.059317806735634804 + }, + { + "epoch": 1.8300843759092231, + "flow_matching_loss": 0.03772878460586071, + "grad_norm": 0.3998326063156128, + "learning_rate": 2.999952428735453e-05, + "loss": 0.0945, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02282879054546356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011524152755737305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057181733474135396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993247801065445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6290, + "weighted_flow_matching_loss": 0.03772878460586071, + "weighted_rkd_loss": 0.056795616075396536 + }, + { + "epoch": 1.8329938900203666, + "flow_matching_loss": 0.03969245404005051, + "grad_norm": 0.46490341424942017, + "learning_rate": 2.9999490796681178e-05, + "loss": 0.0971, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023169916681945325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011418310087174177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05775814391672611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9932263493537903, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6300, + "weighted_flow_matching_loss": 0.03969245404005051, + "weighted_rkd_loss": 0.05736690647900104 + }, + { + "epoch": 1.83590340413151, + "flow_matching_loss": 0.03956506438553333, + "grad_norm": 0.44290101528167725, + "learning_rate": 2.9999456166895485e-05, + "loss": 0.1013, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0246149942278862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012956677563488483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06218666583299637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9932048320770264, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6310, + "weighted_flow_matching_loss": 0.03956506438553333, + "weighted_rkd_loss": 0.061764099448919293 + }, + { + "epoch": 1.8388129182426534, + "flow_matching_loss": 0.03831705749034882, + "grad_norm": 0.47608673572540283, + "learning_rate": 2.999942039800008e-05, + "loss": 0.0982, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024158140085637568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011958973202854394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060275253280997275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993183308839798, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6320, + "weighted_flow_matching_loss": 0.03831705749034882, + "weighted_rkd_loss": 0.05986436493694782 + }, + { + "epoch": 1.841722432353797, + "flow_matching_loss": 0.04116914141923189, + "grad_norm": 0.4636126756668091, + "learning_rate": 2.999938348999769e-05, + "loss": 0.0985, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023018734902143477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011729050613939761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05776652060449124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9931617558002472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6330, + "weighted_flow_matching_loss": 0.04116914141923189, + "weighted_rkd_loss": 0.05737150423228741 + }, + { + "epoch": 1.8446319464649403, + "flow_matching_loss": 0.03724840600043535, + "grad_norm": 0.5083436369895935, + "learning_rate": 2.999934544289111e-05, + "loss": 0.0977, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02426362633705139, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012350485939532519, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060877738520503044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9931401610374451, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6340, + "weighted_flow_matching_loss": 0.03724840600043535, + "weighted_rkd_loss": 0.060460123047232625 + }, + { + "epoch": 1.8475414605760838, + "flow_matching_loss": 0.03968631625175476, + "grad_norm": 0.4334544539451599, + "learning_rate": 2.9999306256683227e-05, + "loss": 0.1013, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02467487696558237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012666721735149622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06201647520065308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9931185364723205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6350, + "weighted_flow_matching_loss": 0.03968631625175476, + "weighted_rkd_loss": 0.061589712649583815 + }, + { + "epoch": 1.8504509746872273, + "flow_matching_loss": 0.040991802141070365, + "grad_norm": 0.42212769389152527, + "learning_rate": 2.9999265931377026e-05, + "loss": 0.1028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02486097738146782, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012555733788758516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06227768957614899, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9930968761444092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6360, + "weighted_flow_matching_loss": 0.040991802141070365, + "weighted_rkd_loss": 0.061847792193293574 + }, + { + "epoch": 1.8533604887983706, + "flow_matching_loss": 0.03847973272204399, + "grad_norm": 0.46970874071121216, + "learning_rate": 2.999922446697556e-05, + "loss": 0.0993, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024617513827979564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012014780566096306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06124980822205543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9930751860141754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6370, + "weighted_flow_matching_loss": 0.03847973272204399, + "weighted_rkd_loss": 0.06082566678524017 + }, + { + "epoch": 1.8562700029095143, + "flow_matching_loss": 0.03820000514388085, + "grad_norm": 0.49185606837272644, + "learning_rate": 2.9999181863481982e-05, + "loss": 0.0945, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02249791193753481, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011718188971281051, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05671401284635067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9930534601211548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6380, + "weighted_flow_matching_loss": 0.03820000514388085, + "weighted_rkd_loss": 0.056320051476359365 + }, + { + "epoch": 1.8591795170206575, + "flow_matching_loss": 0.03907622992992401, + "grad_norm": 0.578765332698822, + "learning_rate": 2.999913812089953e-05, + "loss": 0.099, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024058481119573115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012191412318497897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060308373719453814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9930316984653473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6390, + "weighted_flow_matching_loss": 0.03907622992992401, + "weighted_rkd_loss": 0.059888112172484395 + }, + { + "epoch": 1.862089031131801, + "flow_matching_loss": 0.040074772387743, + "grad_norm": 0.499105304479599, + "learning_rate": 2.9999093239231523e-05, + "loss": 0.0979, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023313805088400842, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011568554677069187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05819616541266441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9930099070072174, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6400, + "weighted_flow_matching_loss": 0.040074772387743, + "weighted_rkd_loss": 0.05778936371207237 + }, + { + "epoch": 1.8649985452429445, + "flow_matching_loss": 0.037907208874821664, + "grad_norm": 0.4408312737941742, + "learning_rate": 2.9999047218481365e-05, + "loss": 0.1012, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025207454338669777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013282082695513963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06369699127972125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9929880678653717, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6410, + "weighted_flow_matching_loss": 0.037907208874821664, + "weighted_rkd_loss": 0.06325037106871605 + }, + { + "epoch": 1.8679080593540878, + "flow_matching_loss": 0.0388478834182024, + "grad_norm": 0.4274986982345581, + "learning_rate": 2.9999000058652565e-05, + "loss": 0.1017, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025050492584705354, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013213674072176218, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06331465877592564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9929662108421325, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6420, + "weighted_flow_matching_loss": 0.0388478834182024, + "weighted_rkd_loss": 0.06286931671202182 + }, + { + "epoch": 1.8708175734652313, + "flow_matching_loss": 0.041455207765102385, + "grad_norm": 0.4496871531009674, + "learning_rate": 2.9998951759748687e-05, + "loss": 0.1012, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024110378138720988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011930449865758419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06015120707452297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9929443180561066, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6430, + "weighted_flow_matching_loss": 0.041455207765102385, + "weighted_rkd_loss": 0.05972678139805794 + }, + { + "epoch": 1.8737270875763747, + "flow_matching_loss": 0.038104334846138954, + "grad_norm": 0.4346208870410919, + "learning_rate": 2.999890232177341e-05, + "loss": 0.0988, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02446536663919687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01215802002698183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061088753119111064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9929223835468293, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6440, + "weighted_flow_matching_loss": 0.038104334846138954, + "weighted_rkd_loss": 0.06065637767314911 + }, + { + "epoch": 1.8766366016875182, + "flow_matching_loss": 0.038983119279146196, + "grad_norm": 0.3865571618080139, + "learning_rate": 2.9998851744730487e-05, + "loss": 0.0953, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02258184663951397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01158563932403922, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05674933269619942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.992900425195694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6450, + "weighted_flow_matching_loss": 0.038983119279146196, + "weighted_rkd_loss": 0.056346450746059415 + }, + { + "epoch": 1.8795461157986617, + "flow_matching_loss": 0.036757560819387435, + "grad_norm": 0.4112933874130249, + "learning_rate": 2.9998800028623755e-05, + "loss": 0.0954, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02349663469940424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012113811820745468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.059107080847024915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9928784132003784, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6460, + "weighted_flow_matching_loss": 0.036757560819387435, + "weighted_rkd_loss": 0.05868613123893738 + }, + { + "epoch": 1.882455629909805, + "flow_matching_loss": 0.038433685526251794, + "grad_norm": 0.4708874821662903, + "learning_rate": 2.9998747173457145e-05, + "loss": 0.0992, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024436602741479872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012299058865755797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0611722644418478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9928564131259918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6470, + "weighted_flow_matching_loss": 0.038433685526251794, + "weighted_rkd_loss": 0.06073525883257389 + }, + { + "epoch": 1.8853651440209485, + "flow_matching_loss": 0.03813743069767952, + "grad_norm": 0.38441330194473267, + "learning_rate": 2.9998693179234674e-05, + "loss": 0.0993, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024639768339693547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01236974410712719, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061649280786514285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9928343296051025, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6480, + "weighted_flow_matching_loss": 0.03813743069767952, + "weighted_rkd_loss": 0.06120753400027752 + }, + { + "epoch": 1.888274658132092, + "flow_matching_loss": 0.03802985716611147, + "grad_norm": 0.4616728127002716, + "learning_rate": 2.9998638045960433e-05, + "loss": 0.0986, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02430623359978199, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012350178975611924, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06096264570951462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9928122282028198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6490, + "weighted_flow_matching_loss": 0.03802985716611147, + "weighted_rkd_loss": 0.06052445396780968 + }, + { + "epoch": 1.8911841722432354, + "flow_matching_loss": 0.03789335917681456, + "grad_norm": 0.42849001288414, + "learning_rate": 2.9998581773638622e-05, + "loss": 0.1, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025124169513583185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012314007431268693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06256234720349312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9927900969982147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6500, + "weighted_flow_matching_loss": 0.03789335917681456, + "weighted_rkd_loss": 0.062111296504735944 + }, + { + "epoch": 1.894093686354379, + "flow_matching_loss": 0.0388945247977972, + "grad_norm": 0.3944694995880127, + "learning_rate": 2.9998524362273504e-05, + "loss": 0.0985, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02387033384293318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01228398410603404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06002465188503266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9927679240703583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6510, + "weighted_flow_matching_loss": 0.0388945247977972, + "weighted_rkd_loss": 0.05959054194390774 + }, + { + "epoch": 1.8970032004655222, + "flow_matching_loss": 0.038153485581278804, + "grad_norm": 0.5015727877616882, + "learning_rate": 2.9998465811869437e-05, + "loss": 0.096, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023143769055604935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011934261117130519, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05822180025279522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9927457273006439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6520, + "weighted_flow_matching_loss": 0.038153485581278804, + "weighted_rkd_loss": 0.05779944062232971 + }, + { + "epoch": 1.8999127145766657, + "flow_matching_loss": 0.037144382297992704, + "grad_norm": 0.4395347833633423, + "learning_rate": 2.9998406122430878e-05, + "loss": 0.0916, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021918125823140143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010993257723748684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05482950955629349, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9927234947681427, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6530, + "weighted_flow_matching_loss": 0.037144382297992704, + "weighted_rkd_loss": 0.05443053878843784 + }, + { + "epoch": 1.9028222286878091, + "flow_matching_loss": 0.03888048343360424, + "grad_norm": 0.43833085894584656, + "learning_rate": 2.9998345293962358e-05, + "loss": 0.1008, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024685210175812246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012963362224400043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.062333783507347106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9927012264728546, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6540, + "weighted_flow_matching_loss": 0.03888048343360424, + "weighted_rkd_loss": 0.061878835409879686 + }, + { + "epoch": 1.9057317427989524, + "flow_matching_loss": 0.03741937279701233, + "grad_norm": 0.44044411182403564, + "learning_rate": 2.9998283326468487e-05, + "loss": 0.1009, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025483552366495132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01299546118825674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06396256648004055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9926789224147796, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6550, + "weighted_flow_matching_loss": 0.03741937279701233, + "weighted_rkd_loss": 0.06349429897964001 + }, + { + "epoch": 1.9086412569100961, + "flow_matching_loss": 0.04041320905089378, + "grad_norm": 0.4260371923446655, + "learning_rate": 2.999822021995398e-05, + "loss": 0.0935, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021326327323913576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010835910961031914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053488565236330034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9926565885543823, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6560, + "weighted_flow_matching_loss": 0.04041320905089378, + "weighted_rkd_loss": 0.05309578962624073 + }, + { + "epoch": 1.9115507710212394, + "flow_matching_loss": 0.03631091713905334, + "grad_norm": 0.45540180802345276, + "learning_rate": 2.999815597442363e-05, + "loss": 0.0937, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023217508010566233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011383616644889117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0578186322003603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9926342248916626, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6570, + "weighted_flow_matching_loss": 0.03631091713905334, + "weighted_rkd_loss": 0.057392753660678864 + }, + { + "epoch": 1.9144602851323829, + "flow_matching_loss": 0.03891849797219038, + "grad_norm": 0.5326258540153503, + "learning_rate": 2.999809058988231e-05, + "loss": 0.1014, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02494589500129223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013046766258776188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0629385557025671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9926118195056916, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6580, + "weighted_flow_matching_loss": 0.03891849797219038, + "weighted_rkd_loss": 0.062473561987280844 + }, + { + "epoch": 1.9173697992435264, + "flow_matching_loss": 0.038607310876250266, + "grad_norm": 0.448886513710022, + "learning_rate": 2.9998024066334984e-05, + "loss": 0.0982, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023990726470947264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01205930495634675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06004075780510902, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9925893843173981, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6590, + "weighted_flow_matching_loss": 0.038607310876250266, + "weighted_rkd_loss": 0.05959580391645432 + }, + { + "epoch": 1.9202793133546696, + "flow_matching_loss": 0.037878958880901335, + "grad_norm": 0.48559868335723877, + "learning_rate": 2.9997956403786718e-05, + "loss": 0.0949, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022993313148617744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011455552838742732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05744217894971371, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9925669074058533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6600, + "weighted_flow_matching_loss": 0.037878958880901335, + "weighted_rkd_loss": 0.05701521299779415 + }, + { + "epoch": 1.9231888274658133, + "flow_matching_loss": 0.037378919683396815, + "grad_norm": 0.4342787563800812, + "learning_rate": 2.9997887602242635e-05, + "loss": 0.0945, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022880566120147706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01175599666312337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057517129182815555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9925444066524506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6610, + "weighted_flow_matching_loss": 0.037378919683396815, + "weighted_rkd_loss": 0.05708829164505005 + }, + { + "epoch": 1.9260983415769566, + "flow_matching_loss": 0.03715268634259701, + "grad_norm": 0.46533313393592834, + "learning_rate": 2.999781766170797e-05, + "loss": 0.0946, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023071508668363095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01176702519878745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057910042628645894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9925218641757965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6620, + "weighted_flow_matching_loss": 0.03715268634259701, + "weighted_rkd_loss": 0.05747696682810784 + }, + { + "epoch": 1.9290078556881, + "flow_matching_loss": 0.040160923078656194, + "grad_norm": 0.42283979058265686, + "learning_rate": 2.9997746582188027e-05, + "loss": 0.0945, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021937506273388863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010893302410840989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05476831533014774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9924992978572845, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6630, + "weighted_flow_matching_loss": 0.040160923078656194, + "weighted_rkd_loss": 0.05435751304030419 + }, + { + "epoch": 1.9319173697992436, + "flow_matching_loss": 0.03582555539906025, + "grad_norm": 0.4944574236869812, + "learning_rate": 2.9997674363688207e-05, + "loss": 0.0944, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02365888748317957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011718966532498598, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05903674103319645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9924767017364502, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6640, + "weighted_flow_matching_loss": 0.03582555539906025, + "weighted_rkd_loss": 0.05859256163239479 + }, + { + "epoch": 1.9348268839103868, + "flow_matching_loss": 0.039623007923364637, + "grad_norm": 0.4056139886379242, + "learning_rate": 2.9997601006213998e-05, + "loss": 0.0948, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02227275222539902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01101962043903768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05556512400507927, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9924540519714355, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6650, + "weighted_flow_matching_loss": 0.039623007923364637, + "weighted_rkd_loss": 0.05514582134783268 + }, + { + "epoch": 1.9377363980215305, + "flow_matching_loss": 0.037799007073044774, + "grad_norm": 0.47791793942451477, + "learning_rate": 2.9997526509770963e-05, + "loss": 0.0985, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024363812431693076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012470508553087711, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06119813323020935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9924313843250274, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6660, + "weighted_flow_matching_loss": 0.037799007073044774, + "weighted_rkd_loss": 0.06073494479060173 + }, + { + "epoch": 1.9406459121326738, + "flow_matching_loss": 0.039032533206045626, + "grad_norm": 0.46571415662765503, + "learning_rate": 2.999745087436477e-05, + "loss": 0.0946, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0222476739436388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011530534829944372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056025882437825206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9924086689949035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6670, + "weighted_flow_matching_loss": 0.039032533206045626, + "weighted_rkd_loss": 0.055600560083985326 + }, + { + "epoch": 1.9435554262438173, + "flow_matching_loss": 0.040359489247202875, + "grad_norm": 0.46602413058280945, + "learning_rate": 2.9997374100001158e-05, + "loss": 0.0961, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02233085539191961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011511293239891529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0561730045825243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9923859357833862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6680, + "weighted_flow_matching_loss": 0.040359489247202875, + "weighted_rkd_loss": 0.0557453028857708 + }, + { + "epoch": 1.9464649403549608, + "flow_matching_loss": 0.03846383262425661, + "grad_norm": 0.4364841878414154, + "learning_rate": 2.999729618668595e-05, + "loss": 0.0932, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021999271772801875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01114572212100029, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05514426566660404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9923631548881531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6690, + "weighted_flow_matching_loss": 0.03846383262425661, + "weighted_rkd_loss": 0.05472313389182091 + }, + { + "epoch": 1.949374454466104, + "flow_matching_loss": 0.037705046311020854, + "grad_norm": 0.40721049904823303, + "learning_rate": 2.9997217134425082e-05, + "loss": 0.0973, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023851054534316064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012303061876446008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060005170106887815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.992340350151062, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6700, + "weighted_flow_matching_loss": 0.037705046311020854, + "weighted_rkd_loss": 0.0595455314964056 + }, + { + "epoch": 1.9522839685772477, + "flow_matching_loss": 0.037348099052906036, + "grad_norm": 0.44735267758369446, + "learning_rate": 2.9997136943224536e-05, + "loss": 0.0953, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023449705354869364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011478881351649761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05837829187512398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.992317509651184, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6710, + "weighted_flow_matching_loss": 0.037348099052906036, + "weighted_rkd_loss": 0.0579298123717308 + }, + { + "epoch": 1.955193482688391, + "flow_matching_loss": 0.039927102625370026, + "grad_norm": 0.4613598585128784, + "learning_rate": 2.9997055613090414e-05, + "loss": 0.0993, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023932132869958878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011983673553913832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.059847939386963846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9922946333885193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6720, + "weighted_flow_matching_loss": 0.039927102625370026, + "weighted_rkd_loss": 0.059386759996414185 + }, + { + "epoch": 1.9581029967995345, + "flow_matching_loss": 0.0359384099021554, + "grad_norm": 0.4021774232387543, + "learning_rate": 2.999697314402889e-05, + "loss": 0.0933, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0230465529486537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011666928697377444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05776003524661064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9922717213630676, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6730, + "weighted_flow_matching_loss": 0.0359384099021554, + "weighted_rkd_loss": 0.057313647121191025 + }, + { + "epoch": 1.961012510910678, + "flow_matching_loss": 0.03956333100795746, + "grad_norm": 0.4601964056491852, + "learning_rate": 2.999688953604623e-05, + "loss": 0.0936, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021705724485218524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011002559307962656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05441400818526745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9922487735748291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6740, + "weighted_flow_matching_loss": 0.03956333100795746, + "weighted_rkd_loss": 0.05399221889674664 + }, + { + "epoch": 1.9639220250218212, + "flow_matching_loss": 0.03826456144452095, + "grad_norm": 0.4871009290218353, + "learning_rate": 2.9996804789148777e-05, + "loss": 0.0925, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02177297677844763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011091974005103112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05463792756199837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9922258079051971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6750, + "weighted_flow_matching_loss": 0.03826456144452095, + "weighted_rkd_loss": 0.05421314872801304 + }, + { + "epoch": 1.966831539132965, + "flow_matching_loss": 0.03841538019478321, + "grad_norm": 0.4903894066810608, + "learning_rate": 2.9996718903342972e-05, + "loss": 0.0946, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022806429862976076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010965165961533786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056578025594353674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9922027945518493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6760, + "weighted_flow_matching_loss": 0.03841538019478321, + "weighted_rkd_loss": 0.0561368603259325 + }, + { + "epoch": 1.9697410532441082, + "flow_matching_loss": 0.03691608905792236, + "grad_norm": 0.47264090180397034, + "learning_rate": 2.9996631878635338e-05, + "loss": 0.0963, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023918920382857322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011968562100082635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05980640202760697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9921797513961792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6770, + "weighted_flow_matching_loss": 0.03691608905792236, + "weighted_rkd_loss": 0.0593387208878994 + }, + { + "epoch": 1.9726505673552517, + "flow_matching_loss": 0.03772129938006401, + "grad_norm": 0.39036694169044495, + "learning_rate": 2.999654371503248e-05, + "loss": 0.0944, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022857950441539286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011374398320913314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05709029920399189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9921566665172576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6780, + "weighted_flow_matching_loss": 0.03772129938006401, + "weighted_rkd_loss": 0.0566425122320652 + }, + { + "epoch": 1.9755600814663952, + "flow_matching_loss": 0.038407350331544875, + "grad_norm": 0.5180209875106812, + "learning_rate": 2.9996454412541092e-05, + "loss": 0.0963, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023177746497094632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012030282150954008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.058385775238275525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9921335577964783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6790, + "weighted_flow_matching_loss": 0.038407350331544875, + "weighted_rkd_loss": 0.05792648307979107 + }, + { + "epoch": 1.9784695955775384, + "flow_matching_loss": 0.038353579491376875, + "grad_norm": 0.4399592876434326, + "learning_rate": 2.9996363971167965e-05, + "loss": 0.099, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024394075386226177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012358993012458085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061147143691778184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9921104192733765, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6800, + "weighted_flow_matching_loss": 0.038353579491376875, + "weighted_rkd_loss": 0.06066470220685005 + }, + { + "epoch": 1.9813791096886821, + "flow_matching_loss": 0.03945258315652609, + "grad_norm": 0.4047827124595642, + "learning_rate": 2.9996272390919962e-05, + "loss": 0.0999, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024226094223558904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01249068435281515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06094287224113941, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9920872390270233, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6810, + "weighted_flow_matching_loss": 0.03945258315652609, + "weighted_rkd_loss": 0.06046065129339695 + }, + { + "epoch": 1.9842886237998254, + "flow_matching_loss": 0.037887361273169516, + "grad_norm": 0.42154136300086975, + "learning_rate": 2.999617967180403e-05, + "loss": 0.0935, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022343487292528153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01133246198296547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0560194369405508, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9920640289783478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6820, + "weighted_flow_matching_loss": 0.037887361273169516, + "weighted_rkd_loss": 0.055574874952435495 + }, + { + "epoch": 1.9871981379109689, + "flow_matching_loss": 0.03633020296692848, + "grad_norm": 0.4533792734146118, + "learning_rate": 2.9996085813827224e-05, + "loss": 0.0928, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022638280317187308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011668454483151436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056945016235113145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9920407831668854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6830, + "weighted_flow_matching_loss": 0.03633020296692848, + "weighted_rkd_loss": 0.056491787731647494 + }, + { + "epoch": 1.9901076520221124, + "flow_matching_loss": 0.03787475693970919, + "grad_norm": 0.4104945957660675, + "learning_rate": 2.9995990816996666e-05, + "loss": 0.0916, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02166785541921854, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010852553881704807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054188264161348344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9920174956321717, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6840, + "weighted_flow_matching_loss": 0.03787475693970919, + "weighted_rkd_loss": 0.05375570021569729 + }, + { + "epoch": 1.9930171661332556, + "flow_matching_loss": 0.038157789036631584, + "grad_norm": 0.4915007948875427, + "learning_rate": 2.9995894681319573e-05, + "loss": 0.0945, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02288084663450718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011055598128587007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05681729130446911, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9919941902160645, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6850, + "weighted_flow_matching_loss": 0.038157789036631584, + "weighted_rkd_loss": 0.056362415477633473 + }, + { + "epoch": 1.9959266802443993, + "flow_matching_loss": 0.03815061151981354, + "grad_norm": 0.4792054295539856, + "learning_rate": 2.9995797406803236e-05, + "loss": 0.0963, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023386815190315248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011845264304429292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05861889384686947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9919708371162415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6860, + "weighted_flow_matching_loss": 0.03815061151981354, + "weighted_rkd_loss": 0.0581482321023941 + }, + { + "epoch": 1.9988361943555426, + "flow_matching_loss": 0.03615482430905104, + "grad_norm": 0.45210692286491394, + "learning_rate": 2.9995698993455056e-05, + "loss": 0.0961, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024224625341594218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012014493253082037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060463744401931765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.991947454214096, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6870, + "weighted_flow_matching_loss": 0.03615482430905104, + "weighted_rkd_loss": 0.0599768552929163 + }, + { + "epoch": 2.001745708466686, + "flow_matching_loss": 0.04005715511739254, + "grad_norm": 0.438740611076355, + "learning_rate": 2.99955994412825e-05, + "loss": 0.0954, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022229982540011406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011347516067326069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05580748133361339, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9919240415096283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6880, + "weighted_flow_matching_loss": 0.04005715511739254, + "weighted_rkd_loss": 0.05535678304731846 + }, + { + "epoch": 2.0046552225778296, + "flow_matching_loss": 0.03755410052835941, + "grad_norm": 0.4954073429107666, + "learning_rate": 2.9995498750293127e-05, + "loss": 0.093, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02233322076499462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011192662268877029, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.055859103426337244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9919005930423737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6890, + "weighted_flow_matching_loss": 0.03755410052835941, + "weighted_rkd_loss": 0.055406691133975984 + }, + { + "epoch": 2.007564736688973, + "flow_matching_loss": 0.03693399652838707, + "grad_norm": 0.5003223419189453, + "learning_rate": 2.9995396920494583e-05, + "loss": 0.0918, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022176246903836728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01100874189287424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05536123588681221, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9918771028518677, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6900, + "weighted_flow_matching_loss": 0.03693399652838707, + "weighted_rkd_loss": 0.054911531135439874 + }, + { + "epoch": 2.0104742508001165, + "flow_matching_loss": 0.03807167466729879, + "grad_norm": 0.41478878259658813, + "learning_rate": 2.9995293951894608e-05, + "loss": 0.0919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02169553181156516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01090005817823112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054291121661663055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9918535947799683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6910, + "weighted_flow_matching_loss": 0.03807167466729879, + "weighted_rkd_loss": 0.05384884551167488 + }, + { + "epoch": 2.01338376491126, + "flow_matching_loss": 0.0389493465423584, + "grad_norm": 0.4641740024089813, + "learning_rate": 2.9995189844501014e-05, + "loss": 0.0951, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022549036145210265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011534849740564823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05663292221724987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9918300449848175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6920, + "weighted_flow_matching_loss": 0.0389493465423584, + "weighted_rkd_loss": 0.056170228868722916 + }, + { + "epoch": 2.016293279022403, + "flow_matching_loss": 0.03617174308747053, + "grad_norm": 0.45683398842811584, + "learning_rate": 2.9995084598321715e-05, + "loss": 0.0924, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022705657593905925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011257899273186923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05666921436786652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9918064594268798, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6930, + "weighted_flow_matching_loss": 0.03617174308747053, + "weighted_rkd_loss": 0.05620489679276943 + }, + { + "epoch": 2.019202793133547, + "flow_matching_loss": 0.038202015310525896, + "grad_norm": 0.41230452060699463, + "learning_rate": 2.9994978213364696e-05, + "loss": 0.0952, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02296258881688118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011509208008646965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05743438489735127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.991782832145691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6940, + "weighted_flow_matching_loss": 0.038202015310525896, + "weighted_rkd_loss": 0.056962430477142334 + }, + { + "epoch": 2.02211230724469, + "flow_matching_loss": 0.039693552628159524, + "grad_norm": 0.5219956636428833, + "learning_rate": 2.999487068963804e-05, + "loss": 0.094, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0218574658036232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01102395188063383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054738883301615714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9917591869831085, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6950, + "weighted_flow_matching_loss": 0.039693552628159524, + "weighted_rkd_loss": 0.05428778566420078 + }, + { + "epoch": 2.0250218213558338, + "flow_matching_loss": 0.0369633624330163, + "grad_norm": 0.42649027705192566, + "learning_rate": 2.9994762027149915e-05, + "loss": 0.0905, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021650524251163004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010712173348292709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05401322320103645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9917355060577393, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6960, + "weighted_flow_matching_loss": 0.0369633624330163, + "weighted_rkd_loss": 0.05356681235134601 + }, + { + "epoch": 2.027931335466977, + "flow_matching_loss": 0.03658410180360079, + "grad_norm": 0.39066851139068604, + "learning_rate": 2.999465222590857e-05, + "loss": 0.0894, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021469473652541638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010284257307648658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05322320461273193, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9917117834091187, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6970, + "weighted_flow_matching_loss": 0.03658410180360079, + "weighted_rkd_loss": 0.05278207324445248 + }, + { + "epoch": 2.0308408495781203, + "flow_matching_loss": 0.03769893255084753, + "grad_norm": 0.3595276474952698, + "learning_rate": 2.9994541285922345e-05, + "loss": 0.0908, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02140278834849596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010712067130953074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053517643734812735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9916880369186402, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6980, + "weighted_flow_matching_loss": 0.03769893255084753, + "weighted_rkd_loss": 0.05307284966111183 + }, + { + "epoch": 2.033750363689264, + "flow_matching_loss": 0.034713924117386344, + "grad_norm": 0.4431023597717285, + "learning_rate": 2.9994429207199663e-05, + "loss": 0.087, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021208334341645242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010288344789296388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052705013379454616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9916642427444458, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6990, + "weighted_flow_matching_loss": 0.034713924117386344, + "weighted_rkd_loss": 0.052265670523047446 + }, + { + "epoch": 2.0366598778004072, + "flow_matching_loss": 0.03739578071981668, + "grad_norm": 0.44168922305107117, + "learning_rate": 2.999431598974904e-05, + "loss": 0.092, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022031258046627044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010983948688954115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05504646487534046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9916404485702515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7000, + "weighted_flow_matching_loss": 0.03739578071981668, + "weighted_rkd_loss": 0.054586301743984225 + }, + { + "epoch": 2.039569391911551, + "flow_matching_loss": 0.034029382094740865, + "grad_norm": 0.44692716002464294, + "learning_rate": 2.999420163357907e-05, + "loss": 0.0888, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022218599915504456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010785099770873784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05522229969501495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9916165769100189, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7010, + "weighted_flow_matching_loss": 0.034029382094740865, + "weighted_rkd_loss": 0.05475932583212852 + }, + { + "epoch": 2.042478906022694, + "flow_matching_loss": 0.03683652579784393, + "grad_norm": 0.38898754119873047, + "learning_rate": 2.9994086138698443e-05, + "loss": 0.0909, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021840810030698776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010840932559221982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05452255196869373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9915926873683929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7020, + "weighted_flow_matching_loss": 0.03683652579784393, + "weighted_rkd_loss": 0.05406417064368725 + }, + { + "epoch": 2.0453884201338375, + "flow_matching_loss": 0.03763440977782011, + "grad_norm": 0.37231171131134033, + "learning_rate": 2.999396950511592e-05, + "loss": 0.0931, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02252916432917118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01089028948917985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05594861917197704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9915687680244446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7030, + "weighted_flow_matching_loss": 0.03763440977782011, + "weighted_rkd_loss": 0.05547690652310848 + }, + { + "epoch": 2.048297934244981, + "flow_matching_loss": 0.036886153370141984, + "grad_norm": 0.3557099401950836, + "learning_rate": 2.9993851732840363e-05, + "loss": 0.095, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02336722705513239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011869171727448702, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05860362611711025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9915448069572449, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7040, + "weighted_flow_matching_loss": 0.036886153370141984, + "weighted_rkd_loss": 0.05810812897980213 + }, + { + "epoch": 2.0512074483561245, + "flow_matching_loss": 0.037182371318340304, + "grad_norm": 0.4081055819988251, + "learning_rate": 2.9993732821880725e-05, + "loss": 0.0912, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02159420158714056, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011268603429198266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05445700660347939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9915208220481873, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7050, + "weighted_flow_matching_loss": 0.037182371318340304, + "weighted_rkd_loss": 0.053995246067643166 + }, + { + "epoch": 2.054116962467268, + "flow_matching_loss": 0.03632760718464852, + "grad_norm": 0.540682852268219, + "learning_rate": 2.9993612772246026e-05, + "loss": 0.0923, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0225372027605772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011360380426049232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056434785947203636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9914968073368072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7060, + "weighted_flow_matching_loss": 0.03632760718464852, + "weighted_rkd_loss": 0.055954893678426744 + }, + { + "epoch": 2.0570264765784114, + "flow_matching_loss": 0.03555865455418825, + "grad_norm": 0.432623028755188, + "learning_rate": 2.9993491583945383e-05, + "loss": 0.092, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022715581022202967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011472396180033683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05690355859696865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9914727449417114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7070, + "weighted_flow_matching_loss": 0.03555865455418825, + "weighted_rkd_loss": 0.05641833283007145 + }, + { + "epoch": 2.0599359906895547, + "flow_matching_loss": 0.0362356785684824, + "grad_norm": 0.4211861789226532, + "learning_rate": 2.999336925698801e-05, + "loss": 0.0892, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021151181869208814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011072848364710808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053375212103128435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9914486587047577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7080, + "weighted_flow_matching_loss": 0.0362356785684824, + "weighted_rkd_loss": 0.052918770164251325 + }, + { + "epoch": 2.0628455048006984, + "flow_matching_loss": 0.03963070809841156, + "grad_norm": 0.4119345545768738, + "learning_rate": 2.9993245791383183e-05, + "loss": 0.0962, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022881683707237244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011320954095572234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05708432197570801, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9914245307445526, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7090, + "weighted_flow_matching_loss": 0.03963070809841156, + "weighted_rkd_loss": 0.056594790890812874 + }, + { + "epoch": 2.0657550189118417, + "flow_matching_loss": 0.03672750256955624, + "grad_norm": 0.4462577998638153, + "learning_rate": 2.9993121187140288e-05, + "loss": 0.0921, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022158318385481834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011554905585944653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.055871542543172836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9914003789424897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7100, + "weighted_flow_matching_loss": 0.03672750256955624, + "weighted_rkd_loss": 0.05539106652140617 + }, + { + "epoch": 2.0686645330229854, + "flow_matching_loss": 0.036813709139823916, + "grad_norm": 0.44013917446136475, + "learning_rate": 2.9992995444268783e-05, + "loss": 0.0884, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020940536446869374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010112964548170567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05199403688311577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9913761913776398, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7110, + "weighted_flow_matching_loss": 0.036813709139823916, + "weighted_rkd_loss": 0.0515456423163414 + }, + { + "epoch": 2.0715740471341286, + "flow_matching_loss": 0.03719753883779049, + "grad_norm": 0.49004966020584106, + "learning_rate": 2.9992868562778222e-05, + "loss": 0.0916, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022044669650495053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010772965289652348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054862304404377935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9913519620895386, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7120, + "weighted_flow_matching_loss": 0.03719753883779049, + "weighted_rkd_loss": 0.05438787341117859 + }, + { + "epoch": 2.074483561245272, + "flow_matching_loss": 0.038229535520076754, + "grad_norm": 0.4378277659416199, + "learning_rate": 2.9992740542678235e-05, + "loss": 0.0957, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02307715155184269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01186332544311881, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05801762826740742, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9913277089595794, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7130, + "weighted_flow_matching_loss": 0.038229535520076754, + "weighted_rkd_loss": 0.057514464110136033 + }, + { + "epoch": 2.0773930753564156, + "flow_matching_loss": 0.03902163989841938, + "grad_norm": 0.3569454848766327, + "learning_rate": 2.999261138397855e-05, + "loss": 0.091, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02097407467663288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01050404030829668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05245218984782696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9913034200668335, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7140, + "weighted_flow_matching_loss": 0.03902163989841938, + "weighted_rkd_loss": 0.0519960418343544 + }, + { + "epoch": 2.080302589467559, + "flow_matching_loss": 0.038513701595366004, + "grad_norm": 0.4551214873790741, + "learning_rate": 2.999248108668897e-05, + "loss": 0.0932, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022014809027314185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01110352175310254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05513314045965671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9912790894508362, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7150, + "weighted_flow_matching_loss": 0.038513701595366004, + "weighted_rkd_loss": 0.05465231947600842 + }, + { + "epoch": 2.0832121035787026, + "flow_matching_loss": 0.03537702988833189, + "grad_norm": 0.4461776912212372, + "learning_rate": 2.9992349650819396e-05, + "loss": 0.085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020026468485593796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010057346522808075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050110283121466634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.991254734992981, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7160, + "weighted_flow_matching_loss": 0.03537702988833189, + "weighted_rkd_loss": 0.04967204444110394 + }, + { + "epoch": 2.086121617689846, + "flow_matching_loss": 0.03637293372303248, + "grad_norm": 0.43753474950790405, + "learning_rate": 2.9992217076379807e-05, + "loss": 0.0902, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021675002202391623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010903667844831943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054253671690821646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9912303388118744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7170, + "weighted_flow_matching_loss": 0.03637293372303248, + "weighted_rkd_loss": 0.05377788953483105 + }, + { + "epoch": 2.089031131800989, + "flow_matching_loss": 0.03688667863607407, + "grad_norm": 0.459587037563324, + "learning_rate": 2.9992083363380266e-05, + "loss": 0.0884, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020906066708266734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010130362864583731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05194249674677849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9912059009075165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7180, + "weighted_flow_matching_loss": 0.03688667863607407, + "weighted_rkd_loss": 0.051485703140497205 + }, + { + "epoch": 2.091940645912133, + "flow_matching_loss": 0.03710534125566482, + "grad_norm": 0.460720032453537, + "learning_rate": 2.999194851183094e-05, + "loss": 0.0904, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02152817025780678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01070749033242464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05376383103430271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9911814630031586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7190, + "weighted_flow_matching_loss": 0.03710534125566482, + "weighted_rkd_loss": 0.05328970178961754 + }, + { + "epoch": 2.094850160023276, + "flow_matching_loss": 0.03648792337626219, + "grad_norm": 0.3762485086917877, + "learning_rate": 2.9991812521742056e-05, + "loss": 0.0909, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021965480595827102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010965587571263313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054896549135446546, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9911569595336914, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7200, + "weighted_flow_matching_loss": 0.03648792337626219, + "weighted_rkd_loss": 0.054411111027002336 + }, + { + "epoch": 2.0977596741344193, + "flow_matching_loss": 0.03511032685637474, + "grad_norm": 0.3824444115161896, + "learning_rate": 2.9991675393123953e-05, + "loss": 0.091, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022519554384052752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01130162924528122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056340738013386724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9911324381828308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7210, + "weighted_flow_matching_loss": 0.03511032685637474, + "weighted_rkd_loss": 0.055841134116053584 + }, + { + "epoch": 2.100669188245563, + "flow_matching_loss": 0.03765986263751984, + "grad_norm": 0.4026249647140503, + "learning_rate": 2.999153712598704e-05, + "loss": 0.0906, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02147585414350033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010485139675438404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053436847403645515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9911078751087189, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7220, + "weighted_flow_matching_loss": 0.03765986263751984, + "weighted_rkd_loss": 0.052961692586541174 + }, + { + "epoch": 2.1035787023567063, + "flow_matching_loss": 0.033841415867209436, + "grad_norm": 0.45709505677223206, + "learning_rate": 2.9991397720341817e-05, + "loss": 0.0887, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022297032177448273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010745767271146178, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05533983111381531, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9910832822322846, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7230, + "weighted_flow_matching_loss": 0.033841415867209436, + "weighted_rkd_loss": 0.05484635643661022 + }, + { + "epoch": 2.10648821646785, + "flow_matching_loss": 0.036909409984946254, + "grad_norm": 0.3844383656978607, + "learning_rate": 2.9991257176198873e-05, + "loss": 0.0909, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022125593014061452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010249642096459866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05450082831084728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9910586476325989, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7240, + "weighted_flow_matching_loss": 0.036909409984946254, + "weighted_rkd_loss": 0.054013531655073166 + }, + { + "epoch": 2.1093977305789933, + "flow_matching_loss": 0.034722805954515934, + "grad_norm": 0.39973512291908264, + "learning_rate": 2.9991115493568876e-05, + "loss": 0.0885, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021604342572391032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011094934493303298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05430361926555634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9910339891910553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7250, + "weighted_flow_matching_loss": 0.034722805954515934, + "weighted_rkd_loss": 0.053816693276166915 + }, + { + "epoch": 2.112307244690137, + "flow_matching_loss": 0.03571246583014727, + "grad_norm": 0.46277886629104614, + "learning_rate": 2.9990972672462592e-05, + "loss": 0.087, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02070563454180956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010364356311038136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05177562460303307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9910092949867249, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7260, + "weighted_flow_matching_loss": 0.03571246583014727, + "weighted_rkd_loss": 0.05131011977791786 + }, + { + "epoch": 2.1152167588012802, + "flow_matching_loss": 0.03579765930771828, + "grad_norm": 0.4425850808620453, + "learning_rate": 2.9990828712890864e-05, + "loss": 0.0853, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01998839285224676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010019590519368649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049996376410126685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9909845590591431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7270, + "weighted_flow_matching_loss": 0.03579765930771828, + "weighted_rkd_loss": 0.04954565651714802 + }, + { + "epoch": 2.1181262729124235, + "flow_matching_loss": 0.03720210231840611, + "grad_norm": 0.4831775426864624, + "learning_rate": 2.9990683614864633e-05, + "loss": 0.0896, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021390204690396784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010124138463288546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05290454775094986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9909598112106324, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7280, + "weighted_flow_matching_loss": 0.03720210231840611, + "weighted_rkd_loss": 0.052426275610923764 + }, + { + "epoch": 2.121035787023567, + "flow_matching_loss": 0.03517989590764046, + "grad_norm": 0.4717470109462738, + "learning_rate": 2.9990537378394905e-05, + "loss": 0.087, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020906108245253564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010450625233352185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05226284153759479, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9909350097179412, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7290, + "weighted_flow_matching_loss": 0.03517989590764046, + "weighted_rkd_loss": 0.05178906843066215 + }, + { + "epoch": 2.1239453011347105, + "flow_matching_loss": 0.036431129463016985, + "grad_norm": 0.512520968914032, + "learning_rate": 2.9990390003492794e-05, + "loss": 0.0918, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022169429063796996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01149428803473711, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05583314597606659, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9909101784229278, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7300, + "weighted_flow_matching_loss": 0.036431129463016985, + "weighted_rkd_loss": 0.055325638875365256 + }, + { + "epoch": 2.1268548152458537, + "flow_matching_loss": 0.03398188222199679, + "grad_norm": 0.42860621213912964, + "learning_rate": 2.999024149016949e-05, + "loss": 0.091, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0230555122718215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011413088254630565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05752411298453808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.990885317325592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7310, + "weighted_flow_matching_loss": 0.03398188222199679, + "weighted_rkd_loss": 0.056999800354242326 + }, + { + "epoch": 2.1297643293569974, + "flow_matching_loss": 0.03691830076277256, + "grad_norm": 0.39471927285194397, + "learning_rate": 2.9990091838436272e-05, + "loss": 0.087, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02032548952847719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009918548073619604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05056952759623527, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9908604204654694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7320, + "weighted_flow_matching_loss": 0.03691830076277256, + "weighted_rkd_loss": 0.050107328593730925 + }, + { + "epoch": 2.1326738434681407, + "flow_matching_loss": 0.037823998183012006, + "grad_norm": 0.3766818940639496, + "learning_rate": 2.9989941048304505e-05, + "loss": 0.0944, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022706297039985657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01168526690453291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05709786079823971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9908354818820954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7330, + "weighted_flow_matching_loss": 0.037823998183012006, + "weighted_rkd_loss": 0.056574607640504836 + }, + { + "epoch": 2.1355833575792844, + "flow_matching_loss": 0.03573935646563768, + "grad_norm": 0.40288054943084717, + "learning_rate": 2.9989789119785636e-05, + "loss": 0.0908, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022145858220756053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011257321387529374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.055549037829041484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9908105194568634, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7340, + "weighted_flow_matching_loss": 0.03573935646563768, + "weighted_rkd_loss": 0.05503856241703033 + }, + { + "epoch": 2.1384928716904277, + "flow_matching_loss": 0.03719951417297125, + "grad_norm": 0.4811195731163025, + "learning_rate": 2.998963605289121e-05, + "loss": 0.0859, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019591855630278588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009966703690588474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049150414764881134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.990785539150238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7350, + "weighted_flow_matching_loss": 0.03719951417297125, + "weighted_rkd_loss": 0.04869750924408436 + }, + { + "epoch": 2.141402385801571, + "flow_matching_loss": 0.03428837694227695, + "grad_norm": 0.4060567021369934, + "learning_rate": 2.998948184763285e-05, + "loss": 0.084, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020209949091076852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009741459041833878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05016135759651661, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9907605051994324, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7360, + "weighted_flow_matching_loss": 0.03428837694227695, + "weighted_rkd_loss": 0.049697884172201154 + }, + { + "epoch": 2.1443118999127146, + "flow_matching_loss": 0.036729994043707845, + "grad_norm": 0.44716766476631165, + "learning_rate": 2.9989326504022268e-05, + "loss": 0.0898, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02133522043004632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01088936603628099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0535598061978817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9907354235649108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7370, + "weighted_flow_matching_loss": 0.036729994043707845, + "weighted_rkd_loss": 0.05306355990469456 + }, + { + "epoch": 2.147221414023858, + "flow_matching_loss": 0.0348359240218997, + "grad_norm": 0.4482835829257965, + "learning_rate": 2.9989170022071258e-05, + "loss": 0.0842, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02018679566681385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009423697274178267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049797289073467255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.990710335969925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7380, + "weighted_flow_matching_loss": 0.0348359240218997, + "weighted_rkd_loss": 0.049334684014320375 + }, + { + "epoch": 2.1501309281350016, + "flow_matching_loss": 0.03522690311074257, + "grad_norm": 0.4175436794757843, + "learning_rate": 2.99890124017917e-05, + "loss": 0.0888, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021713786013424396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010668859910219907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05409643203020096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9906851947307587, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7390, + "weighted_flow_matching_loss": 0.03522690311074257, + "weighted_rkd_loss": 0.05359249897301197 + }, + { + "epoch": 2.153040442246145, + "flow_matching_loss": 0.03586749881505966, + "grad_norm": 0.4606802463531494, + "learning_rate": 2.998885364319557e-05, + "loss": 0.087, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02065847646445036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010343563091009856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051660516485571864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99066002368927, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7400, + "weighted_flow_matching_loss": 0.03586749881505966, + "weighted_rkd_loss": 0.05117800384759903 + }, + { + "epoch": 2.155949956357288, + "flow_matching_loss": 0.03521877471357584, + "grad_norm": 0.39057403802871704, + "learning_rate": 2.9988693746294926e-05, + "loss": 0.0843, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019898227974772455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009769340883940459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049565797299146654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9906348288059235, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7410, + "weighted_flow_matching_loss": 0.03521877471357584, + "weighted_rkd_loss": 0.049101597443223 + }, + { + "epoch": 2.158859470468432, + "flow_matching_loss": 0.040537115186452866, + "grad_norm": 0.5246657729148865, + "learning_rate": 2.9988532711101913e-05, + "loss": 0.0957, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022406486421823503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010836570616811514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05564954243600369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99060959815979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7420, + "weighted_flow_matching_loss": 0.040537115186452866, + "weighted_rkd_loss": 0.05512697733938694 + }, + { + "epoch": 2.161768984579575, + "flow_matching_loss": 0.0360575009137392, + "grad_norm": 0.44642969965934753, + "learning_rate": 2.998837053762875e-05, + "loss": 0.0844, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019722362235188483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00934345372952521, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04878817908465862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9905843317508698, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7430, + "weighted_flow_matching_loss": 0.0360575009137392, + "weighted_rkd_loss": 0.04832881353795528 + }, + { + "epoch": 2.164678498690719, + "flow_matching_loss": 0.03607506901025772, + "grad_norm": 0.4519486725330353, + "learning_rate": 2.9988207225887757e-05, + "loss": 0.0923, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02281950730830431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011085703503340483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05672471784055233, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9905590355396271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7440, + "weighted_flow_matching_loss": 0.03607506901025772, + "weighted_rkd_loss": 0.056189189106225966 + }, + { + "epoch": 2.167588012801862, + "flow_matching_loss": 0.034238369576632975, + "grad_norm": 0.42907676100730896, + "learning_rate": 2.9988042775891344e-05, + "loss": 0.0861, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020952518656849863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010483193956315518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052388230711221694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9905336976051331, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7450, + "weighted_flow_matching_loss": 0.034238369576632975, + "weighted_rkd_loss": 0.05189228951931 + }, + { + "epoch": 2.1704975269130053, + "flow_matching_loss": 0.035509121045470236, + "grad_norm": 0.3898758292198181, + "learning_rate": 2.9987877187651996e-05, + "loss": 0.0837, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019678694382309913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009336722362786532, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04869411140680313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9905083239078522, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7460, + "weighted_flow_matching_loss": 0.035509121045470236, + "weighted_rkd_loss": 0.048231934756040575 + }, + { + "epoch": 2.173407041024149, + "flow_matching_loss": 0.03742838092148304, + "grad_norm": 0.3559173345565796, + "learning_rate": 2.9987710461182285e-05, + "loss": 0.087, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019928259961307048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010182629292830824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05003914907574654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9904829263687134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7470, + "weighted_flow_matching_loss": 0.03742838092148304, + "weighted_rkd_loss": 0.04956292696297169 + }, + { + "epoch": 2.1763165551352923, + "flow_matching_loss": 0.03523981031030417, + "grad_norm": 0.3963957130908966, + "learning_rate": 2.9987542596494874e-05, + "loss": 0.0875, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02137475125491619, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009972958918660879, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05272246114909649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9904574930667878, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7480, + "weighted_flow_matching_loss": 0.03523981031030417, + "weighted_rkd_loss": 0.0522193793207407 + }, + { + "epoch": 2.179226069246436, + "flow_matching_loss": 0.03666099421679973, + "grad_norm": 0.3544101119041443, + "learning_rate": 2.998737359360251e-05, + "loss": 0.0894, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021332918386906385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010532998759299517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053198835253715514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9904320240020752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7490, + "weighted_flow_matching_loss": 0.03666099421679973, + "weighted_rkd_loss": 0.05268986150622368 + }, + { + "epoch": 2.1821355833575793, + "flow_matching_loss": 0.035937008820474146, + "grad_norm": 0.417886883020401, + "learning_rate": 2.9987203452518036e-05, + "loss": 0.088, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02113743554800749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01031619617715478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05259106680750847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9904065191745758, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7500, + "weighted_flow_matching_loss": 0.035937008820474146, + "weighted_rkd_loss": 0.05208652690052986 + }, + { + "epoch": 2.1850450974687226, + "flow_matching_loss": 0.039008954912424086, + "grad_norm": 0.4182053804397583, + "learning_rate": 2.998703217325436e-05, + "loss": 0.0852, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018746111169457437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00910413139499724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046596353873610497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9903809905052186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7510, + "weighted_flow_matching_loss": 0.039008954912424086, + "weighted_rkd_loss": 0.04614814817905426 + }, + { + "epoch": 2.1879546115798663, + "flow_matching_loss": 0.03671146295964718, + "grad_norm": 0.38490596413612366, + "learning_rate": 2.99868597558245e-05, + "loss": 0.0876, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020680499076843262, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010058473190292716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051419471949338914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9903554022312164, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7520, + "weighted_flow_matching_loss": 0.03671146295964718, + "weighted_rkd_loss": 0.05092356875538826 + }, + { + "epoch": 2.1908641256910095, + "flow_matching_loss": 0.03479685019701719, + "grad_norm": 0.44559767842292786, + "learning_rate": 2.9986686200241544e-05, + "loss": 0.0826, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01939048580825329, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009439835557714105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04822080731391907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9903298318386078, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7530, + "weighted_flow_matching_loss": 0.03479685019701719, + "weighted_rkd_loss": 0.04775450639426708 + }, + { + "epoch": 2.1937736398021532, + "flow_matching_loss": 0.03678249288350344, + "grad_norm": 0.3554181456565857, + "learning_rate": 2.9986511506518677e-05, + "loss": 0.0822, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018383287452161313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00908052227459848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04584709703922272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9903041779994964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7540, + "weighted_flow_matching_loss": 0.03678249288350344, + "weighted_rkd_loss": 0.04540258273482323 + }, + { + "epoch": 2.1966831539132965, + "flow_matching_loss": 0.035368209891021254, + "grad_norm": 0.3448430299758911, + "learning_rate": 2.9986335674669163e-05, + "loss": 0.0842, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01972241010516882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009863242227584123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04930806271731854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9902785062789917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7550, + "weighted_flow_matching_loss": 0.035368209891021254, + "weighted_rkd_loss": 0.04882872998714447 + }, + { + "epoch": 2.1995926680244398, + "flow_matching_loss": 0.035390432178974154, + "grad_norm": 0.404897004365921, + "learning_rate": 2.9986158704706353e-05, + "loss": 0.0831, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01946596410125494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009241610951721668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048173538967967035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9902527987957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7560, + "weighted_flow_matching_loss": 0.035390432178974154, + "weighted_rkd_loss": 0.04770399555563927 + }, + { + "epoch": 2.2025021821355835, + "flow_matching_loss": 0.03549037426710129, + "grad_norm": 0.4229907989501953, + "learning_rate": 2.9985980596643692e-05, + "loss": 0.0855, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0203544145449996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009809909062460065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050518738478422164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9902270674705506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7570, + "weighted_flow_matching_loss": 0.03549037426710129, + "weighted_rkd_loss": 0.05002502538263798 + }, + { + "epoch": 2.2054116962467267, + "flow_matching_loss": 0.03597133569419384, + "grad_norm": 0.3823700249195099, + "learning_rate": 2.9985801350494705e-05, + "loss": 0.0859, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02008684501051903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010275709815323353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05044940002262592, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9902012884616852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7580, + "weighted_flow_matching_loss": 0.03597133569419384, + "weighted_rkd_loss": 0.04995507560670376 + }, + { + "epoch": 2.2083212103578704, + "flow_matching_loss": 0.036774967424571516, + "grad_norm": 0.3857390284538269, + "learning_rate": 2.9985620966273e-05, + "loss": 0.085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019547623209655286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009580428525805474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04867567494511604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9901754856109619, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7590, + "weighted_flow_matching_loss": 0.036774967424571516, + "weighted_rkd_loss": 0.048197468370199205 + }, + { + "epoch": 2.2112307244690137, + "flow_matching_loss": 0.03751516900956631, + "grad_norm": 0.3798995316028595, + "learning_rate": 2.9985439443992278e-05, + "loss": 0.0856, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019599237758666278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009343271842226386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04854174740612507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9901496469974518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7600, + "weighted_flow_matching_loss": 0.03751516900956631, + "weighted_rkd_loss": 0.04806359633803368 + }, + { + "epoch": 2.214140238580157, + "flow_matching_loss": 0.03706750050187111, + "grad_norm": 0.3675595819950104, + "learning_rate": 2.9985256783666326e-05, + "loss": 0.0852, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019554377906024455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009474583528935909, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04858333915472031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9901237726211548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7610, + "weighted_flow_matching_loss": 0.03706750050187111, + "weighted_rkd_loss": 0.04810353480279446 + }, + { + "epoch": 2.2170497526913007, + "flow_matching_loss": 0.0345027843490243, + "grad_norm": 0.4809371531009674, + "learning_rate": 2.9985072985309018e-05, + "loss": 0.0835, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019753569923341274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00998210427351296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04948924407362938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9900978744029999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7620, + "weighted_flow_matching_loss": 0.0345027843490243, + "weighted_rkd_loss": 0.04899918809533119 + }, + { + "epoch": 2.219959266802444, + "flow_matching_loss": 0.033858468569815156, + "grad_norm": 0.4664188027381897, + "learning_rate": 2.998488804893431e-05, + "loss": 0.0863, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02114516720175743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010659070685505867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0529494047164917, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9900719344615936, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7630, + "weighted_flow_matching_loss": 0.033858468569815156, + "weighted_rkd_loss": 0.0524236973375082 + }, + { + "epoch": 2.2228687809135876, + "flow_matching_loss": 0.03530678041279316, + "grad_norm": 0.42486104369163513, + "learning_rate": 2.9984701974556245e-05, + "loss": 0.0848, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020042764768004418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009941395791247486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05002692528069019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9900459527969361, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7640, + "weighted_flow_matching_loss": 0.03530678041279316, + "weighted_rkd_loss": 0.04952896423637867 + }, + { + "epoch": 2.225778295024731, + "flow_matching_loss": 0.03648079670965672, + "grad_norm": 0.4848770499229431, + "learning_rate": 2.9984514762188953e-05, + "loss": 0.089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021366327442228793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010328747797757387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053061402216553685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9900199472904205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7650, + "weighted_flow_matching_loss": 0.03648079670965672, + "weighted_rkd_loss": 0.052531836181879045 + }, + { + "epoch": 2.228687809135874, + "flow_matching_loss": 0.03649526163935661, + "grad_norm": 0.4019011855125427, + "learning_rate": 2.9984326411846658e-05, + "loss": 0.0877, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020516783744096757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010728601180016995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051762168481945994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9899939119815826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7660, + "weighted_flow_matching_loss": 0.03649526163935661, + "weighted_rkd_loss": 0.05124427005648613 + }, + { + "epoch": 2.231597323247018, + "flow_matching_loss": 0.03499767705798149, + "grad_norm": 0.36160415410995483, + "learning_rate": 2.998413692354366e-05, + "loss": 0.083, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01938841734081507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009708335064351559, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04848516918718815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9899678349494934, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7670, + "weighted_flow_matching_loss": 0.03499767705798149, + "weighted_rkd_loss": 0.0479987595230341 + }, + { + "epoch": 2.234506837358161, + "flow_matching_loss": 0.03738203197717667, + "grad_norm": 0.46170586347579956, + "learning_rate": 2.9983946297294346e-05, + "loss": 0.0856, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01957099009305239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009534591203555464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04867657162249088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9899417340755463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7680, + "weighted_flow_matching_loss": 0.03738203197717667, + "weighted_rkd_loss": 0.048186967894434926 + }, + { + "epoch": 2.237416351469305, + "flow_matching_loss": 0.03522681333124637, + "grad_norm": 0.39942339062690735, + "learning_rate": 2.99837545331132e-05, + "loss": 0.0861, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02043668869882822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010499961953610183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05137333944439888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9899155914783477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7690, + "weighted_flow_matching_loss": 0.03522681333124637, + "weighted_rkd_loss": 0.050855261459946635 + }, + { + "epoch": 2.240325865580448, + "flow_matching_loss": 0.0341292617842555, + "grad_norm": 0.38864871859550476, + "learning_rate": 2.9983561631014777e-05, + "loss": 0.0839, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020243818871676922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009834511205554008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050322148576378824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9898894190788269, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7700, + "weighted_flow_matching_loss": 0.0341292617842555, + "weighted_rkd_loss": 0.04981338903307915 + }, + { + "epoch": 2.2432353796915914, + "flow_matching_loss": 0.037283194810152055, + "grad_norm": 0.41773107647895813, + "learning_rate": 2.998336759101373e-05, + "loss": 0.0878, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02057072054594755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00988682359457016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051028264313936235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9898632168769836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7710, + "weighted_flow_matching_loss": 0.037283194810152055, + "weighted_rkd_loss": 0.05051100961863995 + }, + { + "epoch": 2.246144893802735, + "flow_matching_loss": 0.03671380952000618, + "grad_norm": 0.368413507938385, + "learning_rate": 2.99831724131248e-05, + "loss": 0.0905, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021780436113476753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010794492438435555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05435536466538906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.989836972951889, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7720, + "weighted_flow_matching_loss": 0.03671380952000618, + "weighted_rkd_loss": 0.05380294732749462 + }, + { + "epoch": 2.2490544079138783, + "flow_matching_loss": 0.035796545818448065, + "grad_norm": 0.47578299045562744, + "learning_rate": 2.9982976097362807e-05, + "loss": 0.0873, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02070750743150711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010571578703820705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051986593008041385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.989810687303543, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7730, + "weighted_flow_matching_loss": 0.035796545818448065, + "weighted_rkd_loss": 0.05145687647163868 + }, + { + "epoch": 2.2519639220250216, + "flow_matching_loss": 0.03952846694737673, + "grad_norm": 0.33517786860466003, + "learning_rate": 2.9982778643742655e-05, + "loss": 0.0881, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01960249086841941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009879893762990833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04908487498760224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9897843837738037, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7740, + "weighted_flow_matching_loss": 0.03952846694737673, + "weighted_rkd_loss": 0.048583457991480825 + }, + { + "epoch": 2.2548734361361653, + "flow_matching_loss": 0.03530619516968727, + "grad_norm": 0.4678904414176941, + "learning_rate": 2.9982580052279346e-05, + "loss": 0.0858, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0204856188967824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01002290826290846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05099414624273777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.989758038520813, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7750, + "weighted_flow_matching_loss": 0.03530619516968727, + "weighted_rkd_loss": 0.050471866875886916 + }, + { + "epoch": 2.2577829502473086, + "flow_matching_loss": 0.033460847660899164, + "grad_norm": 0.3554935157299042, + "learning_rate": 2.9982380322987955e-05, + "loss": 0.0843, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020640808157622814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010104002337902785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05138561837375164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9897316694259644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7760, + "weighted_flow_matching_loss": 0.033460847660899164, + "weighted_rkd_loss": 0.05085795857012272 + }, + { + "epoch": 2.2606924643584523, + "flow_matching_loss": 0.03479693364351988, + "grad_norm": 0.36883246898651123, + "learning_rate": 2.9982179455883657e-05, + "loss": 0.0845, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020241091772913933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009759605396538973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05024178884923458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9897052526474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7770, + "weighted_flow_matching_loss": 0.03479693364351988, + "weighted_rkd_loss": 0.049724574014544484 + }, + { + "epoch": 2.2636019784695955, + "flow_matching_loss": 0.034511156193912026, + "grad_norm": 0.44542568922042847, + "learning_rate": 2.99819774509817e-05, + "loss": 0.0853, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02063350100070238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010057258047163486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051324259489774704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9896788120269775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7780, + "weighted_flow_matching_loss": 0.034511156193912026, + "weighted_rkd_loss": 0.05079454779624939 + }, + { + "epoch": 2.2665114925807393, + "flow_matching_loss": 0.0355930695310235, + "grad_norm": 0.3665625751018524, + "learning_rate": 2.9981774308297428e-05, + "loss": 0.0857, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0206189913675189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009439785592257976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050677768513560296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9896523356437683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7790, + "weighted_flow_matching_loss": 0.0355930695310235, + "weighted_rkd_loss": 0.050153376534581184 + }, + { + "epoch": 2.2694210066918825, + "flow_matching_loss": 0.034934880770742895, + "grad_norm": 0.4625968039035797, + "learning_rate": 2.998157002784627e-05, + "loss": 0.0844, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020109334401786326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009765067882835866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049983736500144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9896258294582367, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7800, + "weighted_flow_matching_loss": 0.034934880770742895, + "weighted_rkd_loss": 0.04946520067751407 + }, + { + "epoch": 2.2723305208030258, + "flow_matching_loss": 0.03368411883711815, + "grad_norm": 0.3531026244163513, + "learning_rate": 2.998136460964374e-05, + "loss": 0.0787, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018454095534980296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008577259164303542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045485451072454455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9895992815494538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7810, + "weighted_flow_matching_loss": 0.03368411883711815, + "weighted_rkd_loss": 0.045012378320097926 + }, + { + "epoch": 2.2752400349141695, + "flow_matching_loss": 0.034980682842433455, + "grad_norm": 0.34459394216537476, + "learning_rate": 2.998115805370543e-05, + "loss": 0.0813, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018850747123360635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009091292228549718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046792786568403244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9895727038383484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7820, + "weighted_flow_matching_loss": 0.034980682842433455, + "weighted_rkd_loss": 0.04630486220121384 + }, + { + "epoch": 2.2781495490253127, + "flow_matching_loss": 0.03398433905094862, + "grad_norm": 0.5070244073867798, + "learning_rate": 2.998095036004704e-05, + "loss": 0.0832, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01986895836889744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01001574844121933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04975366517901421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9895461022853851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7830, + "weighted_flow_matching_loss": 0.03398433905094862, + "weighted_rkd_loss": 0.049233530089259146 + }, + { + "epoch": 2.281059063136456, + "flow_matching_loss": 0.03435635752975941, + "grad_norm": 0.4130939543247223, + "learning_rate": 2.9980741528684333e-05, + "loss": 0.0845, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020412248000502585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009823843371123075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05064833983778953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9895194470882416, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7840, + "weighted_flow_matching_loss": 0.03435635752975941, + "weighted_rkd_loss": 0.05011752098798752 + }, + { + "epoch": 2.2839685772475997, + "flow_matching_loss": 0.03764599021524191, + "grad_norm": 0.4543352425098419, + "learning_rate": 2.998053155963317e-05, + "loss": 0.0872, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02001530732959509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010081454738974571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05011207014322281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9894927680492401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7850, + "weighted_flow_matching_loss": 0.03764599021524191, + "weighted_rkd_loss": 0.049585509672760965 + }, + { + "epoch": 2.286878091358743, + "flow_matching_loss": 0.03513955231755972, + "grad_norm": 0.43406376242637634, + "learning_rate": 2.99803204529095e-05, + "loss": 0.0813, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018785883113741873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00905583300627768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04662759937345982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9894660651683808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7860, + "weighted_flow_matching_loss": 0.03513955231755972, + "weighted_rkd_loss": 0.046136414632201195 + }, + { + "epoch": 2.2897876054698867, + "flow_matching_loss": 0.035575623624026775, + "grad_norm": 0.37658560276031494, + "learning_rate": 2.9980108208529354e-05, + "loss": 0.0834, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01931106625124812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009756158478558064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048378290981054305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9894393146038055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7870, + "weighted_flow_matching_loss": 0.035575623624026775, + "weighted_rkd_loss": 0.04786739051342011 + }, + { + "epoch": 2.29269711958103, + "flow_matching_loss": 0.035151480697095394, + "grad_norm": 0.3851044178009033, + "learning_rate": 2.9979894826508845e-05, + "loss": 0.087, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020827877055853605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010762055730447173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05241780951619148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9894125282764434, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7880, + "weighted_flow_matching_loss": 0.035151480697095394, + "weighted_rkd_loss": 0.05186285637319088 + }, + { + "epoch": 2.295606633692173, + "flow_matching_loss": 0.03626244366168976, + "grad_norm": 0.38916921615600586, + "learning_rate": 2.9979680306864186e-05, + "loss": 0.0849, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01965137217193842, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009864014061167835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04916675910353661, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9893857419490815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7890, + "weighted_flow_matching_loss": 0.03626244366168976, + "weighted_rkd_loss": 0.04864491745829582 + }, + { + "epoch": 2.298516147803317, + "flow_matching_loss": 0.03501692395657301, + "grad_norm": 0.4350258708000183, + "learning_rate": 2.9979464649611663e-05, + "loss": 0.0843, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01980293560773134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010161128267645836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04976699911057949, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9893588721752167, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7900, + "weighted_flow_matching_loss": 0.03501692395657301, + "weighted_rkd_loss": 0.04923745393753052 + }, + { + "epoch": 2.30142566191446, + "flow_matching_loss": 0.03600369356572628, + "grad_norm": 0.4437600374221802, + "learning_rate": 2.997924785476766e-05, + "loss": 0.0843, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01952673848718405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009812944848090411, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04886642210185528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9893320083618165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7910, + "weighted_flow_matching_loss": 0.03600369356572628, + "weighted_rkd_loss": 0.04834511242806912 + }, + { + "epoch": 2.304335176025604, + "flow_matching_loss": 0.03496904894709587, + "grad_norm": 0.43558210134506226, + "learning_rate": 2.997902992234863e-05, + "loss": 0.0871, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020984864234924315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010725219175219536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05269494764506817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9893050849437713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7920, + "weighted_flow_matching_loss": 0.03496904894709587, + "weighted_rkd_loss": 0.05213138312101364 + }, + { + "epoch": 2.307244690136747, + "flow_matching_loss": 0.034208478219807145, + "grad_norm": 0.4308767020702362, + "learning_rate": 2.9978810852371138e-05, + "loss": 0.0812, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018993821926414967, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009470288595184683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0474579319357872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9892781376838684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7930, + "weighted_flow_matching_loss": 0.034208478219807145, + "weighted_rkd_loss": 0.04694908745586872 + }, + { + "epoch": 2.3101542042478904, + "flow_matching_loss": 0.03367581926286221, + "grad_norm": 0.3861002027988434, + "learning_rate": 2.9978590644851808e-05, + "loss": 0.0832, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02023618370294571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009611620008945465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050083987414836884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9892511665821075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7940, + "weighted_flow_matching_loss": 0.03367581926286221, + "weighted_rkd_loss": 0.049545639753341676 + }, + { + "epoch": 2.313063718359034, + "flow_matching_loss": 0.034127180092036725, + "grad_norm": 0.4215516448020935, + "learning_rate": 2.9978369299807372e-05, + "loss": 0.083, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019799228757619858, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009786408394575119, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04938486628234386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9892241537570954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7950, + "weighted_flow_matching_loss": 0.034127180092036725, + "weighted_rkd_loss": 0.04885269403457641 + }, + { + "epoch": 2.3159732324701774, + "flow_matching_loss": 0.036404764652252196, + "grad_norm": 0.45733365416526794, + "learning_rate": 2.997814681725463e-05, + "loss": 0.0854, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019776845909655096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009960396587848664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04951408877968788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9891971051692963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7960, + "weighted_flow_matching_loss": 0.036404764652252196, + "weighted_rkd_loss": 0.04897919297218323 + }, + { + "epoch": 2.318882746581321, + "flow_matching_loss": 0.034825305081903936, + "grad_norm": 0.4808095395565033, + "learning_rate": 2.9977923197210493e-05, + "loss": 0.0854, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020571497455239295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010010335221886634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05115332938730717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9891700208187103, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7970, + "weighted_flow_matching_loss": 0.034825305081903936, + "weighted_rkd_loss": 0.050599342584609984 + }, + { + "epoch": 2.3217922606924644, + "flow_matching_loss": 0.0346984826028347, + "grad_norm": 0.41793301701545715, + "learning_rate": 2.9977698439691927e-05, + "loss": 0.0798, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018259059451520444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009128777589648963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04564689621329308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9891429126262665, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7980, + "weighted_flow_matching_loss": 0.0346984826028347, + "weighted_rkd_loss": 0.045151307433843615 + }, + { + "epoch": 2.3247017748036076, + "flow_matching_loss": 0.03263101540505886, + "grad_norm": 0.35841408371925354, + "learning_rate": 2.9977472544716015e-05, + "loss": 0.0817, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019838506169617177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009956157486885785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049633169919252394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9891157627105713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7990, + "weighted_flow_matching_loss": 0.03263101540505886, + "weighted_rkd_loss": 0.04909295551478863 + }, + { + "epoch": 2.3276112889147513, + "flow_matching_loss": 0.03676234856247902, + "grad_norm": 0.4159061014652252, + "learning_rate": 2.9977245512299904e-05, + "loss": 0.0828, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01864035949110985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009235195675864815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0465159147977829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9890885770320892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8000, + "weighted_flow_matching_loss": 0.03676234856247902, + "weighted_rkd_loss": 0.04600837118923664 + }, + { + "epoch": 2.3305208030258946, + "flow_matching_loss": 0.037691953033208846, + "grad_norm": 0.3194721043109894, + "learning_rate": 2.9977017342460836e-05, + "loss": 0.0848, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01932062366977334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008971197251230478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04761244468390942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9890613675117492, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8010, + "weighted_flow_matching_loss": 0.037691953033208846, + "weighted_rkd_loss": 0.04709164090454578 + }, + { + "epoch": 2.3334303171370383, + "flow_matching_loss": 0.03172155786305666, + "grad_norm": 0.36009857058525085, + "learning_rate": 2.997678803521614e-05, + "loss": 0.0795, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019511335529386997, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009247452113777399, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048270123079419136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.989034116268158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8020, + "weighted_flow_matching_loss": 0.03172155786305666, + "weighted_rkd_loss": 0.04774081595242023 + }, + { + "epoch": 2.3363398312481816, + "flow_matching_loss": 0.03502483908087015, + "grad_norm": 0.3566228151321411, + "learning_rate": 2.9976557590583236e-05, + "loss": 0.0799, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018343327287584543, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008725362643599511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04541201777756214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9890068352222443, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8030, + "weighted_flow_matching_loss": 0.03502483908087015, + "weighted_rkd_loss": 0.044912799075245854 + }, + { + "epoch": 2.339249345359325, + "flow_matching_loss": 0.034998397529125216, + "grad_norm": 0.404030442237854, + "learning_rate": 2.9976326008579614e-05, + "loss": 0.0831, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019541496224701405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009565055277198553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04864804781973362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9889795243740082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8040, + "weighted_flow_matching_loss": 0.034998397529125216, + "weighted_rkd_loss": 0.04811191968619823 + }, + { + "epoch": 2.3421588594704685, + "flow_matching_loss": 0.03507908936589956, + "grad_norm": 0.31265735626220703, + "learning_rate": 2.997609328922287e-05, + "loss": 0.0833, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01959014404565096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009602332068607212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04878261983394623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9889521718025207, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8050, + "weighted_flow_matching_loss": 0.03507908936589956, + "weighted_rkd_loss": 0.0482436865568161 + }, + { + "epoch": 2.345068373581612, + "flow_matching_loss": 0.03461675979197025, + "grad_norm": 0.39279478788375854, + "learning_rate": 2.9975859432530673e-05, + "loss": 0.0825, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01955568790435791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0092696787789464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04838105402886868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9889248013496399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8060, + "weighted_flow_matching_loss": 0.03461675979197025, + "weighted_rkd_loss": 0.04784521982073784 + }, + { + "epoch": 2.347977887692755, + "flow_matching_loss": 0.037731108255684376, + "grad_norm": 0.39964187145233154, + "learning_rate": 2.9975624438520785e-05, + "loss": 0.0837, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018816681765019893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008888363325968384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04652172699570656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9888973832130432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8070, + "weighted_flow_matching_loss": 0.037731108255684376, + "weighted_rkd_loss": 0.04600520208477974 + }, + { + "epoch": 2.3508874018038988, + "flow_matching_loss": 0.03545483890920877, + "grad_norm": 0.49952664971351624, + "learning_rate": 2.997538830721105e-05, + "loss": 0.0826, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01927406173199415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009106317861005664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04765444062650204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9888699293136597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8080, + "weighted_flow_matching_loss": 0.03545483890920877, + "weighted_rkd_loss": 0.04712402187287808 + }, + { + "epoch": 2.353796915915042, + "flow_matching_loss": 0.03496478646993637, + "grad_norm": 0.42646095156669617, + "learning_rate": 2.9975151038619403e-05, + "loss": 0.0808, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018641811050474643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009098319616168738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04638194255530834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9888424515724182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8090, + "weighted_flow_matching_loss": 0.03496478646993637, + "weighted_rkd_loss": 0.045864424854516986 + }, + { + "epoch": 2.3567064300261857, + "flow_matching_loss": 0.03625951111316681, + "grad_norm": 0.452077716588974, + "learning_rate": 2.997491263276386e-05, + "loss": 0.0849, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01980427335947752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009609936224296688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0492184828966856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9888149261474609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8100, + "weighted_flow_matching_loss": 0.03625951111316681, + "weighted_rkd_loss": 0.048667972162365916 + }, + { + "epoch": 2.359615944137329, + "flow_matching_loss": 0.03954915888607502, + "grad_norm": 0.34244945645332336, + "learning_rate": 2.997467308966253e-05, + "loss": 0.086, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018781921919435264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009422843577340245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04698668755590916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9887873828411102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8110, + "weighted_flow_matching_loss": 0.03954915888607502, + "weighted_rkd_loss": 0.04645987041294575 + }, + { + "epoch": 2.3625254582484727, + "flow_matching_loss": 0.034932082146406175, + "grad_norm": 0.39588549733161926, + "learning_rate": 2.9974432409333605e-05, + "loss": 0.0814, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01891912231221795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009135704534128309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04697394892573357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9887597978115081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8120, + "weighted_flow_matching_loss": 0.034932082146406175, + "weighted_rkd_loss": 0.046445968002080916 + }, + { + "epoch": 2.365434972359616, + "flow_matching_loss": 0.03482930101454258, + "grad_norm": 0.4403431713581085, + "learning_rate": 2.997419059179536e-05, + "loss": 0.0829, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019686351530253887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00924305822700262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04861576072871685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9887321829795838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8130, + "weighted_flow_matching_loss": 0.03482930101454258, + "weighted_rkd_loss": 0.048067957162857056 + }, + { + "epoch": 2.3683444864707592, + "flow_matching_loss": 0.034484421275556086, + "grad_norm": 0.39588358998298645, + "learning_rate": 2.997394763706616e-05, + "loss": 0.0804, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018598552234470843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009237864101305605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046434969455003736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9887045323848724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8140, + "weighted_flow_matching_loss": 0.034484421275556086, + "weighted_rkd_loss": 0.04591045267879963 + }, + { + "epoch": 2.371254000581903, + "flow_matching_loss": 0.03378614541143179, + "grad_norm": 0.40260621905326843, + "learning_rate": 2.9973703545164458e-05, + "loss": 0.0807, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018995144870132208, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00942891975864768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047419209778308866, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9886768460273743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8150, + "weighted_flow_matching_loss": 0.03378614541143179, + "weighted_rkd_loss": 0.04688226580619812 + }, + { + "epoch": 2.374163514693046, + "flow_matching_loss": 0.035284917801618576, + "grad_norm": 0.3905757963657379, + "learning_rate": 2.9973458316108788e-05, + "loss": 0.0797, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01820643786340952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008535281568765641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04494815692305565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9886491298675537, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8160, + "weighted_flow_matching_loss": 0.035284917801618576, + "weighted_rkd_loss": 0.04443795755505562 + }, + { + "epoch": 2.3770730288041895, + "flow_matching_loss": 0.0328581802546978, + "grad_norm": 0.36151155829429626, + "learning_rate": 2.9973211949917773e-05, + "loss": 0.08, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019110383279621602, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009415745129808783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0476365115493536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9886213779449463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8170, + "weighted_flow_matching_loss": 0.0328581802546978, + "weighted_rkd_loss": 0.04709446802735329 + }, + { + "epoch": 2.379982542915333, + "flow_matching_loss": 0.03501864280551672, + "grad_norm": 0.4010675847530365, + "learning_rate": 2.997296444661013e-05, + "loss": 0.0851, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020315498858690263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009999410342425108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050630408152937886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9885935962200165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8180, + "weighted_flow_matching_loss": 0.03501864280551672, + "weighted_rkd_loss": 0.05005291663110256 + }, + { + "epoch": 2.3828920570264764, + "flow_matching_loss": 0.035604358091950415, + "grad_norm": 0.4123130738735199, + "learning_rate": 2.9972715806204648e-05, + "loss": 0.0849, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019867235235869883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010168700199574231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049903170019388196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9885657787322998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8190, + "weighted_flow_matching_loss": 0.035604358091950415, + "weighted_rkd_loss": 0.049332547932863235 + }, + { + "epoch": 2.38580157113762, + "flow_matching_loss": 0.03280797023326158, + "grad_norm": 0.40503478050231934, + "learning_rate": 2.9972466028720212e-05, + "loss": 0.0777, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0179202301427722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009537400817498565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04537786021828651, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9885379254817963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8200, + "weighted_flow_matching_loss": 0.03280797023326158, + "weighted_rkd_loss": 0.04485774710774422 + }, + { + "epoch": 2.3887110852487634, + "flow_matching_loss": 0.03534561023116112, + "grad_norm": 0.3936230540275574, + "learning_rate": 2.997221511417579e-05, + "loss": 0.0845, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020008783042430877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009751196578145027, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04976876266300678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9885100424289703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8210, + "weighted_flow_matching_loss": 0.03534561023116112, + "weighted_rkd_loss": 0.04919690191745758 + }, + { + "epoch": 2.391620599359907, + "flow_matching_loss": 0.031707015447318555, + "grad_norm": 0.3452717959880829, + "learning_rate": 2.9971963062590435e-05, + "loss": 0.0758, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017717146314680576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009159100567921996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04459339268505573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9884821236133575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8220, + "weighted_flow_matching_loss": 0.031707015447318555, + "weighted_rkd_loss": 0.044079771265387535 + }, + { + "epoch": 2.3945301134710504, + "flow_matching_loss": 0.03326019924134016, + "grad_norm": 0.354267954826355, + "learning_rate": 2.9971709873983297e-05, + "loss": 0.0816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019643097929656506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009588131308555603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04887432716786862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9884541749954223, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8230, + "weighted_flow_matching_loss": 0.03326019924134016, + "weighted_rkd_loss": 0.048310057073831555 + }, + { + "epoch": 2.3974396275821936, + "flow_matching_loss": 0.034280334040522574, + "grad_norm": 0.4168093502521515, + "learning_rate": 2.9971455548373595e-05, + "loss": 0.082, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01950067449361086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009302283916622401, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04830363281071186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9884261786937714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8240, + "weighted_flow_matching_loss": 0.034280334040522574, + "weighted_rkd_loss": 0.04774455726146698 + }, + { + "epoch": 2.4003491416933374, + "flow_matching_loss": 0.032736736722290515, + "grad_norm": 0.36961784958839417, + "learning_rate": 2.9971200085780648e-05, + "loss": 0.0767, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01792551390826702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00861607058905065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044467099010944366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9883981645107269, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8250, + "weighted_flow_matching_loss": 0.032736736722290515, + "weighted_rkd_loss": 0.043951195850968364 + }, + { + "epoch": 2.4032586558044806, + "flow_matching_loss": 0.03642626702785492, + "grad_norm": 0.3782845735549927, + "learning_rate": 2.997094348622386e-05, + "loss": 0.0834, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019089404959231615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009306679153814912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047485489025712015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9883701205253601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8260, + "weighted_flow_matching_loss": 0.03642626702785492, + "weighted_rkd_loss": 0.04693325608968735 + }, + { + "epoch": 2.406168169915624, + "flow_matching_loss": 0.03641131576150656, + "grad_norm": 0.4026985764503479, + "learning_rate": 2.9970685749722712e-05, + "loss": 0.0836, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019216130673885345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009345797682181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04777805879712105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9883420407772064, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8270, + "weighted_flow_matching_loss": 0.03641131576150656, + "weighted_rkd_loss": 0.047221063449978826 + }, + { + "epoch": 2.4090776840267676, + "flow_matching_loss": 0.034659940749406815, + "grad_norm": 0.3853141963481903, + "learning_rate": 2.9970426876296775e-05, + "loss": 0.08, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01852312097325921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0087905615568161, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04583680368959904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9883139133453369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8280, + "weighted_flow_matching_loss": 0.034659940749406815, + "weighted_rkd_loss": 0.045301153883337976 + }, + { + "epoch": 2.411987198137911, + "flow_matching_loss": 0.034066824987530706, + "grad_norm": 0.35391879081726074, + "learning_rate": 2.9970166865965716e-05, + "loss": 0.0799, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018638418801128865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009142757393419743, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046419594436883926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9882857620716095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8290, + "weighted_flow_matching_loss": 0.034066824987530706, + "weighted_rkd_loss": 0.04587583243846893 + }, + { + "epoch": 2.4148967122490546, + "flow_matching_loss": 0.036250025033950806, + "grad_norm": 0.4677330553531647, + "learning_rate": 2.996990571874928e-05, + "loss": 0.082, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018540828395634888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00917837331071496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046260029822587964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9882575809955597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8300, + "weighted_flow_matching_loss": 0.036250025033950806, + "weighted_rkd_loss": 0.04571683742105961 + }, + { + "epoch": 2.417806226360198, + "flow_matching_loss": 0.035448592342436316, + "grad_norm": 0.39803802967071533, + "learning_rate": 2.9969643434667293e-05, + "loss": 0.0813, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01849896293133497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009370742551982402, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04636866822838783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9882293581962586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8310, + "weighted_flow_matching_loss": 0.035448592342436316, + "weighted_rkd_loss": 0.04582289420068264 + }, + { + "epoch": 2.4207157404713415, + "flow_matching_loss": 0.03422524780035019, + "grad_norm": 0.3260091543197632, + "learning_rate": 2.996938001373968e-05, + "loss": 0.0786, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018061602860689162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008814338454976678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04493754468858242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.988201105594635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8320, + "weighted_flow_matching_loss": 0.03422524780035019, + "weighted_rkd_loss": 0.04440734013915062 + }, + { + "epoch": 2.423625254582485, + "flow_matching_loss": 0.03285160716623068, + "grad_norm": 0.35813435912132263, + "learning_rate": 2.9969115455986443e-05, + "loss": 0.0764, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017650023568421602, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008736405335366727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04403645247220993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9881728231906891, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8330, + "weighted_flow_matching_loss": 0.03285160716623068, + "weighted_rkd_loss": 0.04351565353572369 + }, + { + "epoch": 2.426534768693628, + "flow_matching_loss": 0.03596883397549391, + "grad_norm": 0.37781742215156555, + "learning_rate": 2.996884976142768e-05, + "loss": 0.0807, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01808301480486989, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00909710181877017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045263131335377696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9881445050239563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8340, + "weighted_flow_matching_loss": 0.03596883397549391, + "weighted_rkd_loss": 0.04472649767994881 + }, + { + "epoch": 2.4294442828047718, + "flow_matching_loss": 0.03373211342841387, + "grad_norm": 0.4550575613975525, + "learning_rate": 2.996858293008356e-05, + "loss": 0.0791, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01845589019358158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008979522529989481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04589130319654942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9881161451339722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8350, + "weighted_flow_matching_loss": 0.03373211342841387, + "weighted_rkd_loss": 0.04534591995179653 + }, + { + "epoch": 2.432353796915915, + "flow_matching_loss": 0.033869500271975994, + "grad_norm": 0.43322283029556274, + "learning_rate": 2.9968314961974346e-05, + "loss": 0.0838, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020265922881662846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010031122341752052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050562968105077745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9880877614021302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8360, + "weighted_flow_matching_loss": 0.033869500271975994, + "weighted_rkd_loss": 0.049960635229945184 + }, + { + "epoch": 2.4352633110270583, + "flow_matching_loss": 0.03545719888061285, + "grad_norm": 0.36880984902381897, + "learning_rate": 2.99680458571204e-05, + "loss": 0.0846, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019853885285556316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010026180883869529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04973395131528378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9880593419075012, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8370, + "weighted_flow_matching_loss": 0.03545719888061285, + "weighted_rkd_loss": 0.04914012737572193 + }, + { + "epoch": 2.438172825138202, + "flow_matching_loss": 0.034685599990189074, + "grad_norm": 0.4161757528781891, + "learning_rate": 2.996777561554215e-05, + "loss": 0.0803, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018685541301965713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008827600302174688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046198683232069014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9880308926105499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8380, + "weighted_flow_matching_loss": 0.034685599990189074, + "weighted_rkd_loss": 0.04564572013914585 + }, + { + "epoch": 2.4410823392493453, + "flow_matching_loss": 0.0326385049149394, + "grad_norm": 0.4313952624797821, + "learning_rate": 2.9967504237260117e-05, + "loss": 0.0794, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019062502682209014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009199934871867299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04732493981719017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9880023956298828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8390, + "weighted_flow_matching_loss": 0.0326385049149394, + "weighted_rkd_loss": 0.0467571571469307 + }, + { + "epoch": 2.443991853360489, + "flow_matching_loss": 0.0319276312366128, + "grad_norm": 0.36564901471138, + "learning_rate": 2.9967231722294916e-05, + "loss": 0.0776, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018682977370917796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008885886147618293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04625184088945389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9879738867282868, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8400, + "weighted_flow_matching_loss": 0.0319276312366128, + "weighted_rkd_loss": 0.04569560550153255 + }, + { + "epoch": 2.4469013674716322, + "flow_matching_loss": 0.0341646334156394, + "grad_norm": 0.42845121026039124, + "learning_rate": 2.9966958070667245e-05, + "loss": 0.0806, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018931428156793117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009177600499242544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047040455788373944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9879453182220459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8410, + "weighted_flow_matching_loss": 0.0341646334156394, + "weighted_rkd_loss": 0.046473415568470955 + }, + { + "epoch": 2.4498108815827755, + "flow_matching_loss": 0.03499258756637573, + "grad_norm": 0.42143598198890686, + "learning_rate": 2.9966683282397877e-05, + "loss": 0.0819, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019246533699333667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009011560445651412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047504627332091334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9879167556762696, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8420, + "weighted_flow_matching_loss": 0.03499258756637573, + "weighted_rkd_loss": 0.04693063274025917 + }, + { + "epoch": 2.452720395693919, + "flow_matching_loss": 0.03322344161570072, + "grad_norm": 0.38426923751831055, + "learning_rate": 2.9966407357507688e-05, + "loss": 0.079, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018736667186021804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00883379471488297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04630712941288948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.987888103723526, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8430, + "weighted_flow_matching_loss": 0.03322344161570072, + "weighted_rkd_loss": 0.045746258646249774 + }, + { + "epoch": 2.4556299098050625, + "flow_matching_loss": 0.031757927127182485, + "grad_norm": 0.3717184364795685, + "learning_rate": 2.9966130296017625e-05, + "loss": 0.0784, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019084359891712664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00905601060949266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047224730253219604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9878594636917114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8440, + "weighted_flow_matching_loss": 0.031757927127182485, + "weighted_rkd_loss": 0.04665140062570572 + }, + { + "epoch": 2.458539423916206, + "flow_matching_loss": 0.0313302718102932, + "grad_norm": 0.3914903700351715, + "learning_rate": 2.9965852097948738e-05, + "loss": 0.0785, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019135756138712167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009455577190965413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04772708937525749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9878307819366455, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8450, + "weighted_flow_matching_loss": 0.0313302718102932, + "weighted_rkd_loss": 0.04714631102979183 + }, + { + "epoch": 2.4614489380273494, + "flow_matching_loss": 0.033807830326259135, + "grad_norm": 0.38936060667037964, + "learning_rate": 2.9965572763322148e-05, + "loss": 0.0793, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018635959737002848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008742345171049238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04601426422595978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9878020524978638, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8460, + "weighted_flow_matching_loss": 0.033807830326259135, + "weighted_rkd_loss": 0.04545298591256142 + }, + { + "epoch": 2.4643584521384927, + "flow_matching_loss": 0.034736533649265766, + "grad_norm": 0.4230800271034241, + "learning_rate": 2.9965292292159076e-05, + "loss": 0.0778, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0175975089892745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008389622997492551, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04358464106917381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9877732992172241, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8470, + "weighted_flow_matching_loss": 0.034736533649265766, + "weighted_rkd_loss": 0.04305174835026264 + }, + { + "epoch": 2.4672679662496364, + "flow_matching_loss": 0.03379690181463957, + "grad_norm": 0.33557239174842834, + "learning_rate": 2.9965010684480812e-05, + "loss": 0.081, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019235746189951896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009332140488550068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04780363216996193, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9877445220947265, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8480, + "weighted_flow_matching_loss": 0.03379690181463957, + "weighted_rkd_loss": 0.047217792645096776 + }, + { + "epoch": 2.4701774803607797, + "flow_matching_loss": 0.03513956479728222, + "grad_norm": 0.38519981503486633, + "learning_rate": 2.996472794030875e-05, + "loss": 0.0816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018975333869457246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00906589231453836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047016560286283496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9877156913280487, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8490, + "weighted_flow_matching_loss": 0.03513956479728222, + "weighted_rkd_loss": 0.046439026668667796 + }, + { + "epoch": 2.4730869944719234, + "flow_matching_loss": 0.03325784858316183, + "grad_norm": 0.37962010502815247, + "learning_rate": 2.9964444059664358e-05, + "loss": 0.0771, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017784757539629936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008809128263965248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044378644227981566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.987686836719513, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8500, + "weighted_flow_matching_loss": 0.03325784858316183, + "weighted_rkd_loss": 0.043832216039299966 + }, + { + "epoch": 2.4759965085830666, + "flow_matching_loss": 0.0349629694595933, + "grad_norm": 0.3885883390903473, + "learning_rate": 2.9964159042569194e-05, + "loss": 0.0836, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01992366127669811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009444732265546919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04929205439984798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9876579582691193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8510, + "weighted_flow_matching_loss": 0.0349629694595933, + "weighted_rkd_loss": 0.04868369922041893 + }, + { + "epoch": 2.47890602269421, + "flow_matching_loss": 0.03640696909278631, + "grad_norm": 0.39680808782577515, + "learning_rate": 2.996387288904491e-05, + "loss": 0.0796, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017670161090791225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008386636292561889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04372695907950401, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9876290261745453, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8520, + "weighted_flow_matching_loss": 0.03640696909278631, + "weighted_rkd_loss": 0.04318600706756115 + }, + { + "epoch": 2.4818155368053536, + "flow_matching_loss": 0.03236093670129776, + "grad_norm": 0.3758259117603302, + "learning_rate": 2.996358559911323e-05, + "loss": 0.0773, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01835124483332038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008780322456732392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045482812076807023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9876000702381134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8530, + "weighted_flow_matching_loss": 0.03236093670129776, + "weighted_rkd_loss": 0.04491880275309086 + }, + { + "epoch": 2.484725050916497, + "flow_matching_loss": 0.03530341405421496, + "grad_norm": 0.44934526085853577, + "learning_rate": 2.9963297172795975e-05, + "loss": 0.0823, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018990648537874223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0095659582875669, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04754725508391857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9875710785388947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8540, + "weighted_flow_matching_loss": 0.03530341405421496, + "weighted_rkd_loss": 0.04695629402995109 + }, + { + "epoch": 2.4876345650276406, + "flow_matching_loss": 0.03259676992893219, + "grad_norm": 0.3425244092941284, + "learning_rate": 2.996300761011505e-05, + "loss": 0.0774, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018282982520759106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008821992250159384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04538795687258244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.987542062997818, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8550, + "weighted_flow_matching_loss": 0.03259676992893219, + "weighted_rkd_loss": 0.04482249766588211 + }, + { + "epoch": 2.490544079138784, + "flow_matching_loss": 0.034996080212295055, + "grad_norm": 0.40456491708755493, + "learning_rate": 2.9962716911092444e-05, + "loss": 0.0799, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01823426838964224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009017189964652061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04548572674393654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9875130116939544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8560, + "weighted_flow_matching_loss": 0.034996080212295055, + "weighted_rkd_loss": 0.044917769357562067 + }, + { + "epoch": 2.493453593249927, + "flow_matching_loss": 0.033973250165581705, + "grad_norm": 0.394661009311676, + "learning_rate": 2.996242507575023e-05, + "loss": 0.0771, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017437274288386106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008776436559855938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04365098588168621, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9874839186668396, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8570, + "weighted_flow_matching_loss": 0.033973250165581705, + "weighted_rkd_loss": 0.043104635551571846 + }, + { + "epoch": 2.496363107361071, + "flow_matching_loss": 0.035110085643827914, + "grad_norm": 0.3695663809776306, + "learning_rate": 2.996213210411058e-05, + "loss": 0.0783, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017586412373930216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008575757313519717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043748582154512404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9874547958374024, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8580, + "weighted_flow_matching_loss": 0.035110085643827914, + "weighted_rkd_loss": 0.04319975413382053 + }, + { + "epoch": 2.499272621472214, + "flow_matching_loss": 0.03281041923910379, + "grad_norm": 0.3841492533683777, + "learning_rate": 2.9961837996195737e-05, + "loss": 0.0773, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01818369925022125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00872502950951457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045092427730560304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9874256551265717, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8590, + "weighted_flow_matching_loss": 0.03281041923910379, + "weighted_rkd_loss": 0.04452540576457977 + }, + { + "epoch": 2.5021821355833573, + "flow_matching_loss": 0.032722844555974004, + "grad_norm": 0.34833186864852905, + "learning_rate": 2.996154275202804e-05, + "loss": 0.0767, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017946198396384717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008607866801321506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04450026378035545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9873964488506317, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8600, + "weighted_flow_matching_loss": 0.032722844555974004, + "weighted_rkd_loss": 0.04393942952156067 + }, + { + "epoch": 2.505091649694501, + "flow_matching_loss": 0.033976292423903945, + "grad_norm": 0.3781873285770416, + "learning_rate": 2.99612463716299e-05, + "loss": 0.0802, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018810020852833988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009218700043857098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04683874174952507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9873672425746918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8610, + "weighted_flow_matching_loss": 0.033976292423903945, + "weighted_rkd_loss": 0.046247028186917304 + }, + { + "epoch": 2.5080011638056443, + "flow_matching_loss": 0.03605169244110584, + "grad_norm": 0.3834382891654968, + "learning_rate": 2.9960948855023844e-05, + "loss": 0.081, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0184133630245924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008744754735380411, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04557148069143295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9873379707336426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8620, + "weighted_flow_matching_loss": 0.03605169244110584, + "weighted_rkd_loss": 0.04499446600675583 + }, + { + "epoch": 2.510910677916788, + "flow_matching_loss": 0.03359951227903366, + "grad_norm": 0.3950725495815277, + "learning_rate": 2.9960650202232453e-05, + "loss": 0.0754, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01725719906389713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007853899896144868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042368298023939134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9873086869716644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8630, + "weighted_flow_matching_loss": 0.03359951227903366, + "weighted_rkd_loss": 0.041830582171678544 + }, + { + "epoch": 2.5138201920279313, + "flow_matching_loss": 0.03157621826976538, + "grad_norm": 0.4230753779411316, + "learning_rate": 2.996035041327841e-05, + "loss": 0.0772, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018605322949588297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008956664660945534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04616731144487858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9872793674468994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8640, + "weighted_flow_matching_loss": 0.03157621826976538, + "weighted_rkd_loss": 0.045580020919442174 + }, + { + "epoch": 2.516729706139075, + "flow_matching_loss": 0.03289705794304609, + "grad_norm": 0.3287864327430725, + "learning_rate": 2.9960049488184483e-05, + "loss": 0.0749, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01714058071374893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008295885380357504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042577047273516654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.987250006198883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8650, + "weighted_flow_matching_loss": 0.03289705794304609, + "weighted_rkd_loss": 0.04203421138226986 + }, + { + "epoch": 2.5196392202502182, + "flow_matching_loss": 0.03391328658908606, + "grad_norm": 0.3623948097229004, + "learning_rate": 2.9959747426973522e-05, + "loss": 0.0769, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017487179674208164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008521266421303153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04349562600255012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9872206270694732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8660, + "weighted_flow_matching_loss": 0.03391328658908606, + "weighted_rkd_loss": 0.042939775809645656 + }, + { + "epoch": 2.5225487343613615, + "flow_matching_loss": 0.03477519024163485, + "grad_norm": 0.35128241777420044, + "learning_rate": 2.995944422966847e-05, + "loss": 0.0792, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01810242738574743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008796803746372461, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0450016587972641, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9871911942958832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8670, + "weighted_flow_matching_loss": 0.03477519024163485, + "weighted_rkd_loss": 0.044425243511796 + }, + { + "epoch": 2.525458248472505, + "flow_matching_loss": 0.03189622461795807, + "grad_norm": 0.32722207903862, + "learning_rate": 2.9959139896292354e-05, + "loss": 0.0748, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01764680529013276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008213702356442809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04350731298327446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9871617436408997, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8680, + "weighted_flow_matching_loss": 0.03189622461795807, + "weighted_rkd_loss": 0.042948779463768 + }, + { + "epoch": 2.5283677625836485, + "flow_matching_loss": 0.03093375787138939, + "grad_norm": 0.38248422741889954, + "learning_rate": 2.9958834426868286e-05, + "loss": 0.0739, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017498134449124337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008501359075307847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04349762797355652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9871322512626648, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8690, + "weighted_flow_matching_loss": 0.03093375787138939, + "weighted_rkd_loss": 0.0429379228502512 + }, + { + "epoch": 2.5312772766947917, + "flow_matching_loss": 0.03324422314763069, + "grad_norm": 0.42485323548316956, + "learning_rate": 2.9958527821419462e-05, + "loss": 0.0766, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017845288012176753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008224898530170322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04391547441482544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9871027350425721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8700, + "weighted_flow_matching_loss": 0.03324422314763069, + "weighted_rkd_loss": 0.04334908276796341 + }, + { + "epoch": 2.5341867908059355, + "flow_matching_loss": 0.03390802182257176, + "grad_norm": 0.3867221176624298, + "learning_rate": 2.9958220079969172e-05, + "loss": 0.0741, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01650483664125204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0077151449397206305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040724818408489225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9870731711387635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8710, + "weighted_flow_matching_loss": 0.03390802182257176, + "weighted_rkd_loss": 0.040198370814323425 + }, + { + "epoch": 2.5370963049170787, + "flow_matching_loss": 0.030991026759147645, + "grad_norm": 0.3376215398311615, + "learning_rate": 2.995791120254078e-05, + "loss": 0.0721, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01688890513032675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0078309983946383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04160880856215954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9870435893535614, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8720, + "weighted_flow_matching_loss": 0.030991026759147645, + "weighted_rkd_loss": 0.04106970690190792 + }, + { + "epoch": 2.5400058190282224, + "flow_matching_loss": 0.03416402880102396, + "grad_norm": 0.38728511333465576, + "learning_rate": 2.9957601189157743e-05, + "loss": 0.0756, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017006146349012853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007987844850867986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042000137642025946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9870139598846436, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8730, + "weighted_flow_matching_loss": 0.03416402880102396, + "weighted_rkd_loss": 0.04145474396646023 + }, + { + "epoch": 2.5429153331393657, + "flow_matching_loss": 0.036530463770031926, + "grad_norm": 0.40825626254081726, + "learning_rate": 2.9957290039843608e-05, + "loss": 0.0813, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018231597263365984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008894539810717106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04535773433744907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9869843065738678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8740, + "weighted_flow_matching_loss": 0.036530463770031926, + "weighted_rkd_loss": 0.04476740211248398 + }, + { + "epoch": 2.5458248472505094, + "flow_matching_loss": 0.035729774832725526, + "grad_norm": 0.386592835187912, + "learning_rate": 2.9956977754622007e-05, + "loss": 0.08, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01810309672728181, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008621774706989526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04482796788215637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9869546115398407, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8750, + "weighted_flow_matching_loss": 0.035729774832725526, + "weighted_rkd_loss": 0.044243162125349046 + }, + { + "epoch": 2.5487343613616527, + "flow_matching_loss": 0.03484733421355486, + "grad_norm": 0.33209866285324097, + "learning_rate": 2.995666433351665e-05, + "loss": 0.0775, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017486955411732195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00824969345703721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043223604559898376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9869248926639557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8760, + "weighted_flow_matching_loss": 0.03484733421355486, + "weighted_rkd_loss": 0.04265844747424126 + }, + { + "epoch": 2.551643875472796, + "flow_matching_loss": 0.034991465881466864, + "grad_norm": 0.41071391105651855, + "learning_rate": 2.9956349776551343e-05, + "loss": 0.0796, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018252368643879892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008708658860996366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045213396474719045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9868951439857483, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8770, + "weighted_flow_matching_loss": 0.034991465881466864, + "weighted_rkd_loss": 0.04462086893618107 + }, + { + "epoch": 2.5545533895839396, + "flow_matching_loss": 0.03182583022862673, + "grad_norm": 0.3513837158679962, + "learning_rate": 2.9956034083749975e-05, + "loss": 0.0758, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017938990239053966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008677634550258518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04455561526119709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9868653416633606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8780, + "weighted_flow_matching_loss": 0.03182583022862673, + "weighted_rkd_loss": 0.04397042542695999 + }, + { + "epoch": 2.557462903695083, + "flow_matching_loss": 0.03610475994646549, + "grad_norm": 0.48639318346977234, + "learning_rate": 2.9955717255136517e-05, + "loss": 0.0813, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018298765644431114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00922982902266085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04582736045122147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.986835527420044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8790, + "weighted_flow_matching_loss": 0.03610475994646549, + "weighted_rkd_loss": 0.045224054530262944 + }, + { + "epoch": 2.560372417806226, + "flow_matching_loss": 0.03354213256388903, + "grad_norm": 0.35035672783851624, + "learning_rate": 2.9955399290735035e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017406134959310295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008472336130216718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04328460618853569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9868056654930115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8800, + "weighted_flow_matching_loss": 0.03354213256388903, + "weighted_rkd_loss": 0.04271349348127842 + }, + { + "epoch": 2.56328193191737, + "flow_matching_loss": 0.034662524983286855, + "grad_norm": 0.4451592266559601, + "learning_rate": 2.995508019056967e-05, + "loss": 0.0807, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018701996468007565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00925956517457962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04666355811059475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9867757797241211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8810, + "weighted_flow_matching_loss": 0.034662524983286855, + "weighted_rkd_loss": 0.04604645110666752 + }, + { + "epoch": 2.566191446028513, + "flow_matching_loss": 0.03297423273324966, + "grad_norm": 0.4033888876438141, + "learning_rate": 2.9954759954664663e-05, + "loss": 0.0802, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01917344145476818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00954480143263936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04789168499410153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9867458462715148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8820, + "weighted_flow_matching_loss": 0.03297423273324966, + "weighted_rkd_loss": 0.04725695438683033 + }, + { + "epoch": 2.569100960139657, + "flow_matching_loss": 0.031472759880125524, + "grad_norm": 0.44643521308898926, + "learning_rate": 2.9954438583044328e-05, + "loss": 0.076, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018244906049221753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008628893084824086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04511870518326759, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9867158949375152, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8830, + "weighted_flow_matching_loss": 0.031472759880125524, + "weighted_rkd_loss": 0.04451931789517403 + }, + { + "epoch": 2.5720104742508, + "flow_matching_loss": 0.03230247646570206, + "grad_norm": 0.3644046485424042, + "learning_rate": 2.995411607573308e-05, + "loss": 0.0784, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0187172275967896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009286284586414695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04672073982656002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9866859018802643, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8840, + "weighted_flow_matching_loss": 0.03230247646570206, + "weighted_rkd_loss": 0.046098703145980836 + }, + { + "epoch": 2.574919988361944, + "flow_matching_loss": 0.03167949840426445, + "grad_norm": 0.3815555274486542, + "learning_rate": 2.9953792432755397e-05, + "loss": 0.0751, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017650453187525273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008746705250814557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044047611951828006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.986655879020691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8850, + "weighted_flow_matching_loss": 0.03167949840426445, + "weighted_rkd_loss": 0.0434598196297884 + }, + { + "epoch": 2.577829502473087, + "flow_matching_loss": 0.03211899287998676, + "grad_norm": 0.45303964614868164, + "learning_rate": 2.9953467654135868e-05, + "loss": 0.0775, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018321071099489928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009331086743623018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04597322903573513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9866258263587951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8860, + "weighted_flow_matching_loss": 0.03211899287998676, + "weighted_rkd_loss": 0.045358354598283766 + }, + { + "epoch": 2.5807390165842303, + "flow_matching_loss": 0.03367602415382862, + "grad_norm": 0.4330901503562927, + "learning_rate": 2.9953141739899155e-05, + "loss": 0.079, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018361840024590494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009186520520597697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045910200104117393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.986595743894577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8870, + "weighted_flow_matching_loss": 0.03367602415382862, + "weighted_rkd_loss": 0.045294782519340514 + }, + { + "epoch": 2.583648530695374, + "flow_matching_loss": 0.03275075629353523, + "grad_norm": 0.3942927420139313, + "learning_rate": 2.9952814690070008e-05, + "loss": 0.0759, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017678959760814905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008395345229655505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04375326484441757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9865656137466431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8880, + "weighted_flow_matching_loss": 0.03275075629353523, + "weighted_rkd_loss": 0.04316545017063618 + }, + { + "epoch": 2.5865580448065173, + "flow_matching_loss": 0.0343131598085165, + "grad_norm": 0.31530672311782837, + "learning_rate": 2.9952486504673265e-05, + "loss": 0.0775, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017402988206595183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008968928782269358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043774905055761336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9865354597568512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8890, + "weighted_flow_matching_loss": 0.0343131598085165, + "weighted_rkd_loss": 0.04318550750613213 + }, + { + "epoch": 2.5894675589176606, + "flow_matching_loss": 0.032794019766151906, + "grad_norm": 0.40769872069358826, + "learning_rate": 2.995215718373385e-05, + "loss": 0.077, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018073508981615305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008657571533694864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044804589450359346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9865052700042725, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8900, + "weighted_flow_matching_loss": 0.032794019766151906, + "weighted_rkd_loss": 0.04419996254146099 + }, + { + "epoch": 2.5923770730288043, + "flow_matching_loss": 0.03204831536859274, + "grad_norm": 0.4519806504249573, + "learning_rate": 2.9951826727276778e-05, + "loss": 0.0759, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017938878014683725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00858648233115673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044464238360524176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9864750444889069, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8910, + "weighted_flow_matching_loss": 0.03204831536859274, + "weighted_rkd_loss": 0.043862859904766086 + }, + { + "epoch": 2.5952865871399475, + "flow_matching_loss": 0.033675954304635526, + "grad_norm": 0.3889714479446411, + "learning_rate": 2.9951495135327133e-05, + "loss": 0.0735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016312294639647007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007705326471477747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04032991584390402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9864447832107544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8920, + "weighted_flow_matching_loss": 0.033675954304635526, + "weighted_rkd_loss": 0.039783237501978874 + }, + { + "epoch": 2.598196101251091, + "flow_matching_loss": 0.03472023233771324, + "grad_norm": 0.3440926969051361, + "learning_rate": 2.9951162407910107e-05, + "loss": 0.077, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017287997249513864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008256228407844901, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04283222332596779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.986414498090744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8930, + "weighted_flow_matching_loss": 0.03472023233771324, + "weighted_rkd_loss": 0.042250349372625354 + }, + { + "epoch": 2.6011056153622345, + "flow_matching_loss": 0.031688017770648, + "grad_norm": 0.39125803112983704, + "learning_rate": 2.995082854505096e-05, + "loss": 0.0731, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01701173037290573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008008565194904805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04203202575445175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9863841712474823, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8940, + "weighted_flow_matching_loss": 0.031688017770648, + "weighted_rkd_loss": 0.04145973846316338 + }, + { + "epoch": 2.604015129473378, + "flow_matching_loss": 0.03395451605319977, + "grad_norm": 0.43077871203422546, + "learning_rate": 2.9950493546775056e-05, + "loss": 0.0767, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0173890920355916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008512120274826885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04329030402004719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9863538205623626, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8950, + "weighted_flow_matching_loss": 0.03395451605319977, + "weighted_rkd_loss": 0.042699556797742844 + }, + { + "epoch": 2.6069246435845215, + "flow_matching_loss": 0.030329924076795578, + "grad_norm": 0.3537796139717102, + "learning_rate": 2.9950157413107834e-05, + "loss": 0.0758, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01837334604933858, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009354914026334881, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046101606264710424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9863234460353851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8960, + "weighted_flow_matching_loss": 0.030329924076795578, + "weighted_rkd_loss": 0.04547109603881836 + }, + { + "epoch": 2.6098341576956647, + "flow_matching_loss": 0.031754227541387084, + "grad_norm": 0.3941766619682312, + "learning_rate": 2.9949820144074813e-05, + "loss": 0.0756, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017859210167080163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008725761389359831, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04444418139755726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9862929940223694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8970, + "weighted_flow_matching_loss": 0.031754227541387084, + "weighted_rkd_loss": 0.04383498877286911 + }, + { + "epoch": 2.6127436718068084, + "flow_matching_loss": 0.034551605209708215, + "grad_norm": 0.3670554459095001, + "learning_rate": 2.9949481739701618e-05, + "loss": 0.0817, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019107714109122754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00954873776063323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04776416607201099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9862625479698182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8980, + "weighted_flow_matching_loss": 0.034551605209708215, + "weighted_rkd_loss": 0.04710799343883991 + }, + { + "epoch": 2.6156531859179517, + "flow_matching_loss": 0.036941443011164664, + "grad_norm": 0.3930760324001312, + "learning_rate": 2.994914220001394e-05, + "loss": 0.0794, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01728403242304921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008483609976246953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04305167496204376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.986232054233551, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8990, + "weighted_flow_matching_loss": 0.036941443011164664, + "weighted_rkd_loss": 0.04245895184576511 + }, + { + "epoch": 2.618562700029095, + "flow_matching_loss": 0.032465063780546186, + "grad_norm": 0.38637110590934753, + "learning_rate": 2.9948801525037567e-05, + "loss": 0.0728, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016448497772216797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008030417235568165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04092741310596466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.986201536655426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9000, + "weighted_flow_matching_loss": 0.032465063780546186, + "weighted_rkd_loss": 0.040362673625349996 + }, + { + "epoch": 2.6214722141402387, + "flow_matching_loss": 0.03258148413151503, + "grad_norm": 0.3732664883136749, + "learning_rate": 2.9948459714798376e-05, + "loss": 0.0759, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01769507201388478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008538078982383013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043928223103284834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9861709773540497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9010, + "weighted_flow_matching_loss": 0.03258148413151503, + "weighted_rkd_loss": 0.043320754915475844 + }, + { + "epoch": 2.624381728251382, + "flow_matching_loss": 0.03062690682709217, + "grad_norm": 0.3310674726963043, + "learning_rate": 2.994811676932232e-05, + "loss": 0.0679, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015389664005488158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007035750895738602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03781507909297943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.986140388250351, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9020, + "weighted_flow_matching_loss": 0.03062690682709217, + "weighted_rkd_loss": 0.03729098588228226 + }, + { + "epoch": 2.627291242362525, + "flow_matching_loss": 0.031882813200354576, + "grad_norm": 0.3996850550174713, + "learning_rate": 2.994777268863544e-05, + "loss": 0.0723, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016578080039471386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007815036829560995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040971196815371515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9861097633838654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9030, + "weighted_flow_matching_loss": 0.031882813200354576, + "weighted_rkd_loss": 0.04040207825601101 + }, + { + "epoch": 2.630200756473669, + "flow_matching_loss": 0.03614100199192762, + "grad_norm": 0.32148391008377075, + "learning_rate": 2.994742747276387e-05, + "loss": 0.0774, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017049990221858026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007712863432243466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04181284420192242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9860791027545929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9040, + "weighted_flow_matching_loss": 0.03614100199192762, + "weighted_rkd_loss": 0.04123078733682632 + }, + { + "epoch": 2.633110270584812, + "flow_matching_loss": 0.034007074870169164, + "grad_norm": 0.35842275619506836, + "learning_rate": 2.9947081121733835e-05, + "loss": 0.0761, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017240553349256515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008257898595184087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042739004269242285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9860484182834626, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9050, + "weighted_flow_matching_loss": 0.034007074870169164, + "weighted_rkd_loss": 0.042142726853489874 + }, + { + "epoch": 2.636019784695956, + "flow_matching_loss": 0.03355587068945169, + "grad_norm": 0.35519489645957947, + "learning_rate": 2.9946733635571625e-05, + "loss": 0.0773, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01806049644947052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008207767875865102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04432876035571098, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9860176920890809, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9060, + "weighted_flow_matching_loss": 0.03355587068945169, + "weighted_rkd_loss": 0.043708961084485055 + }, + { + "epoch": 2.638929298807099, + "flow_matching_loss": 0.03219871949404478, + "grad_norm": 0.31526580452919006, + "learning_rate": 2.994638501430364e-05, + "loss": 0.0752, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01773009579628706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008201261656358838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0436614528298378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9859869301319122, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9070, + "weighted_flow_matching_loss": 0.03219871949404478, + "weighted_rkd_loss": 0.043049604073166846 + }, + { + "epoch": 2.641838812918243, + "flow_matching_loss": 0.03222274202853441, + "grad_norm": 0.3946211636066437, + "learning_rate": 2.9946035257956345e-05, + "loss": 0.0733, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016784299723804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008108861651271582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041677461192011836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9859561443328857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9080, + "weighted_flow_matching_loss": 0.03222274202853441, + "weighted_rkd_loss": 0.041092167049646376 + }, + { + "epoch": 2.644748327029386, + "flow_matching_loss": 0.03253653980791569, + "grad_norm": 0.3740198016166687, + "learning_rate": 2.9945684366556313e-05, + "loss": 0.075, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017360109742730855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008372050477191806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04309227019548416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9859253168106079, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9090, + "weighted_flow_matching_loss": 0.03253653980791569, + "weighted_rkd_loss": 0.04248575419187546 + }, + { + "epoch": 2.6476578411405294, + "flow_matching_loss": 0.03225305937230587, + "grad_norm": 0.3571323752403259, + "learning_rate": 2.9945332340130183e-05, + "loss": 0.072, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016199218947440384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007903466233983636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04030190408229828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9858944654464722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9100, + "weighted_flow_matching_loss": 0.03225305937230587, + "weighted_rkd_loss": 0.039733439311385156 + }, + { + "epoch": 2.650567355251673, + "flow_matching_loss": 0.0332100436091423, + "grad_norm": 0.4065572917461395, + "learning_rate": 2.994497917870469e-05, + "loss": 0.0745, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01700648237019777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007917254790663719, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041930219158530235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9858635783195495, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9110, + "weighted_flow_matching_loss": 0.0332100436091423, + "weighted_rkd_loss": 0.04133744686841965 + }, + { + "epoch": 2.6534768693628163, + "flow_matching_loss": 0.03393442500382662, + "grad_norm": 0.36083051562309265, + "learning_rate": 2.994462488230666e-05, + "loss": 0.0735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016160492226481436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007804545946419239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04012553039938212, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9858326494693757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9120, + "weighted_flow_matching_loss": 0.03393442500382662, + "weighted_rkd_loss": 0.03955704811960459 + }, + { + "epoch": 2.6563863834739596, + "flow_matching_loss": 0.03266864717006683, + "grad_norm": 0.36165064573287964, + "learning_rate": 2.9944269450962997e-05, + "loss": 0.0727, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0163576596416533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007869899459183215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040585218369960784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9858016967773438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9130, + "weighted_flow_matching_loss": 0.03266864717006683, + "weighted_rkd_loss": 0.04000896997749805 + }, + { + "epoch": 2.6592958975851033, + "flow_matching_loss": 0.03315623104572296, + "grad_norm": 0.3696068823337555, + "learning_rate": 2.994391288470069e-05, + "loss": 0.0724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016154191736131907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007465002732351423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03977338597178459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9857707023620605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9140, + "weighted_flow_matching_loss": 0.03315623104572296, + "weighted_rkd_loss": 0.03920743465423584 + }, + { + "epoch": 2.6622054116962466, + "flow_matching_loss": 0.033333659544587134, + "grad_norm": 0.30891865491867065, + "learning_rate": 2.994355518354682e-05, + "loss": 0.0743, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016698401421308517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008145892666652799, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041542695462703706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9857396841049194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9150, + "weighted_flow_matching_loss": 0.033333659544587134, + "weighted_rkd_loss": 0.0409502886235714 + }, + { + "epoch": 2.6651149258073903, + "flow_matching_loss": 0.034586962312459946, + "grad_norm": 0.3782028257846832, + "learning_rate": 2.994319634752856e-05, + "loss": 0.0766, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017171496339142323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008266795240342618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04260978773236275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.985708624124527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9160, + "weighted_flow_matching_loss": 0.034586962312459946, + "weighted_rkd_loss": 0.04200081825256348 + }, + { + "epoch": 2.6680244399185336, + "flow_matching_loss": 0.035193653963506225, + "grad_norm": 0.37349647283554077, + "learning_rate": 2.9942836376673145e-05, + "loss": 0.0794, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01793433362618089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009025025554001332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04489369280636311, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9856775343418122, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9170, + "weighted_flow_matching_loss": 0.035193653963506225, + "weighted_rkd_loss": 0.04425070285797119 + }, + { + "epoch": 2.6709339540296773, + "flow_matching_loss": 0.0339345833286643, + "grad_norm": 0.3676605820655823, + "learning_rate": 2.994247527100793e-05, + "loss": 0.0767, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017382218223065137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008653162745758891, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043417598307132724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.985646414756775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9180, + "weighted_flow_matching_loss": 0.0339345833286643, + "weighted_rkd_loss": 0.042794400453567506 + }, + { + "epoch": 2.6738434681408205, + "flow_matching_loss": 0.03255172204226255, + "grad_norm": 0.3981539309024811, + "learning_rate": 2.9942113030560327e-05, + "loss": 0.0755, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01757704494521022, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008443629741668702, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04359771907329559, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9856152594089508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9190, + "weighted_flow_matching_loss": 0.03255172204226255, + "weighted_rkd_loss": 0.04297056309878826 + }, + { + "epoch": 2.676752982251964, + "flow_matching_loss": 0.033663198351860046, + "grad_norm": 0.37696510553359985, + "learning_rate": 2.9941749655357844e-05, + "loss": 0.0722, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015722927730530502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0076254080981016156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03907126411795616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9855840682983399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9200, + "weighted_flow_matching_loss": 0.033663198351860046, + "weighted_rkd_loss": 0.038508028537034986 + }, + { + "epoch": 2.6796624963631075, + "flow_matching_loss": 0.03112845718860626, + "grad_norm": 0.3404398560523987, + "learning_rate": 2.9941385145428083e-05, + "loss": 0.074, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017673385702073573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008166508004069328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04351327903568745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9855528473854065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9210, + "weighted_flow_matching_loss": 0.03112845718860626, + "weighted_rkd_loss": 0.04288463667035103 + }, + { + "epoch": 2.6825720104742508, + "flow_matching_loss": 0.03375545646995306, + "grad_norm": 0.4140242338180542, + "learning_rate": 2.994101950079873e-05, + "loss": 0.0777, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017964500468224287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008637960813939572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044566961377859114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9855215847492218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9220, + "weighted_flow_matching_loss": 0.03375545646995306, + "weighted_rkd_loss": 0.04392170123755932 + }, + { + "epoch": 2.685481524585394, + "flow_matching_loss": 0.03329048678278923, + "grad_norm": 0.40866023302078247, + "learning_rate": 2.9940652721497543e-05, + "loss": 0.0729, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01621517911553383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007732199691236019, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04016255848109722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9854902923107147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9230, + "weighted_flow_matching_loss": 0.03329048678278923, + "weighted_rkd_loss": 0.03957982901483774 + }, + { + "epoch": 2.6883910386965377, + "flow_matching_loss": 0.034661256708204746, + "grad_norm": 0.3775288164615631, + "learning_rate": 2.9940284807552384e-05, + "loss": 0.0787, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018111691810190678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008510576002299786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04473395943641663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9854589760303497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9240, + "weighted_flow_matching_loss": 0.034661256708204746, + "weighted_rkd_loss": 0.0440835066139698 + }, + { + "epoch": 2.691300552807681, + "flow_matching_loss": 0.03444662410765886, + "grad_norm": 0.3616732954978943, + "learning_rate": 2.993991575899119e-05, + "loss": 0.0766, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01717905392870307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00846426202915609, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04282236993312836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9854276180267334, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9250, + "weighted_flow_matching_loss": 0.03444662410765886, + "weighted_rkd_loss": 0.04219831973314285 + }, + { + "epoch": 2.6942100669188247, + "flow_matching_loss": 0.032260856591165066, + "grad_norm": 0.3813704252243042, + "learning_rate": 2.9939545575841986e-05, + "loss": 0.0732, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01675600167363882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008029969967901706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04154197350144386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9853962242603302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9260, + "weighted_flow_matching_loss": 0.032260856591165066, + "weighted_rkd_loss": 0.04093530066311359 + }, + { + "epoch": 2.697119581029968, + "flow_matching_loss": 0.03310307487845421, + "grad_norm": 0.4243256449699402, + "learning_rate": 2.993917425813289e-05, + "loss": 0.0758, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017485542595386504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008403942128643394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043375026807188986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9853648126125336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9270, + "weighted_flow_matching_loss": 0.03310307487845421, + "weighted_rkd_loss": 0.042740239202976225 + }, + { + "epoch": 2.7000290951411117, + "flow_matching_loss": 0.03173750787973404, + "grad_norm": 0.3283553421497345, + "learning_rate": 2.99388018058921e-05, + "loss": 0.0764, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018307171389460565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008684996888041497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04529934003949165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9853333592414856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9280, + "weighted_flow_matching_loss": 0.03173750787973404, + "weighted_rkd_loss": 0.04463495761156082 + }, + { + "epoch": 2.702938609252255, + "flow_matching_loss": 0.031396870873868464, + "grad_norm": 0.31427860260009766, + "learning_rate": 2.9938428219147902e-05, + "loss": 0.0711, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01636983556672931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007517072511836887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040256743878126146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9853018641471862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9290, + "weighted_flow_matching_loss": 0.031396870873868464, + "weighted_rkd_loss": 0.039665070548653605 + }, + { + "epoch": 2.705848123363398, + "flow_matching_loss": 0.03131113406270743, + "grad_norm": 0.39976197481155396, + "learning_rate": 2.9938053497928666e-05, + "loss": 0.072, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016489831637591122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008338838210329413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04131850153207779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9852703452110291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9300, + "weighted_flow_matching_loss": 0.03131113406270743, + "weighted_rkd_loss": 0.040709884651005265 + }, + { + "epoch": 2.708757637474542, + "flow_matching_loss": 0.03398359213024378, + "grad_norm": 0.37832027673721313, + "learning_rate": 2.9937677642262844e-05, + "loss": 0.076, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01723440093919635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008194815600290895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04266361780464649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.985238778591156, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9310, + "weighted_flow_matching_loss": 0.03398359213024378, + "weighted_rkd_loss": 0.042033829540014264 + }, + { + "epoch": 2.711667151585685, + "flow_matching_loss": 0.03166589289903641, + "grad_norm": 0.348960280418396, + "learning_rate": 2.993730065217899e-05, + "loss": 0.0741, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017380377650260924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008317678235471249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043078433349728586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9852071940898895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9320, + "weighted_flow_matching_loss": 0.03166589289903641, + "weighted_rkd_loss": 0.04244118183851242 + }, + { + "epoch": 2.7145766656968284, + "flow_matching_loss": 0.031068926118314265, + "grad_norm": 0.37718600034713745, + "learning_rate": 2.993692252770573e-05, + "loss": 0.0704, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0161864323541522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007512303907424212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03988516833633184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9851755797863007, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9330, + "weighted_flow_matching_loss": 0.031068926118314265, + "weighted_rkd_loss": 0.03929391019046306 + }, + { + "epoch": 2.717486179807972, + "flow_matching_loss": 0.03476364314556122, + "grad_norm": 0.400436669588089, + "learning_rate": 2.9936543268871776e-05, + "loss": 0.0734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015790146216750144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007629223726689815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03920951634645462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.985143917798996, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9340, + "weighted_flow_matching_loss": 0.03476364314556122, + "weighted_rkd_loss": 0.038627023622393605 + }, + { + "epoch": 2.7203956939191154, + "flow_matching_loss": 0.032447518408298494, + "grad_norm": 0.40866604447364807, + "learning_rate": 2.9936162875705933e-05, + "loss": 0.075, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01730534080415964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008535022428259254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04314570426940918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9851122379302979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9350, + "weighted_flow_matching_loss": 0.032447518408298494, + "weighted_rkd_loss": 0.04250333942472935 + }, + { + "epoch": 2.723305208030259, + "flow_matching_loss": 0.03231866341084242, + "grad_norm": 0.32472050189971924, + "learning_rate": 2.993578134823709e-05, + "loss": 0.0777, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01838787840679288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00925437775440514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04603013470768928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9850805103778839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9360, + "weighted_flow_matching_loss": 0.03231866341084242, + "weighted_rkd_loss": 0.04534340761601925 + }, + { + "epoch": 2.7262147221414024, + "flow_matching_loss": 0.03258288353681564, + "grad_norm": 0.322629451751709, + "learning_rate": 2.9935398686494226e-05, + "loss": 0.0726, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016551873460412026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007477038726210594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040580785274505614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9850487530231475, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9370, + "weighted_flow_matching_loss": 0.03258288353681564, + "weighted_rkd_loss": 0.03997407704591751 + }, + { + "epoch": 2.729124236252546, + "flow_matching_loss": 0.03178622648119926, + "grad_norm": 0.3475531339645386, + "learning_rate": 2.993501489050639e-05, + "loss": 0.0715, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01647890852764249, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007405781885609031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04036359861493111, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9850169658660889, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9380, + "weighted_flow_matching_loss": 0.03178622648119926, + "weighted_rkd_loss": 0.03975883238017559 + }, + { + "epoch": 2.7320337503636893, + "flow_matching_loss": 0.03272149953991175, + "grad_norm": 0.3601875901222229, + "learning_rate": 2.9934629960302742e-05, + "loss": 0.0735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016714587341994047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007951465342193842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041380640119314194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9849851429462433, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9390, + "weighted_flow_matching_loss": 0.03272149953991175, + "weighted_rkd_loss": 0.0407593097537756 + }, + { + "epoch": 2.7349432644748326, + "flow_matching_loss": 0.034858956933021545, + "grad_norm": 0.5074306130409241, + "learning_rate": 2.99342438959125e-05, + "loss": 0.0755, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01671040914952755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007873423164710402, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04129424095153809, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9849532842636108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9400, + "weighted_flow_matching_loss": 0.034858956933021545, + "weighted_rkd_loss": 0.04067287966609001 + }, + { + "epoch": 2.7378527785859763, + "flow_matching_loss": 0.029763745330274104, + "grad_norm": 0.32271808385849, + "learning_rate": 2.9933856697364998e-05, + "loss": 0.0669, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015295048803091049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007087560184299946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0376776572316885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.984921395778656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9410, + "weighted_flow_matching_loss": 0.029763745330274104, + "weighted_rkd_loss": 0.03710955493152142 + }, + { + "epoch": 2.7407622926971196, + "flow_matching_loss": 0.030448013544082643, + "grad_norm": 0.36356550455093384, + "learning_rate": 2.9933468364689632e-05, + "loss": 0.0708, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016657991521060466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007606720877811312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04092270396649837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9848894834518432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9420, + "weighted_flow_matching_loss": 0.030448013544082643, + "weighted_rkd_loss": 0.04030437245965004 + }, + { + "epoch": 2.743671806808263, + "flow_matching_loss": 0.0321541441604495, + "grad_norm": 0.3707292377948761, + "learning_rate": 2.9933078897915903e-05, + "loss": 0.0731, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016662775166332723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008230211958289147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04155576191842556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9848575294017792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9430, + "weighted_flow_matching_loss": 0.0321541441604495, + "weighted_rkd_loss": 0.04092649221420288 + }, + { + "epoch": 2.7465813209194065, + "flow_matching_loss": 0.029191220365464687, + "grad_norm": 0.3490999639034271, + "learning_rate": 2.9932688297073373e-05, + "loss": 0.0679, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016033448465168475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007194084953516722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039260981418192384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9848255395889283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9440, + "weighted_flow_matching_loss": 0.029191220365464687, + "weighted_rkd_loss": 0.03866522666066885 + }, + { + "epoch": 2.74949083503055, + "flow_matching_loss": 0.03186376430094242, + "grad_norm": 0.353290319442749, + "learning_rate": 2.993229656219172e-05, + "loss": 0.0695, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015429917816072702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007308833627030253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03816866874694824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9847935140132904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9450, + "weighted_flow_matching_loss": 0.03186376430094242, + "weighted_rkd_loss": 0.037588228657841685 + }, + { + "epoch": 2.752400349141693, + "flow_matching_loss": 0.032522836700081825, + "grad_norm": 0.38686302304267883, + "learning_rate": 2.993190369330068e-05, + "loss": 0.0743, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017151571717113255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008079646108672022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04238278903067112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9847614645957947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9460, + "weighted_flow_matching_loss": 0.032522836700081825, + "weighted_rkd_loss": 0.04173693396151066 + }, + { + "epoch": 2.7553098632528368, + "flow_matching_loss": 0.03390190247446299, + "grad_norm": 0.3367585837841034, + "learning_rate": 2.9931509690430102e-05, + "loss": 0.0793, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018564689066261054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008968143537640572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04609752111136913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9847293794155121, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9470, + "weighted_flow_matching_loss": 0.03390190247446299, + "weighted_rkd_loss": 0.04539359472692013 + }, + { + "epoch": 2.7582193773639805, + "flow_matching_loss": 0.033120812475681306, + "grad_norm": 0.320253849029541, + "learning_rate": 2.99311145536099e-05, + "loss": 0.0734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01647438183426857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007918986724689603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04086774997413158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9846972584724426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9480, + "weighted_flow_matching_loss": 0.033120812475681306, + "weighted_rkd_loss": 0.04024237245321274 + }, + { + "epoch": 2.7611288914751237, + "flow_matching_loss": 0.03475774247199297, + "grad_norm": 0.35280904173851013, + "learning_rate": 2.993071828287009e-05, + "loss": 0.074, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016052634920924902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007715147547423839, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039820417389273645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9846650958061218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9490, + "weighted_flow_matching_loss": 0.03475774247199297, + "weighted_rkd_loss": 0.03920978717505932 + }, + { + "epoch": 2.764038405586267, + "flow_matching_loss": 0.035489279218018054, + "grad_norm": 0.37050989270210266, + "learning_rate": 2.9930320878240753e-05, + "loss": 0.0755, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016333271004259586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00793660837225616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04060315042734146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9846329092979431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9500, + "weighted_flow_matching_loss": 0.035489279218018054, + "weighted_rkd_loss": 0.03997919335961342 + }, + { + "epoch": 2.7669479196974107, + "flow_matching_loss": 0.03191109523177147, + "grad_norm": 0.31680673360824585, + "learning_rate": 2.9929922339752084e-05, + "loss": 0.0725, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016696301847696306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007865663338452577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041258266940712926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9846006989479065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9510, + "weighted_flow_matching_loss": 0.03191109523177147, + "weighted_rkd_loss": 0.04062293730676174 + }, + { + "epoch": 2.769857433808554, + "flow_matching_loss": 0.03634234182536602, + "grad_norm": 0.3788840174674988, + "learning_rate": 2.9929522667434334e-05, + "loss": 0.0753, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016033822298049928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00754232220351696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039609966799616814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9845684468746185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9520, + "weighted_flow_matching_loss": 0.03634234182536602, + "weighted_rkd_loss": 0.03899872303009033 + }, + { + "epoch": 2.7727669479196972, + "flow_matching_loss": 0.03347853347659111, + "grad_norm": 0.37849316000938416, + "learning_rate": 2.9929121861317872e-05, + "loss": 0.0746, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016751994378864765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008294438617303968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04179842695593834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9845361530780792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9530, + "weighted_flow_matching_loss": 0.03347853347659111, + "weighted_rkd_loss": 0.04115205518901348 + }, + { + "epoch": 2.775676462030841, + "flow_matching_loss": 0.03534841146320104, + "grad_norm": 0.428821325302124, + "learning_rate": 2.9928719921433122e-05, + "loss": 0.0751, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016282312106341124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007819684781134128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04038430862128735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.984503835439682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9540, + "weighted_flow_matching_loss": 0.03534841146320104, + "weighted_rkd_loss": 0.03975850120186806 + }, + { + "epoch": 2.778585976141984, + "flow_matching_loss": 0.030847177654504777, + "grad_norm": 0.33441880345344543, + "learning_rate": 2.9928316847810618e-05, + "loss": 0.0711, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016620219685137272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007630565669387579, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04087100476026535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9844714820384979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9550, + "weighted_flow_matching_loss": 0.030847177654504777, + "weighted_rkd_loss": 0.04023633599281311 + }, + { + "epoch": 2.7814954902531275, + "flow_matching_loss": 0.031632243655622005, + "grad_norm": 0.37050652503967285, + "learning_rate": 2.9927912640480966e-05, + "loss": 0.0708, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015909710619598626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007923185545951127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03974260687828064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9844390988349915, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9560, + "weighted_flow_matching_loss": 0.031632243655622005, + "weighted_rkd_loss": 0.039124172553420065 + }, + { + "epoch": 2.784405004364271, + "flow_matching_loss": 0.034043876267969606, + "grad_norm": 0.37115731835365295, + "learning_rate": 2.9927507299474865e-05, + "loss": 0.0724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01569540109485388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007542272610589862, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038933074474334715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9844066739082337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9570, + "weighted_flow_matching_loss": 0.034043876267969606, + "weighted_rkd_loss": 0.03832596633583307 + }, + { + "epoch": 2.7873145184754144, + "flow_matching_loss": 0.033995925635099414, + "grad_norm": 0.37043341994285583, + "learning_rate": 2.99271008248231e-05, + "loss": 0.0759, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017288217786699533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008028578665107488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04260501451790333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9843742251396179, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9580, + "weighted_flow_matching_loss": 0.033995925635099414, + "weighted_rkd_loss": 0.04193927347660065 + }, + { + "epoch": 2.790224032586558, + "flow_matching_loss": 0.03274632822722197, + "grad_norm": 0.3300260305404663, + "learning_rate": 2.9926693216556536e-05, + "loss": 0.0727, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01648354884237051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007577646104618907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04054474346339702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9843417406082153, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9590, + "weighted_flow_matching_loss": 0.03274632822722197, + "weighted_rkd_loss": 0.039909875765442845 + }, + { + "epoch": 2.7931335466977014, + "flow_matching_loss": 0.03262751139700413, + "grad_norm": 0.3628789186477661, + "learning_rate": 2.992628447470613e-05, + "loss": 0.0719, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016173854004591703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007565731601789594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03991343937814236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9843092203140259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9600, + "weighted_flow_matching_loss": 0.03262751139700413, + "weighted_rkd_loss": 0.03928717672824859 + }, + { + "epoch": 2.796043060808845, + "flow_matching_loss": 0.034192664176225664, + "grad_norm": 0.4006335735321045, + "learning_rate": 2.9925874599302922e-05, + "loss": 0.0753, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016838318761438133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00808781930245459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04176445715129375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9842766642570495, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9610, + "weighted_flow_matching_loss": 0.034192664176225664, + "weighted_rkd_loss": 0.04110778346657753 + }, + { + "epoch": 2.7989525749199884, + "flow_matching_loss": 0.031932453624904156, + "grad_norm": 0.3338753581047058, + "learning_rate": 2.9925463590378043e-05, + "loss": 0.0688, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015059919841587544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007301147561520338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037420987710356714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9842440843582153, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9620, + "weighted_flow_matching_loss": 0.031932453624904156, + "weighted_rkd_loss": 0.036831393279135226 + }, + { + "epoch": 2.8018620890311317, + "flow_matching_loss": 0.03255791943520307, + "grad_norm": 0.40395668148994446, + "learning_rate": 2.99250514479627e-05, + "loss": 0.0736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016911129746586085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007924163760617376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04174642339348793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9842114627361298, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9630, + "weighted_flow_matching_loss": 0.03255791943520307, + "weighted_rkd_loss": 0.04108729138970375 + }, + { + "epoch": 2.8047716031422754, + "flow_matching_loss": 0.03314264845103025, + "grad_norm": 0.3472961187362671, + "learning_rate": 2.9924638172088196e-05, + "loss": 0.0734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016426072921603918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008015703130513429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040867849439382556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9841788113117218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9640, + "weighted_flow_matching_loss": 0.03314264845103025, + "weighted_rkd_loss": 0.040221283584833144 + }, + { + "epoch": 2.8076811172534186, + "flow_matching_loss": 0.03252452574670315, + "grad_norm": 0.4119461178779602, + "learning_rate": 2.9924223762785922e-05, + "loss": 0.0722, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016256777010858058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007850980991497636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04036453515291214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.984146124124527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9650, + "weighted_flow_matching_loss": 0.03252452574670315, + "weighted_rkd_loss": 0.039724599570035934 + }, + { + "epoch": 2.810590631364562, + "flow_matching_loss": 0.032090894505381586, + "grad_norm": 0.42918094992637634, + "learning_rate": 2.992380822008734e-05, + "loss": 0.0729, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016691435780376195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008064496843144298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0414473682641983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9841134071350097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9660, + "weighted_flow_matching_loss": 0.032090894505381586, + "weighted_rkd_loss": 0.04078889098018408 + }, + { + "epoch": 2.8135001454757056, + "flow_matching_loss": 0.033771219477057454, + "grad_norm": 0.3638564348220825, + "learning_rate": 2.9923391544024015e-05, + "loss": 0.0724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01586361937224865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007492495281621814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039219734445214274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9840806424617767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9670, + "weighted_flow_matching_loss": 0.033771219477057454, + "weighted_rkd_loss": 0.038595366105437276 + }, + { + "epoch": 2.816409659586849, + "flow_matching_loss": 0.032958795130252835, + "grad_norm": 0.3299553096294403, + "learning_rate": 2.9922973734627586e-05, + "loss": 0.0722, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016171194706112148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007534890901297331, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03987728022038937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9840478599071503, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9680, + "weighted_flow_matching_loss": 0.032958795130252835, + "weighted_rkd_loss": 0.03924115523695946 + }, + { + "epoch": 2.8193191736979926, + "flow_matching_loss": 0.03312576059252024, + "grad_norm": 0.41634586453437805, + "learning_rate": 2.9922554791929786e-05, + "loss": 0.0707, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015423815883696079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00733520402573049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038182835839688775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9840150535106659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9690, + "weighted_flow_matching_loss": 0.03312576059252024, + "weighted_rkd_loss": 0.03757249340415001 + }, + { + "epoch": 2.822228687809136, + "flow_matching_loss": 0.031542889215052126, + "grad_norm": 0.33394762873649597, + "learning_rate": 2.992213471596243e-05, + "loss": 0.0693, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01575246937572956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006894022412598133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03839896135032177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9839822053909302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9700, + "weighted_flow_matching_loss": 0.031542889215052126, + "weighted_rkd_loss": 0.03778393715620041 + }, + { + "epoch": 2.8251382019202795, + "flow_matching_loss": 0.03135459125041962, + "grad_norm": 0.33073967695236206, + "learning_rate": 2.992171350675742e-05, + "loss": 0.0696, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015755964629352092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007313283206894994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03882521260529757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9839493155479431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9710, + "weighted_flow_matching_loss": 0.03135459125041962, + "weighted_rkd_loss": 0.03820204511284828 + }, + { + "epoch": 2.828047716031423, + "flow_matching_loss": 0.03326356709003449, + "grad_norm": 0.3194456696510315, + "learning_rate": 2.9921291164346743e-05, + "loss": 0.0713, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015628766547888518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007361145643517375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03861867859959602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9839164018630981, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9720, + "weighted_flow_matching_loss": 0.03326356709003449, + "weighted_rkd_loss": 0.03799757193773985 + }, + { + "epoch": 2.830957230142566, + "flow_matching_loss": 0.03175548892468214, + "grad_norm": 0.3474823832511902, + "learning_rate": 2.9920867688762476e-05, + "loss": 0.0687, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015254756156355142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00700174174271524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037511253729462624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9838834524154663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9730, + "weighted_flow_matching_loss": 0.03175548892468214, + "weighted_rkd_loss": 0.03690672945231199 + }, + { + "epoch": 2.8338667442537098, + "flow_matching_loss": 0.033047156780958174, + "grad_norm": 0.43931087851524353, + "learning_rate": 2.9920443080036777e-05, + "loss": 0.0721, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015912967454642057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007836744887754321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03966267928481102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9838504672050477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9740, + "weighted_flow_matching_loss": 0.033047156780958174, + "weighted_rkd_loss": 0.03902214914560318 + }, + { + "epoch": 2.836776258364853, + "flow_matching_loss": 0.03346973843872547, + "grad_norm": 0.385677695274353, + "learning_rate": 2.992001733820189e-05, + "loss": 0.0704, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01510533932596445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007313322555273771, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03752400148659944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9838174462318421, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9750, + "weighted_flow_matching_loss": 0.03346973843872547, + "weighted_rkd_loss": 0.03691677432507277 + }, + { + "epoch": 2.8396857724759963, + "flow_matching_loss": 0.030062059685587884, + "grad_norm": 0.34968283772468567, + "learning_rate": 2.9919590463290152e-05, + "loss": 0.0669, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015189805068075657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007045012339949608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03742462247610092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.983784407377243, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9760, + "weighted_flow_matching_loss": 0.030062059685587884, + "weighted_rkd_loss": 0.03681777529418469 + }, + { + "epoch": 2.84259528658714, + "flow_matching_loss": 0.03177945427596569, + "grad_norm": 0.37083834409713745, + "learning_rate": 2.9919162455333975e-05, + "loss": 0.0712, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016230959072709085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007655415311455726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04011733345687389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9837513208389282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9770, + "weighted_flow_matching_loss": 0.03177945427596569, + "weighted_rkd_loss": 0.03946550153195858 + }, + { + "epoch": 2.8455048006982833, + "flow_matching_loss": 0.033196520805358884, + "grad_norm": 0.4317930042743683, + "learning_rate": 2.991873331436587e-05, + "loss": 0.0747, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017055809777230025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00811913339421153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04223075322806835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9837182044982911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9780, + "weighted_flow_matching_loss": 0.033196520805358884, + "weighted_rkd_loss": 0.04154314827173948 + }, + { + "epoch": 2.848414314809427, + "flow_matching_loss": 0.0320734366774559, + "grad_norm": 0.32648271322250366, + "learning_rate": 2.9918303040418426e-05, + "loss": 0.0704, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015720235370099544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007547970674932003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03898844122886658, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9836850464344025, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9790, + "weighted_flow_matching_loss": 0.0320734366774559, + "weighted_rkd_loss": 0.038352370634675025 + }, + { + "epoch": 2.8513238289205702, + "flow_matching_loss": 0.035033502615988255, + "grad_norm": 0.38783562183380127, + "learning_rate": 2.9917871633524318e-05, + "loss": 0.0758, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016850259993225335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007741156546398997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04144167602062225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.983651864528656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9800, + "weighted_flow_matching_loss": 0.035033502615988255, + "weighted_rkd_loss": 0.04076417721807957 + }, + { + "epoch": 2.854233343031714, + "flow_matching_loss": 0.030180023051798344, + "grad_norm": 0.3611852526664734, + "learning_rate": 2.991743909371631e-05, + "loss": 0.0678, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015385758318006992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007508266577497124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038279783353209494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9836186468601227, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9810, + "weighted_flow_matching_loss": 0.030180023051798344, + "weighted_rkd_loss": 0.037652704119682315 + }, + { + "epoch": 2.857142857142857, + "flow_matching_loss": 0.03149999901652336, + "grad_norm": 0.4599561393260956, + "learning_rate": 2.9917005421027242e-05, + "loss": 0.072, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016439219191670417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008248054888099432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041126493737101553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9835854053497315, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9820, + "weighted_flow_matching_loss": 0.03149999901652336, + "weighted_rkd_loss": 0.040451401844620705 + }, + { + "epoch": 2.8600523712540005, + "flow_matching_loss": 0.032608835212886336, + "grad_norm": 0.3389822542667389, + "learning_rate": 2.9916570615490066e-05, + "loss": 0.0699, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015260648541152477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007369847828522324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037891144305467604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9835521221160889, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9830, + "weighted_flow_matching_loss": 0.032608835212886336, + "weighted_rkd_loss": 0.037267911806702615 + }, + { + "epoch": 2.862961885365144, + "flow_matching_loss": 0.032133198156952855, + "grad_norm": 0.327466756105423, + "learning_rate": 2.9916134677137785e-05, + "loss": 0.0709, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01606281390413642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00731092607602477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039436554349958894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9835188090801239, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9840, + "weighted_flow_matching_loss": 0.032133198156952855, + "weighted_rkd_loss": 0.03878660127520561 + }, + { + "epoch": 2.8658713994762874, + "flow_matching_loss": 0.03245924673974514, + "grad_norm": 0.39171215891838074, + "learning_rate": 2.991569760600352e-05, + "loss": 0.0718, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01601665522903204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007984007010236382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04001731760799885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9834854602813721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9850, + "weighted_flow_matching_loss": 0.03245924673974514, + "weighted_rkd_loss": 0.039356441050767896 + }, + { + "epoch": 2.8687809135874307, + "flow_matching_loss": 0.03233292791992426, + "grad_norm": 0.31100985407829285, + "learning_rate": 2.9915259402120454e-05, + "loss": 0.0702, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01551229963079095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0074503427837044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03847494199872017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9834520816802979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9860, + "weighted_flow_matching_loss": 0.03233292791992426, + "weighted_rkd_loss": 0.037838276103138925 + }, + { + "epoch": 2.8716904276985744, + "flow_matching_loss": 0.03160697873681784, + "grad_norm": 0.37392279505729675, + "learning_rate": 2.9914820065521866e-05, + "loss": 0.0707, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016043714340776206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00769015122205019, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03977758027613163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9834186613559723, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9870, + "weighted_flow_matching_loss": 0.03160697873681784, + "weighted_rkd_loss": 0.0391180157661438 + }, + { + "epoch": 2.8745999418097177, + "flow_matching_loss": 0.031828411668539044, + "grad_norm": 0.4290027320384979, + "learning_rate": 2.9914379596241124e-05, + "loss": 0.0738, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01749071441590786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007720392756164074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04270182177424431, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9833852171897888, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9880, + "weighted_flow_matching_loss": 0.031828411668539044, + "weighted_rkd_loss": 0.04199232254177332 + }, + { + "epoch": 2.8775094559208614, + "flow_matching_loss": 0.031906531378626826, + "grad_norm": 0.45078974962234497, + "learning_rate": 2.991393799431168e-05, + "loss": 0.074, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017144015803933144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008506317622959614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04279434941709041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9833517372608185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9890, + "weighted_flow_matching_loss": 0.031906531378626826, + "weighted_rkd_loss": 0.042081900872290134 + }, + { + "epoch": 2.8804189700320046, + "flow_matching_loss": 0.0323579341173172, + "grad_norm": 0.3046727478504181, + "learning_rate": 2.991349525976707e-05, + "loss": 0.0694, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015141044650226832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0074094146024435755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03769150413572788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9833182215690612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9900, + "weighted_flow_matching_loss": 0.0323579341173172, + "weighted_rkd_loss": 0.03706272710114718 + }, + { + "epoch": 2.8833284841431484, + "flow_matching_loss": 0.031273848563432696, + "grad_norm": 0.40819016098976135, + "learning_rate": 2.991305139264091e-05, + "loss": 0.0684, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015210665389895439, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0073007556144148115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037722086533904076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9832846760749817, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9910, + "weighted_flow_matching_loss": 0.031273848563432696, + "weighted_rkd_loss": 0.03709154985845089 + }, + { + "epoch": 2.8862379982542916, + "flow_matching_loss": 0.03307069577276707, + "grad_norm": 0.3417457938194275, + "learning_rate": 2.9912606392966918e-05, + "loss": 0.0703, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015309077873826026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007239536801353097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03785769231617451, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9832510948181152, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9920, + "weighted_flow_matching_loss": 0.03307069577276707, + "weighted_rkd_loss": 0.03722363412380218 + }, + { + "epoch": 2.889147512365435, + "flow_matching_loss": 0.0342417323961854, + "grad_norm": 0.26762691140174866, + "learning_rate": 2.9912160260778885e-05, + "loss": 0.0721, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015517135802656411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007453261502087116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03848753310739994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9832174777984619, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9930, + "weighted_flow_matching_loss": 0.0342417323961854, + "weighted_rkd_loss": 0.03784165661782026 + }, + { + "epoch": 2.8920570264765786, + "flow_matching_loss": 0.03164354115724564, + "grad_norm": 0.359514981508255, + "learning_rate": 2.9911712996110688e-05, + "loss": 0.0704, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01583087518811226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007762524951249361, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039424275048077106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9831838309764862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9940, + "weighted_flow_matching_loss": 0.03164354115724564, + "weighted_rkd_loss": 0.038761311769485475 + }, + { + "epoch": 2.894966540587722, + "flow_matching_loss": 0.03353297226130962, + "grad_norm": 0.38460659980773926, + "learning_rate": 2.9911264598996304e-05, + "loss": 0.0706, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01511990986764431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007441256288439035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037681075558066365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9831501543521881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9950, + "weighted_flow_matching_loss": 0.03353297226130962, + "weighted_rkd_loss": 0.037046154774725434 + }, + { + "epoch": 2.897876054698865, + "flow_matching_loss": 0.03182348627597094, + "grad_norm": 0.34038281440734863, + "learning_rate": 2.991081506946978e-05, + "loss": 0.071, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016137734334915877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007588516315445304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039863985031843185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9831164360046387, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9960, + "weighted_flow_matching_loss": 0.03182348627597094, + "weighted_rkd_loss": 0.03919092882424593 + }, + { + "epoch": 2.900785568810009, + "flow_matching_loss": 0.03181076515465975, + "grad_norm": 0.34281909465789795, + "learning_rate": 2.9910364407565246e-05, + "loss": 0.0673, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014491645153611898, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007155481213703752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03613877091556787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9830826938152313, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9970, + "weighted_flow_matching_loss": 0.03181076515465975, + "weighted_rkd_loss": 0.03552740104496479 + }, + { + "epoch": 2.903695082921152, + "flow_matching_loss": 0.03137392792850733, + "grad_norm": 0.3736158311367035, + "learning_rate": 2.9909912613316942e-05, + "loss": 0.0678, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014873151481151582, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007274535112082958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03702083770185709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9830489158630371, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9980, + "weighted_flow_matching_loss": 0.03137392792850733, + "weighted_rkd_loss": 0.0363932890817523 + }, + { + "epoch": 2.9066045970322953, + "flow_matching_loss": 0.030413644388318062, + "grad_norm": 0.3982498049736023, + "learning_rate": 2.990945968675917e-05, + "loss": 0.0702, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0163942682556808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00770409656688571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04049263373017311, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.983015102148056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9990, + "weighted_flow_matching_loss": 0.030413644388318062, + "weighted_rkd_loss": 0.03980484940111637 + }, + { + "epoch": 2.909514111143439, + "flow_matching_loss": 0.03023599795997143, + "grad_norm": 0.35264045000076294, + "learning_rate": 2.990900562792633e-05, + "loss": 0.0691, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01604433488100767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007436945103108883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03952561467885971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.982981264591217, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10000, + "weighted_flow_matching_loss": 0.03023599795997143, + "weighted_rkd_loss": 0.038852959871292114 + }, + { + "epoch": 2.9124236252545828, + "flow_matching_loss": 0.031494008377194405, + "grad_norm": 0.36976921558380127, + "learning_rate": 2.9908550436852902e-05, + "loss": 0.0699, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015583949536085129, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007874644035473466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03904254361987114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9829473793506622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10010, + "weighted_flow_matching_loss": 0.031494008377194405, + "weighted_rkd_loss": 0.038376758247613905 + }, + { + "epoch": 2.915333139365726, + "flow_matching_loss": 0.0293395958840847, + "grad_norm": 0.3373558819293976, + "learning_rate": 2.9908094113573457e-05, + "loss": 0.0677, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015726190619170667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007577961031347513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03903034143149853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.982913464307785, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10020, + "weighted_flow_matching_loss": 0.0293395958840847, + "weighted_rkd_loss": 0.0383634289726615 + }, + { + "epoch": 2.9182426534768693, + "flow_matching_loss": 0.03389300983399153, + "grad_norm": 0.3296702802181244, + "learning_rate": 2.9907636658122645e-05, + "loss": 0.0716, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015399516187608242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007528478000313044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0383275106549263, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9828795373439789, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10030, + "weighted_flow_matching_loss": 0.03389300983399153, + "weighted_rkd_loss": 0.03767133485525846 + }, + { + "epoch": 2.921152167588013, + "flow_matching_loss": 0.030153940059244634, + "grad_norm": 0.3441750407218933, + "learning_rate": 2.9907178070535216e-05, + "loss": 0.0689, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01592646911740303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007587145501747728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03944008350372315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9828455567359924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10040, + "weighted_flow_matching_loss": 0.030153940059244634, + "weighted_rkd_loss": 0.03876351304352284 + }, + { + "epoch": 2.9240616816991563, + "flow_matching_loss": 0.033441484160721305, + "grad_norm": 0.39122074842453003, + "learning_rate": 2.990671835084599e-05, + "loss": 0.0732, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01626425180584192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007878070883452892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040406574681401256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9828115403652191, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10050, + "weighted_flow_matching_loss": 0.033441484160721305, + "weighted_rkd_loss": 0.039712049067020416 + }, + { + "epoch": 2.9269711958102995, + "flow_matching_loss": 0.030326366797089575, + "grad_norm": 0.3168375492095947, + "learning_rate": 2.990625749908988e-05, + "loss": 0.067, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015056203585118055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007232245150953531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03734465315937996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9827774882316589, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10060, + "weighted_flow_matching_loss": 0.030326366797089575, + "weighted_rkd_loss": 0.036701503396034243 + }, + { + "epoch": 2.9298807099214432, + "flow_matching_loss": 0.03085656389594078, + "grad_norm": 0.37854281067848206, + "learning_rate": 2.9905795515301883e-05, + "loss": 0.0682, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015342346765100956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00731017030775547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03799486383795738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9827434182167053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10070, + "weighted_flow_matching_loss": 0.03085656389594078, + "weighted_rkd_loss": 0.03733922392129898 + }, + { + "epoch": 2.9327902240325865, + "flow_matching_loss": 0.03427367229014635, + "grad_norm": 0.3288387656211853, + "learning_rate": 2.990533239951709e-05, + "loss": 0.0717, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015353884268552064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007332031475380063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03803979977965355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9827093064785004, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10080, + "weighted_flow_matching_loss": 0.03427367229014635, + "weighted_rkd_loss": 0.03738208748400211 + }, + { + "epoch": 2.9356997381437298, + "flow_matching_loss": 0.029458657279610632, + "grad_norm": 0.34328263998031616, + "learning_rate": 2.990486815177067e-05, + "loss": 0.0649, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014574660174548625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006935508223250508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036084828339517115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.982675164937973, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10090, + "weighted_flow_matching_loss": 0.029458657279610632, + "weighted_rkd_loss": 0.03545965645462275 + }, + { + "epoch": 2.9386092522548735, + "flow_matching_loss": 0.03238974791020155, + "grad_norm": 0.3504005968570709, + "learning_rate": 2.990440277209787e-05, + "loss": 0.0709, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015755252726376057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007639575516805053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03915008082985878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9826409876346588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10100, + "weighted_flow_matching_loss": 0.03238974791020155, + "weighted_rkd_loss": 0.03847045823931694 + }, + { + "epoch": 2.9415187663660167, + "flow_matching_loss": 0.029572294279932975, + "grad_norm": 0.3614276051521301, + "learning_rate": 2.9903936260534045e-05, + "loss": 0.0704, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016674296744167805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008175422344356776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04152401499450207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9826067805290222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10110, + "weighted_flow_matching_loss": 0.029572294279932975, + "weighted_rkd_loss": 0.04080182239413262 + }, + { + "epoch": 2.9444282804771604, + "flow_matching_loss": 0.031763963401317596, + "grad_norm": 0.3408048450946808, + "learning_rate": 2.990346861711462e-05, + "loss": 0.0665, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014270813949406148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006780191790312529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03532181903719902, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9825725257396698, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10120, + "weighted_flow_matching_loss": 0.031763963401317596, + "weighted_rkd_loss": 0.03470623455941677 + }, + { + "epoch": 2.9473377945883037, + "flow_matching_loss": 0.030370801500976086, + "grad_norm": 0.3750506639480591, + "learning_rate": 2.9902999841875105e-05, + "loss": 0.0686, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015776203479617834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007321971422061324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03887437842786312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.982538253068924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10130, + "weighted_flow_matching_loss": 0.030370801500976086, + "weighted_rkd_loss": 0.03819554410874844 + }, + { + "epoch": 2.9502473086994474, + "flow_matching_loss": 0.03086136169731617, + "grad_norm": 0.3329547047615051, + "learning_rate": 2.9902529934851107e-05, + "loss": 0.0698, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016134290397167204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00736929876729846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03963787965476513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9825039505958557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10140, + "weighted_flow_matching_loss": 0.03086136169731617, + "weighted_rkd_loss": 0.03894434720277786 + }, + { + "epoch": 2.9531568228105907, + "flow_matching_loss": 0.031074356846511365, + "grad_norm": 0.3452816605567932, + "learning_rate": 2.9902058896078304e-05, + "loss": 0.0683, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015228460170328617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00741058993153274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037867509946227076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9824696063995362, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10150, + "weighted_flow_matching_loss": 0.031074356846511365, + "weighted_rkd_loss": 0.0372036587446928 + }, + { + "epoch": 2.956066336921734, + "flow_matching_loss": 0.03282598052173853, + "grad_norm": 0.30603817105293274, + "learning_rate": 2.9901586725592475e-05, + "loss": 0.0699, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015311923250555992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007110016793012619, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037733863294124606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9824352324008941, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10160, + "weighted_flow_matching_loss": 0.03282598052173853, + "weighted_rkd_loss": 0.037071116268634796 + }, + { + "epoch": 2.9589758510328776, + "flow_matching_loss": 0.030659819953143597, + "grad_norm": 0.35752949118614197, + "learning_rate": 2.9901113423429473e-05, + "loss": 0.0673, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015006026066839696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007285771006718278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03729782290756702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9824008345603943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10170, + "weighted_flow_matching_loss": 0.030659819953143597, + "weighted_rkd_loss": 0.03664139844477177 + }, + { + "epoch": 2.961885365144021, + "flow_matching_loss": 0.028654619306325912, + "grad_norm": 0.33402180671691895, + "learning_rate": 2.9900638989625252e-05, + "loss": 0.0638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014484564308077097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006819050433114171, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03578817918896675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9823663890361786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10180, + "weighted_flow_matching_loss": 0.028654619306325912, + "weighted_rkd_loss": 0.03515710551291704 + }, + { + "epoch": 2.964794879255164, + "flow_matching_loss": 0.031355003081262114, + "grad_norm": 0.32568246126174927, + "learning_rate": 2.9900163424215832e-05, + "loss": 0.0695, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01576733561232686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007332663889974356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03886733502149582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9823319137096405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10190, + "weighted_flow_matching_loss": 0.031355003081262114, + "weighted_rkd_loss": 0.038180659897625445 + }, + { + "epoch": 2.967704393366308, + "flow_matching_loss": 0.029318691045045853, + "grad_norm": 0.38815945386886597, + "learning_rate": 2.9899686727237328e-05, + "loss": 0.0683, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015918875671923162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007858403585851192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03969615437090397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9822974085807801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10200, + "weighted_flow_matching_loss": 0.029318691045045853, + "weighted_rkd_loss": 0.03899342752993107 + }, + { + "epoch": 2.970613907477451, + "flow_matching_loss": 0.031431431882083416, + "grad_norm": 0.38513970375061035, + "learning_rate": 2.9899208898725954e-05, + "loss": 0.0653, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013959819544106722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006562385894358158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03448202479630709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9822628498077393, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10210, + "weighted_flow_matching_loss": 0.031431431882083416, + "weighted_rkd_loss": 0.03387041017413139 + }, + { + "epoch": 2.973523421588595, + "flow_matching_loss": 0.029533158987760544, + "grad_norm": 0.363389790058136, + "learning_rate": 2.9898729938717984e-05, + "loss": 0.0662, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015101512894034386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007146383356302977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037349409237504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9822282791137695, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10220, + "weighted_flow_matching_loss": 0.029533158987760544, + "weighted_rkd_loss": 0.036685635522007945 + }, + { + "epoch": 2.976432935699738, + "flow_matching_loss": 0.02966185975819826, + "grad_norm": 0.3418305218219757, + "learning_rate": 2.98982498472498e-05, + "loss": 0.0649, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014578954316675663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006719372468069196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03587728086858988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9821936905384063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10230, + "weighted_flow_matching_loss": 0.02966185975819826, + "weighted_rkd_loss": 0.035238408111035825 + }, + { + "epoch": 2.979342449810882, + "flow_matching_loss": 0.03128401339054108, + "grad_norm": 0.35166648030281067, + "learning_rate": 2.9897768624357865e-05, + "loss": 0.0681, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01527472548186779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0069660032633692024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03751545380800962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9821590483188629, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10240, + "weighted_flow_matching_loss": 0.03128401339054108, + "weighted_rkd_loss": 0.03684612680226564 + }, + { + "epoch": 2.982251963922025, + "flow_matching_loss": 0.028522592782974244, + "grad_norm": 0.28424859046936035, + "learning_rate": 2.9897286270078713e-05, + "loss": 0.0622, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01404599631205201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006208900967612863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034300893731415275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9821243822574616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10250, + "weighted_flow_matching_loss": 0.028522592782974244, + "weighted_rkd_loss": 0.033687744103372094 + }, + { + "epoch": 2.9851614780331683, + "flow_matching_loss": 0.033113792166113855, + "grad_norm": 0.3943287432193756, + "learning_rate": 2.9896802784448986e-05, + "loss": 0.0724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016132846660912036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007705228170379996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03997092172503471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9820896685123444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10260, + "weighted_flow_matching_loss": 0.033113792166113855, + "weighted_rkd_loss": 0.039255044981837274 + }, + { + "epoch": 2.988070992144312, + "flow_matching_loss": 0.03133232444524765, + "grad_norm": 0.31870242953300476, + "learning_rate": 2.9896318167505393e-05, + "loss": 0.0681, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015027325972914696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007433414319530129, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03748806677758694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9820549368858338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10270, + "weighted_flow_matching_loss": 0.03133232444524765, + "weighted_rkd_loss": 0.03681534845381975 + }, + { + "epoch": 2.9909805062554553, + "flow_matching_loss": 0.03087249081581831, + "grad_norm": 0.34085342288017273, + "learning_rate": 2.9895832419284748e-05, + "loss": 0.067, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015010929852724075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006764618260785937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03678647764027119, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9820201635360718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10280, + "weighted_flow_matching_loss": 0.03087249081581831, + "weighted_rkd_loss": 0.03612504675984383 + }, + { + "epoch": 2.9938900203665986, + "flow_matching_loss": 0.02931579276919365, + "grad_norm": 0.41113346815109253, + "learning_rate": 2.9895345539823934e-05, + "loss": 0.0662, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015203847736120223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007135013351216912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03754270821809769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9819853603839874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10290, + "weighted_flow_matching_loss": 0.02931579276919365, + "weighted_rkd_loss": 0.03686637114733458 + }, + { + "epoch": 2.9967995344777423, + "flow_matching_loss": 0.0295780124142766, + "grad_norm": 0.3289874196052551, + "learning_rate": 2.9894857529159927e-05, + "loss": 0.0664, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015164956357330084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007195488829165697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037525401636958125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9819505214691162, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10300, + "weighted_flow_matching_loss": 0.0295780124142766, + "weighted_rkd_loss": 0.036848079226911065 + }, + { + "epoch": 2.9997090485888855, + "flow_matching_loss": 0.032499029859900476, + "grad_norm": 0.3378007709980011, + "learning_rate": 2.9894368387329788e-05, + "loss": 0.0682, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014731732755899429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006943063158541918, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03640652820467949, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9819156527519226, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10310, + "weighted_flow_matching_loss": 0.032499029859900476, + "weighted_rkd_loss": 0.03574811853468418 + }, + { + "epoch": 3.0026185627000292, + "flow_matching_loss": 0.03248726539313793, + "grad_norm": 0.31760597229003906, + "learning_rate": 2.989387811437066e-05, + "loss": 0.0673, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014388822577893734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006639323476701975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03541696928441525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9818807482719422, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10320, + "weighted_flow_matching_loss": 0.03248726539313793, + "weighted_rkd_loss": 0.034775204956531525 + }, + { + "epoch": 3.0055280768111725, + "flow_matching_loss": 0.03238500859588385, + "grad_norm": 0.366856187582016, + "learning_rate": 2.9893386710319786e-05, + "loss": 0.0684, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014933092333376408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006783575424924493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03664976023137569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9818458080291748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10330, + "weighted_flow_matching_loss": 0.03238500859588385, + "weighted_rkd_loss": 0.03598442357033491 + }, + { + "epoch": 3.0084375909223158, + "flow_matching_loss": 0.031213024631142616, + "grad_norm": 0.36720022559165955, + "learning_rate": 2.9892894175214474e-05, + "loss": 0.0691, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015527112316340208, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007496835757046938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03855106104165316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9818108439445495, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10340, + "weighted_flow_matching_loss": 0.031213024631142616, + "weighted_rkd_loss": 0.03784984797239303 + }, + { + "epoch": 3.0113471050334595, + "flow_matching_loss": 0.0303227785974741, + "grad_norm": 0.32786181569099426, + "learning_rate": 2.9892400509092136e-05, + "loss": 0.0674, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015246870461851358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007313787937164307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037807528674602506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.981775838136673, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10350, + "weighted_flow_matching_loss": 0.0303227785974741, + "weighted_rkd_loss": 0.037118513509631156 + }, + { + "epoch": 3.0142566191446027, + "flow_matching_loss": 0.030800014920532703, + "grad_norm": 0.31551143527030945, + "learning_rate": 2.989190571199026e-05, + "loss": 0.066, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01448621265590191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006850304221734405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03582272976636887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9817408084869385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10360, + "weighted_flow_matching_loss": 0.030800014920532703, + "weighted_rkd_loss": 0.035168631747364995 + }, + { + "epoch": 3.0171661332557465, + "flow_matching_loss": 0.029436804354190826, + "grad_norm": 0.3219139575958252, + "learning_rate": 2.9891409783946426e-05, + "loss": 0.0637, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014201289135962724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006530951336026191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034933529794216156, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9817057311534881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10370, + "weighted_flow_matching_loss": 0.029436804354190826, + "weighted_rkd_loss": 0.03429445549845696 + }, + { + "epoch": 3.0200756473668897, + "flow_matching_loss": 0.029693375714123248, + "grad_norm": 0.3354239761829376, + "learning_rate": 2.9890912724998293e-05, + "loss": 0.0663, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015164616703987121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007004358293488622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03733359221369028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9816706359386445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10380, + "weighted_flow_matching_loss": 0.029693375714123248, + "weighted_rkd_loss": 0.03664927054196596 + }, + { + "epoch": 3.022985161478033, + "flow_matching_loss": 0.03003329299390316, + "grad_norm": 0.34405848383903503, + "learning_rate": 2.9890414535183605e-05, + "loss": 0.0661, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014879832696169614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0069656061008572575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03672527112066746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9816354811191559, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10390, + "weighted_flow_matching_loss": 0.03003329299390316, + "weighted_rkd_loss": 0.03605082146823406 + }, + { + "epoch": 3.0258946755891767, + "flow_matching_loss": 0.03170877378433943, + "grad_norm": 0.32547327876091003, + "learning_rate": 2.9889915214540206e-05, + "loss": 0.0675, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014664823561906815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007170584611594677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03650023210793733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9816003143787384, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10400, + "weighted_flow_matching_loss": 0.03170877378433943, + "weighted_rkd_loss": 0.03582864869385958 + }, + { + "epoch": 3.02880418970032, + "flow_matching_loss": 0.030512782372534276, + "grad_norm": 0.3456466495990753, + "learning_rate": 2.9889414763106013e-05, + "loss": 0.0672, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015187212824821472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007022306276485324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037396731600165364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.981565135717392, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10410, + "weighted_flow_matching_loss": 0.030512782372534276, + "weighted_rkd_loss": 0.036707335710525514 + }, + { + "epoch": 3.0317137038114637, + "flow_matching_loss": 0.03388255685567856, + "grad_norm": 0.4176626205444336, + "learning_rate": 2.9888913180919028e-05, + "loss": 0.0699, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014786242973059416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007083292119204998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0366557776927948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9815298974514007, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10420, + "weighted_flow_matching_loss": 0.03388255685567856, + "weighted_rkd_loss": 0.03597874771803618 + }, + { + "epoch": 3.034623217922607, + "flow_matching_loss": 0.032636011578142644, + "grad_norm": 0.3811602294445038, + "learning_rate": 2.9888410468017346e-05, + "loss": 0.0683, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014681679755449295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007015439169481397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03637879844754934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9814946293830872, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10430, + "weighted_flow_matching_loss": 0.032636011578142644, + "weighted_rkd_loss": 0.03570557367056608 + }, + { + "epoch": 3.03753273203375, + "flow_matching_loss": 0.029281178675591945, + "grad_norm": 0.3670819401741028, + "learning_rate": 2.9887906624439142e-05, + "loss": 0.0644, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01459452398121357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006637755827978253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035826803743839265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9814593374729157, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10440, + "weighted_flow_matching_loss": 0.029281178675591945, + "weighted_rkd_loss": 0.03516255244612694 + }, + { + "epoch": 3.040442246144894, + "flow_matching_loss": 0.030722532607614993, + "grad_norm": 0.39905425906181335, + "learning_rate": 2.988740165022268e-05, + "loss": 0.0677, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015060537215322255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007526475144550205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0376475490629673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9814240038394928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10450, + "weighted_flow_matching_loss": 0.030722532607614993, + "weighted_rkd_loss": 0.03694818690419197 + }, + { + "epoch": 3.043351760256037, + "flow_matching_loss": 0.03268843274563551, + "grad_norm": 0.3799138069152832, + "learning_rate": 2.9886895545406317e-05, + "loss": 0.0698, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015384653117507696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006995430449023843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03776473701000214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9813886344432831, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10460, + "weighted_flow_matching_loss": 0.03268843274563551, + "weighted_rkd_loss": 0.03706188052892685 + }, + { + "epoch": 3.046261274367181, + "flow_matching_loss": 0.02991221658885479, + "grad_norm": 0.35373756289482117, + "learning_rate": 2.9886388310028478e-05, + "loss": 0.0677, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01564167132601142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00723342914134264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03851677142083645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9813532412052155, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10470, + "weighted_flow_matching_loss": 0.02991221658885479, + "weighted_rkd_loss": 0.037798546627163886 + }, + { + "epoch": 3.049170788478324, + "flow_matching_loss": 0.030286012031137944, + "grad_norm": 0.3757373094558716, + "learning_rate": 2.988587994412769e-05, + "loss": 0.0655, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014605336543172597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006673857104033232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03588453065603971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9813178062438965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10480, + "weighted_flow_matching_loss": 0.030286012031137944, + "weighted_rkd_loss": 0.03521410450339317 + }, + { + "epoch": 3.0520803025894674, + "flow_matching_loss": 0.031802080757915976, + "grad_norm": 0.34756240248680115, + "learning_rate": 2.9885370447742556e-05, + "loss": 0.0702, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015924621745944022, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007261075731366873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039110319688916206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9812823474407196, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10490, + "weighted_flow_matching_loss": 0.031802080757915976, + "weighted_rkd_loss": 0.03837827481329441 + }, + { + "epoch": 3.054989816700611, + "flow_matching_loss": 0.029601309262216092, + "grad_norm": 0.3968396484851837, + "learning_rate": 2.9884859820911773e-05, + "loss": 0.0684, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01578692886978388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00791957527399063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039493432641029357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9812468409538269, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10500, + "weighted_flow_matching_loss": 0.029601309262216092, + "weighted_rkd_loss": 0.03875279501080513 + }, + { + "epoch": 3.0578993308117544, + "flow_matching_loss": 0.03040149752050638, + "grad_norm": 0.3993563652038574, + "learning_rate": 2.988434806367412e-05, + "loss": 0.0655, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014548169635236263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006636997731402516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035733336582779884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9812113106250763, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10510, + "weighted_flow_matching_loss": 0.03040149752050638, + "weighted_rkd_loss": 0.03506194483488798 + }, + { + "epoch": 3.060808844922898, + "flow_matching_loss": 0.029813078232109547, + "grad_norm": 0.3366228938102722, + "learning_rate": 2.9883835176068455e-05, + "loss": 0.0672, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015440325438976287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007247445592656732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038128096610307693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9811757504940033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10520, + "weighted_flow_matching_loss": 0.029813078232109547, + "weighted_rkd_loss": 0.03741036206483841 + }, + { + "epoch": 3.0637183590340413, + "flow_matching_loss": 0.03099121432751417, + "grad_norm": 0.36431437730789185, + "learning_rate": 2.9883321158133736e-05, + "loss": 0.0692, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015694080386310816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00758096044883132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0389691211283207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9811401486396789, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10530, + "weighted_flow_matching_loss": 0.03099121432751417, + "weighted_rkd_loss": 0.03823417387902737 + }, + { + "epoch": 3.0666278731451846, + "flow_matching_loss": 0.030911387130618095, + "grad_norm": 0.3318611979484558, + "learning_rate": 2.9882806009908996e-05, + "loss": 0.0655, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01443038359284401, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006437523290514946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03529829103499651, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9811045229434967, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10540, + "weighted_flow_matching_loss": 0.030911387130618095, + "weighted_rkd_loss": 0.034631319157779214 + }, + { + "epoch": 3.0695373872563283, + "flow_matching_loss": 0.02974409405142069, + "grad_norm": 0.36098772287368774, + "learning_rate": 2.988228973143336e-05, + "loss": 0.0648, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014442999381572008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006837596325203777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035723595321178435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9810688674449921, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10550, + "weighted_flow_matching_loss": 0.02974409405142069, + "weighted_rkd_loss": 0.0350473091006279 + }, + { + "epoch": 3.0724469013674716, + "flow_matching_loss": 0.03076146300882101, + "grad_norm": 0.3366888761520386, + "learning_rate": 2.988177232274603e-05, + "loss": 0.0667, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014977442380040885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006664529722183943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03661941420286894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9810331642627717, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10560, + "weighted_flow_matching_loss": 0.03076146300882101, + "weighted_rkd_loss": 0.03592487089335918 + }, + { + "epoch": 3.0753564154786153, + "flow_matching_loss": 0.03066011667251587, + "grad_norm": 0.31057241559028625, + "learning_rate": 2.9881253783886305e-05, + "loss": 0.0657, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014497394394129514, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00676913377828896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03576392251998186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9809974431991577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10570, + "weighted_flow_matching_loss": 0.03066011667251587, + "weighted_rkd_loss": 0.03508431576192379 + }, + { + "epoch": 3.0782659295897585, + "flow_matching_loss": 0.029927757941186427, + "grad_norm": 0.34214097261428833, + "learning_rate": 2.988073411489356e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014819363504648209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006517060706391931, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03615578766912222, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9809616804122925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10580, + "weighted_flow_matching_loss": 0.029927757941186427, + "weighted_rkd_loss": 0.035467441007494924 + }, + { + "epoch": 3.081175443700902, + "flow_matching_loss": 0.03095151074230671, + "grad_norm": 0.4146866798400879, + "learning_rate": 2.988021331580727e-05, + "loss": 0.0676, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015126648545265197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007072295993566513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037325592897832396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9809258818626404, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10590, + "weighted_flow_matching_loss": 0.03095151074230671, + "weighted_rkd_loss": 0.036613631434738635 + }, + { + "epoch": 3.0840849578120455, + "flow_matching_loss": 0.02893105410039425, + "grad_norm": 0.2938387095928192, + "learning_rate": 2.9879691386666976e-05, + "loss": 0.0646, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014703127089887857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00693670785985887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0363429618999362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9808900594711304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10600, + "weighted_flow_matching_loss": 0.02893105410039425, + "weighted_rkd_loss": 0.03564848732203245 + }, + { + "epoch": 3.0869944719231888, + "flow_matching_loss": 0.02990221567451954, + "grad_norm": 0.3231969177722931, + "learning_rate": 2.9879168327512315e-05, + "loss": 0.0663, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014935999829322099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071947521064430475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037066752463579176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.980854195356369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10610, + "weighted_flow_matching_loss": 0.02990221567451954, + "weighted_rkd_loss": 0.03635706789791584 + }, + { + "epoch": 3.0899039860343325, + "flow_matching_loss": 0.030821909569203852, + "grad_norm": 0.383573055267334, + "learning_rate": 2.987864413838302e-05, + "loss": 0.0685, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015449199732393027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007538378145545721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038436776772141454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9808183014392853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10620, + "weighted_flow_matching_loss": 0.030821909569203852, + "weighted_rkd_loss": 0.03769950661808252 + }, + { + "epoch": 3.0928135001454757, + "flow_matching_loss": 0.030145752429962158, + "grad_norm": 0.34465664625167847, + "learning_rate": 2.9878118819318887e-05, + "loss": 0.0663, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014913569763302803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007008671341463923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0368358101695776, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9807823777198792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10630, + "weighted_flow_matching_loss": 0.030145752429962158, + "weighted_rkd_loss": 0.036127909831702706 + }, + { + "epoch": 3.095723014256619, + "flow_matching_loss": 0.032468338310718534, + "grad_norm": 0.4033108055591583, + "learning_rate": 2.987759237035982e-05, + "loss": 0.0694, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015333165228366853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006967388791963458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03763371892273426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9807464182376862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10640, + "weighted_flow_matching_loss": 0.032468338310718534, + "weighted_rkd_loss": 0.036909115873277186 + }, + { + "epoch": 3.0986325283677627, + "flow_matching_loss": 0.030924981459975243, + "grad_norm": 0.35899507999420166, + "learning_rate": 2.9877064791545795e-05, + "loss": 0.0655, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014345829002559185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0066061326768249275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03529779072850943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9807104229927063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10650, + "weighted_flow_matching_loss": 0.030924981459975243, + "weighted_rkd_loss": 0.03461691401898861 + }, + { + "epoch": 3.101542042478906, + "flow_matching_loss": 0.028863153792917727, + "grad_norm": 0.4276405870914459, + "learning_rate": 2.9876536082916878e-05, + "loss": 0.0649, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014926972705870867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006847201799973846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036701147072017194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9806744039058686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10660, + "weighted_flow_matching_loss": 0.028863153792917727, + "weighted_rkd_loss": 0.035991879738867284 + }, + { + "epoch": 3.1044515565900497, + "flow_matching_loss": 0.03175907842814922, + "grad_norm": 0.318152517080307, + "learning_rate": 2.9876006244513228e-05, + "loss": 0.0683, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014994627516716718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072617890313267704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03725104369223118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9806383490562439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10670, + "weighted_flow_matching_loss": 0.03175907842814922, + "weighted_rkd_loss": 0.036529820412397385 + }, + { + "epoch": 3.107361070701193, + "flow_matching_loss": 0.03160724509507418, + "grad_norm": 0.33375513553619385, + "learning_rate": 2.987547527637507e-05, + "loss": 0.0666, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014470606949180365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0067911673802882435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03573238104581833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.980602252483368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10680, + "weighted_flow_matching_loss": 0.03160724509507418, + "weighted_rkd_loss": 0.03503926321864128 + }, + { + "epoch": 3.110270584812336, + "flow_matching_loss": 0.030068336613476278, + "grad_norm": 0.3397270739078522, + "learning_rate": 2.987494317854274e-05, + "loss": 0.0666, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015011134278029203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007274024887010455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037296293303370476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9805661320686341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10690, + "weighted_flow_matching_loss": 0.030068336613476278, + "weighted_rkd_loss": 0.036571497842669484 + }, + { + "epoch": 3.11318009892348, + "flow_matching_loss": 0.031574487872421744, + "grad_norm": 0.3658752739429474, + "learning_rate": 2.9874409951056634e-05, + "loss": 0.0661, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014420701749622821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006342009687796235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03518341388553381, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9805299699306488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10700, + "weighted_flow_matching_loss": 0.031574487872421744, + "weighted_rkd_loss": 0.03449836242944002 + }, + { + "epoch": 3.116089613034623, + "flow_matching_loss": 0.03251878675073385, + "grad_norm": 0.38086798787117004, + "learning_rate": 2.987387559395726e-05, + "loss": 0.0723, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01628651553764939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007949976017698646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04052300713956356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9804937839508057, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10710, + "weighted_flow_matching_loss": 0.03251878675073385, + "weighted_rkd_loss": 0.03973255418241024 + }, + { + "epoch": 3.1189991271457664, + "flow_matching_loss": 0.02950261440128088, + "grad_norm": 0.3198232650756836, + "learning_rate": 2.9873340107285197e-05, + "loss": 0.0622, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013458691071718931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006438853032886982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03335623573511839, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9804575562477111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10720, + "weighted_flow_matching_loss": 0.02950261440128088, + "weighted_rkd_loss": 0.032704390771687034 + }, + { + "epoch": 3.12190864125691, + "flow_matching_loss": 0.03059883490204811, + "grad_norm": 0.3952866792678833, + "learning_rate": 2.9872803491081105e-05, + "loss": 0.0685, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015590534452348948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007510083401575685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03869115263223648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9804213047027588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10730, + "weighted_flow_matching_loss": 0.03059883490204811, + "weighted_rkd_loss": 0.03793363533914089 + }, + { + "epoch": 3.1248181553680534, + "flow_matching_loss": 0.03162776492536068, + "grad_norm": 0.35058748722076416, + "learning_rate": 2.9872265745385737e-05, + "loss": 0.0674, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014788756519556046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006914553744718432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03649206701666117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9803850173950195, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10740, + "weighted_flow_matching_loss": 0.03162776492536068, + "weighted_rkd_loss": 0.035776298493146896 + }, + { + "epoch": 3.127727669479197, + "flow_matching_loss": 0.02983020767569542, + "grad_norm": 0.4208681583404541, + "learning_rate": 2.9871726870239937e-05, + "loss": 0.0642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01421569287776947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006655228463932872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03508661389350891, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9803486764431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10750, + "weighted_flow_matching_loss": 0.02983020767569542, + "weighted_rkd_loss": 0.034397108294069764 + }, + { + "epoch": 3.1306371835903404, + "flow_matching_loss": 0.031107492186129092, + "grad_norm": 0.3781690299510956, + "learning_rate": 2.9871186865684624e-05, + "loss": 0.0641, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013543580844998359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006533853569999337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03362101558595896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.980312317609787, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10760, + "weighted_flow_matching_loss": 0.031107492186129092, + "weighted_rkd_loss": 0.03295907136052847 + }, + { + "epoch": 3.133546697701484, + "flow_matching_loss": 0.028162371553480624, + "grad_norm": 0.29834046959877014, + "learning_rate": 2.987064573176081e-05, + "loss": 0.0628, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01438142154365778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0065693173091858625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03533216044306755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.980275958776474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10770, + "weighted_flow_matching_loss": 0.028162371553480624, + "weighted_rkd_loss": 0.0346352893859148 + }, + { + "epoch": 3.1364562118126273, + "flow_matching_loss": 0.030513253062963486, + "grad_norm": 0.3533119857311249, + "learning_rate": 2.9870103468509595e-05, + "loss": 0.0662, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014727573003619909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0069133595563471316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036368504725396635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9802395284175873, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10780, + "weighted_flow_matching_loss": 0.030513253062963486, + "weighted_rkd_loss": 0.03564983084797859 + }, + { + "epoch": 3.1393657259237706, + "flow_matching_loss": 0.030064966529607773, + "grad_norm": 0.34244832396507263, + "learning_rate": 2.986956007597215e-05, + "loss": 0.0627, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013421309366822242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006493420107290149, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03333603851497173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9802030801773072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10790, + "weighted_flow_matching_loss": 0.030064966529607773, + "weighted_rkd_loss": 0.03267607931047678 + }, + { + "epoch": 3.1422752400349143, + "flow_matching_loss": 0.029571990482509136, + "grad_norm": 0.3440578281879425, + "learning_rate": 2.986901555418975e-05, + "loss": 0.0679, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015820217691361903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007496656570583582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03913709241896868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9801665961742401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10800, + "weighted_flow_matching_loss": 0.029571990482509136, + "weighted_rkd_loss": 0.038360876590013505 + }, + { + "epoch": 3.1451847541460576, + "flow_matching_loss": 0.028918012417852877, + "grad_norm": 0.37640929222106934, + "learning_rate": 2.9868469903203746e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013935538195073604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006348038744181395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034219115041196346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9801300764083862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10810, + "weighted_flow_matching_loss": 0.028918012417852877, + "weighted_rkd_loss": 0.0335391890257597 + }, + { + "epoch": 3.148094268257201, + "flow_matching_loss": 0.030492972955107688, + "grad_norm": 0.33921051025390625, + "learning_rate": 2.986792312305558e-05, + "loss": 0.0644, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013968116231262684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006639485713094473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034575718268752095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9800935208797454, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10820, + "weighted_flow_matching_loss": 0.030492972955107688, + "weighted_rkd_loss": 0.033887443132698536 + }, + { + "epoch": 3.1510037823683446, + "flow_matching_loss": 0.03234951794147491, + "grad_norm": 0.3108830153942108, + "learning_rate": 2.9867375213786762e-05, + "loss": 0.0664, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014071681443601847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00661075422540307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03475411683320999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9800569355487824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10830, + "weighted_flow_matching_loss": 0.03234951794147491, + "weighted_rkd_loss": 0.034061025269329546 + }, + { + "epoch": 3.153913296479488, + "flow_matching_loss": 0.0330373153090477, + "grad_norm": 0.32859674096107483, + "learning_rate": 2.9866826175438922e-05, + "loss": 0.0694, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01499395314604044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007083098217844963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037071004137396814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9800203144550323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10840, + "weighted_flow_matching_loss": 0.0330373153090477, + "weighted_rkd_loss": 0.03633034154772759 + }, + { + "epoch": 3.1568228105906315, + "flow_matching_loss": 0.030216096714138985, + "grad_norm": 0.3406435549259186, + "learning_rate": 2.9866276008053742e-05, + "loss": 0.0635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013743752241134643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006455982849001884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03394348751753569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.979983675479889, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10850, + "weighted_flow_matching_loss": 0.030216096714138985, + "weighted_rkd_loss": 0.03326407764106989 + }, + { + "epoch": 3.159732324701775, + "flow_matching_loss": 0.027957860194146634, + "grad_norm": 0.30357080698013306, + "learning_rate": 2.986572471167301e-05, + "loss": 0.063, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014469632040709257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0067852107807993885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0357244748622179, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9799469888210297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10860, + "weighted_flow_matching_loss": 0.027957860194146634, + "weighted_rkd_loss": 0.03500808961689472 + }, + { + "epoch": 3.162641838812918, + "flow_matching_loss": 0.03049442023038864, + "grad_norm": 0.3256266117095947, + "learning_rate": 2.986517228633859e-05, + "loss": 0.0659, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014627727214246988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0068484105635434386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0361038651317358, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9799102723598481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10870, + "weighted_flow_matching_loss": 0.03049442023038864, + "weighted_rkd_loss": 0.0353785814717412 + }, + { + "epoch": 3.1655513529240618, + "flow_matching_loss": 0.029552517272531987, + "grad_norm": 0.3193569481372833, + "learning_rate": 2.9864618732092438e-05, + "loss": 0.0667, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015423110406845807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007030159374698997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0378763796761632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.979873526096344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10880, + "weighted_flow_matching_loss": 0.029552517272531987, + "weighted_rkd_loss": 0.03711406234651804 + }, + { + "epoch": 3.168460867035205, + "flow_matching_loss": 0.030192114226520062, + "grad_norm": 0.3838639259338379, + "learning_rate": 2.9864064048976593e-05, + "loss": 0.066, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014859953429549932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00684376354329288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03656367063522339, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9798367440700531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10890, + "weighted_flow_matching_loss": 0.030192114226520062, + "weighted_rkd_loss": 0.03582640346139669 + }, + { + "epoch": 3.1713703811463487, + "flow_matching_loss": 0.03187324721366167, + "grad_norm": 0.32675909996032715, + "learning_rate": 2.9863508237033176e-05, + "loss": 0.0671, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01460365429520607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006734151626005769, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03594146016985178, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9797999262809753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10900, + "weighted_flow_matching_loss": 0.03187324721366167, + "weighted_rkd_loss": 0.03521543741226196 + }, + { + "epoch": 3.174279895257492, + "flow_matching_loss": 0.028120225854218008, + "grad_norm": 0.3644017279148102, + "learning_rate": 2.9862951296304398e-05, + "loss": 0.0613, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013697547279298306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0064609995111823085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03385609406977892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9797630727291107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10910, + "weighted_flow_matching_loss": 0.028120225854218008, + "weighted_rkd_loss": 0.03317097816616297 + }, + { + "epoch": 3.1771894093686353, + "flow_matching_loss": 0.029976310953497887, + "grad_norm": 0.3095386028289795, + "learning_rate": 2.9862393226832555e-05, + "loss": 0.0627, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013599882833659649, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006251218635588885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03345098402351141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9797261953353882, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10920, + "weighted_flow_matching_loss": 0.029976310953497887, + "weighted_rkd_loss": 0.03277280181646347 + }, + { + "epoch": 3.180098923479779, + "flow_matching_loss": 0.03275572843849659, + "grad_norm": 0.37710338830947876, + "learning_rate": 2.986183402866003e-05, + "loss": 0.0695, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01509653739631176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007277495926246047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037470570765435696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9796893000602722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10930, + "weighted_flow_matching_loss": 0.03275572843849659, + "weighted_rkd_loss": 0.03670946005731821 + }, + { + "epoch": 3.1830084375909222, + "flow_matching_loss": 0.029657895863056182, + "grad_norm": 0.3639652132987976, + "learning_rate": 2.9861273701829295e-05, + "loss": 0.0673, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015371038485318423, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007690804079174995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038432881236076355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9796523451805115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10940, + "weighted_flow_matching_loss": 0.029657895863056182, + "weighted_rkd_loss": 0.037650853767991065 + }, + { + "epoch": 3.185917951702066, + "flow_matching_loss": 0.0300249669700861, + "grad_norm": 0.41222435235977173, + "learning_rate": 2.986071224638289e-05, + "loss": 0.0655, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014586922992020846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007004115032032132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03617796134203673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9796153604984283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10950, + "weighted_flow_matching_loss": 0.0300249669700861, + "weighted_rkd_loss": 0.03544043768197298 + }, + { + "epoch": 3.188827465813209, + "flow_matching_loss": 0.02953706681728363, + "grad_norm": 0.3469090163707733, + "learning_rate": 2.9860149662363467e-05, + "loss": 0.0667, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01542530683800578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00707665435038507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037927267514169215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9795783460140228, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10960, + "weighted_flow_matching_loss": 0.02953706681728363, + "weighted_rkd_loss": 0.03715271819382906 + }, + { + "epoch": 3.1917369799243525, + "flow_matching_loss": 0.028660468570888043, + "grad_norm": 0.3432318866252899, + "learning_rate": 2.9859585949813745e-05, + "loss": 0.0632, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014218205586075783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006813503801822662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03524991497397423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9795413017272949, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10970, + "weighted_flow_matching_loss": 0.028660468570888043, + "weighted_rkd_loss": 0.03452873807400465 + }, + { + "epoch": 3.194646494035496, + "flow_matching_loss": 0.03112939912825823, + "grad_norm": 0.4426761567592621, + "learning_rate": 2.9859021108776538e-05, + "loss": 0.067, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014951619319617748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006729527888819575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03663276676088571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9795042276382446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10980, + "weighted_flow_matching_loss": 0.03112939912825823, + "weighted_rkd_loss": 0.035881947912275794 + }, + { + "epoch": 3.1975560081466394, + "flow_matching_loss": 0.03117939326912165, + "grad_norm": 0.3795885741710663, + "learning_rate": 2.9858455139294737e-05, + "loss": 0.065, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01408102000132203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063692627940326926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03453130330890417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.979467111825943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10990, + "weighted_flow_matching_loss": 0.03117939326912165, + "weighted_rkd_loss": 0.03382229562848806 + }, + { + "epoch": 3.200465522257783, + "flow_matching_loss": 0.028322724066674708, + "grad_norm": 0.3729858696460724, + "learning_rate": 2.9857888041411324e-05, + "loss": 0.0644, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014975335635244846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00684155635535717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036792227812111375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9794299721717834, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11000, + "weighted_flow_matching_loss": 0.028322724066674708, + "weighted_rkd_loss": 0.0360354034230113 + }, + { + "epoch": 3.2033750363689264, + "flow_matching_loss": 0.029921170324087143, + "grad_norm": 0.2911995053291321, + "learning_rate": 2.985731981516937e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013434780295938253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061777109745889906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03304727151989937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9793927907943726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11010, + "weighted_flow_matching_loss": 0.029921170324087143, + "weighted_rkd_loss": 0.0323662469163537 + }, + { + "epoch": 3.2062845504800697, + "flow_matching_loss": 0.02961389198899269, + "grad_norm": 0.37303048372268677, + "learning_rate": 2.985675046061202e-05, + "loss": 0.0633, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01405106671154499, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006325650587677955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03442778438329697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9793555796146393, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11020, + "weighted_flow_matching_loss": 0.02961389198899269, + "weighted_rkd_loss": 0.03371702767908573 + }, + { + "epoch": 3.2091940645912134, + "flow_matching_loss": 0.028724913857877253, + "grad_norm": 0.29325661063194275, + "learning_rate": 2.985617997778252e-05, + "loss": 0.0638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014521501399576664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006760867265984416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03580387011170387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9793183505535126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11030, + "weighted_flow_matching_loss": 0.028724913857877253, + "weighted_rkd_loss": 0.035063383355736735 + }, + { + "epoch": 3.2121035787023566, + "flow_matching_loss": 0.029639259353280066, + "grad_norm": 0.416272908449173, + "learning_rate": 2.9855608366724196e-05, + "loss": 0.0643, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014189556892961263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006982477894052863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03536159172654152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9792810618877411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11040, + "weighted_flow_matching_loss": 0.029639259353280066, + "weighted_rkd_loss": 0.03462894707918167 + }, + { + "epoch": 3.2150130928135003, + "flow_matching_loss": 0.0310451565310359, + "grad_norm": 0.35225948691368103, + "learning_rate": 2.9855035627480456e-05, + "loss": 0.0661, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014467387273907662, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0068619792815297845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035796753503382205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9792437553405762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11050, + "weighted_flow_matching_loss": 0.0310451565310359, + "weighted_rkd_loss": 0.03505373802036047 + }, + { + "epoch": 3.2179226069246436, + "flow_matching_loss": 0.03106956575065851, + "grad_norm": 0.38241758942604065, + "learning_rate": 2.985446176009479e-05, + "loss": 0.0662, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0145138886757195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006835978850722313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03586375657469034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9792064189910888, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11060, + "weighted_flow_matching_loss": 0.03106956575065851, + "weighted_rkd_loss": 0.035118036158382895 + }, + { + "epoch": 3.220832121035787, + "flow_matching_loss": 0.030384113639593126, + "grad_norm": 0.3245283365249634, + "learning_rate": 2.985388676461079e-05, + "loss": 0.0647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014247205294668674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00655417125672102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035048581473529336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9791690409183502, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11070, + "weighted_flow_matching_loss": 0.030384113639593126, + "weighted_rkd_loss": 0.034318477660417554 + }, + { + "epoch": 3.2237416351469306, + "flow_matching_loss": 0.030388077907264232, + "grad_norm": 0.35135790705680847, + "learning_rate": 2.9853310641072117e-05, + "loss": 0.0631, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013481579627841711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006470065843313932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03343322593718767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9791316330432892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11080, + "weighted_flow_matching_loss": 0.030388077907264232, + "weighted_rkd_loss": 0.032735512033104895 + }, + { + "epoch": 3.226651149258074, + "flow_matching_loss": 0.02885448858141899, + "grad_norm": 0.33482626080513, + "learning_rate": 2.985273338952252e-05, + "loss": 0.0627, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014039794448763132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006523747742176056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034603336825966835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9790942013263703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11090, + "weighted_flow_matching_loss": 0.02885448858141899, + "weighted_rkd_loss": 0.033879895135760306 + }, + { + "epoch": 3.2295606633692175, + "flow_matching_loss": 0.03075597882270813, + "grad_norm": 0.34509173035621643, + "learning_rate": 2.9852155010005843e-05, + "loss": 0.0629, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013332256488502026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006146048847585917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03281056173145771, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9790567278861999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11100, + "weighted_flow_matching_loss": 0.03075597882270813, + "weighted_rkd_loss": 0.03212340772151947 + }, + { + "epoch": 3.232470177480361, + "flow_matching_loss": 0.030018402822315694, + "grad_norm": 0.38414257764816284, + "learning_rate": 2.985157550256601e-05, + "loss": 0.0634, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013848463166505098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006358111975714564, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03405503854155541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9790192186832428, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11110, + "weighted_flow_matching_loss": 0.030018402822315694, + "weighted_rkd_loss": 0.03334052134305239 + }, + { + "epoch": 3.235379691591504, + "flow_matching_loss": 0.029336138069629668, + "grad_norm": 0.3256610929965973, + "learning_rate": 2.985099486724703e-05, + "loss": 0.0667, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015477767400443553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007182365003973246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038137900084257124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9789817035198212, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11120, + "weighted_flow_matching_loss": 0.029336138069629668, + "weighted_rkd_loss": 0.03733629416674376 + }, + { + "epoch": 3.2382892057026478, + "flow_matching_loss": 0.029422437213361262, + "grad_norm": 0.318444162607193, + "learning_rate": 2.9850413104092997e-05, + "loss": 0.0613, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013264568150043487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0060240069404244425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032553143613040444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9789441049098968, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11130, + "weighted_flow_matching_loss": 0.029422437213361262, + "weighted_rkd_loss": 0.03186770249158144 + }, + { + "epoch": 3.241198719813791, + "flow_matching_loss": 0.029417026601731778, + "grad_norm": 0.35189583897590637, + "learning_rate": 2.984983021314809e-05, + "loss": 0.0652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014767382014542817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007061297772452235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03659606128931046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.978906512260437, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11140, + "weighted_flow_matching_loss": 0.029417026601731778, + "weighted_rkd_loss": 0.03582411035895348 + }, + { + "epoch": 3.2441082339249343, + "flow_matching_loss": 0.0306423993781209, + "grad_norm": 0.28558921813964844, + "learning_rate": 2.984924619445658e-05, + "loss": 0.0658, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01456224862486124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0068235877901315686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035948085226118566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9788688719272614, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11150, + "weighted_flow_matching_loss": 0.0306423993781209, + "weighted_rkd_loss": 0.03518848530948162 + }, + { + "epoch": 3.247017748036078, + "flow_matching_loss": 0.029953986406326294, + "grad_norm": 0.34788715839385986, + "learning_rate": 2.984866104806282e-05, + "loss": 0.0641, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014068337250500917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006787895644083619, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03492457009851933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9788312017917633, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11160, + "weighted_flow_matching_loss": 0.029953986406326294, + "weighted_rkd_loss": 0.03418526370078325 + }, + { + "epoch": 3.2499272621472213, + "flow_matching_loss": 0.0312239408493042, + "grad_norm": 0.3651556670665741, + "learning_rate": 2.984807477401125e-05, + "loss": 0.0681, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015210671350359917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007248155307024717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03766949772834778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9787934958934784, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11170, + "weighted_flow_matching_loss": 0.0312239408493042, + "weighted_rkd_loss": 0.036870636977255344 + }, + { + "epoch": 3.252836776258365, + "flow_matching_loss": 0.032831141352653505, + "grad_norm": 0.34153246879577637, + "learning_rate": 2.9847487372346383e-05, + "loss": 0.0684, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014753119647502899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006846533715724945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03635277319699526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9787557721138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11180, + "weighted_flow_matching_loss": 0.032831141352653505, + "weighted_rkd_loss": 0.03558049220591784 + }, + { + "epoch": 3.2557462903695082, + "flow_matching_loss": 0.03137747552245855, + "grad_norm": 0.3455811142921448, + "learning_rate": 2.9846898843112833e-05, + "loss": 0.0651, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013991023786365986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006507916375994682, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03448996432125569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9787179946899414, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11190, + "weighted_flow_matching_loss": 0.03137747552245855, + "weighted_rkd_loss": 0.03375593367964029 + }, + { + "epoch": 3.258655804480652, + "flow_matching_loss": 0.029870156571269034, + "grad_norm": 0.35150253772735596, + "learning_rate": 2.9846309186355302e-05, + "loss": 0.0633, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013907817099243403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006301867682486773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03411750234663487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9786801993846893, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11200, + "weighted_flow_matching_loss": 0.029870156571269034, + "weighted_rkd_loss": 0.03339012861251831 + }, + { + "epoch": 3.261565318591795, + "flow_matching_loss": 0.027816610410809517, + "grad_norm": 0.3455900251865387, + "learning_rate": 2.9845718402118565e-05, + "loss": 0.0617, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014029142912477255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006589936558157206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034648222476243974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9786423742771149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11210, + "weighted_flow_matching_loss": 0.027816610410809517, + "weighted_rkd_loss": 0.03390819244086742 + }, + { + "epoch": 3.2644748327029385, + "flow_matching_loss": 0.028359479457139968, + "grad_norm": 0.4003836214542389, + "learning_rate": 2.9845126490447487e-05, + "loss": 0.0634, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014458015747368335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006895190803334117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0358112221583724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9786044955253601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11220, + "weighted_flow_matching_loss": 0.028359479457139968, + "weighted_rkd_loss": 0.035045025683939454 + }, + { + "epoch": 3.267384346814082, + "flow_matching_loss": 0.029716773703694344, + "grad_norm": 0.4026486873626709, + "learning_rate": 2.984453345138702e-05, + "loss": 0.0637, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014151626266539097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006428058864548803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03473131153732538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9785665988922119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11230, + "weighted_flow_matching_loss": 0.029716773703694344, + "weighted_rkd_loss": 0.03398689944297075 + }, + { + "epoch": 3.2702938609252254, + "flow_matching_loss": 0.030270567908883095, + "grad_norm": 0.3124600350856781, + "learning_rate": 2.9843939284982203e-05, + "loss": 0.0614, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012926704809069633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0059326667804270984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03178607616573572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9785286664962769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11240, + "weighted_flow_matching_loss": 0.030270567908883095, + "weighted_rkd_loss": 0.03110357541590929 + }, + { + "epoch": 3.2732033750363687, + "flow_matching_loss": 0.02870887387543917, + "grad_norm": 0.305559366941452, + "learning_rate": 2.9843343991278162e-05, + "loss": 0.0626, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013981370069086552, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006691061891615391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03465380165725947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9784907102584839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11250, + "weighted_flow_matching_loss": 0.02870887387543917, + "weighted_rkd_loss": 0.03390843216329813 + }, + { + "epoch": 3.2761128891475124, + "flow_matching_loss": 0.030518930032849313, + "grad_norm": 0.3206128478050232, + "learning_rate": 2.9842747570320104e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01443966617807746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006808989681303501, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03568832166492939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9784527063369751, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11260, + "weighted_flow_matching_loss": 0.030518930032849313, + "weighted_rkd_loss": 0.034919347055256365 + }, + { + "epoch": 3.2790224032586557, + "flow_matching_loss": 0.03051274921745062, + "grad_norm": 0.3373974859714508, + "learning_rate": 2.9842150022153317e-05, + "loss": 0.064, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013922129198908806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006401397474110127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03424565568566322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9784146785736084, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11270, + "weighted_flow_matching_loss": 0.03051274921745062, + "weighted_rkd_loss": 0.03350643217563629 + }, + { + "epoch": 3.2819319173697994, + "flow_matching_loss": 0.031360695511102675, + "grad_norm": 0.3699333071708679, + "learning_rate": 2.984155134682319e-05, + "loss": 0.0666, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014566592779010535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006862328574061394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035995513945817945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9783766150474549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11280, + "weighted_flow_matching_loss": 0.031360695511102675, + "weighted_rkd_loss": 0.0352171627804637 + }, + { + "epoch": 3.2848414314809427, + "flow_matching_loss": 0.02843600194901228, + "grad_norm": 0.27513208985328674, + "learning_rate": 2.984095154437518e-05, + "loss": 0.0601, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0131390861235559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0060763642191886905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03235453646630049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9783385276794434, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11290, + "weighted_flow_matching_loss": 0.02843600194901228, + "weighted_rkd_loss": 0.03165369424968958 + }, + { + "epoch": 3.2877509455920864, + "flow_matching_loss": 0.03255603946745396, + "grad_norm": 0.3459114134311676, + "learning_rate": 2.984035061485484e-05, + "loss": 0.0698, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015209104120731353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007634079875424504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03805228769779205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9783003926277161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11300, + "weighted_flow_matching_loss": 0.03255603946745396, + "weighted_rkd_loss": 0.03722657822072506 + }, + { + "epoch": 3.2906604597032296, + "flow_matching_loss": 0.030050118826329707, + "grad_norm": 0.2817586362361908, + "learning_rate": 2.983974855830781e-05, + "loss": 0.0627, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013486351165920497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063623626250773665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033335065096616746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9782622396945954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11310, + "weighted_flow_matching_loss": 0.030050118826329707, + "weighted_rkd_loss": 0.032610419951379296 + }, + { + "epoch": 3.293569973814373, + "flow_matching_loss": 0.032556737586855886, + "grad_norm": 0.36551374197006226, + "learning_rate": 2.9839145374779817e-05, + "loss": 0.0696, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015153542533516884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00756526542827487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03787235058844089, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9782240331172943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11320, + "weighted_flow_matching_loss": 0.032556737586855886, + "weighted_rkd_loss": 0.03704761527478695 + }, + { + "epoch": 3.2964794879255166, + "flow_matching_loss": 0.029236948862671853, + "grad_norm": 0.3448335826396942, + "learning_rate": 2.9838541064316656e-05, + "loss": 0.0639, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014336737804114819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006770560657605529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0354440363124013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9781858086585998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11330, + "weighted_flow_matching_loss": 0.029236948862671853, + "weighted_rkd_loss": 0.034670878760516645 + }, + { + "epoch": 3.29938900203666, + "flow_matching_loss": 0.029668707214295865, + "grad_norm": 0.2770213782787323, + "learning_rate": 2.983793562696423e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01337408097460866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006247126404196024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032995288446545604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9781475484371185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11340, + "weighted_flow_matching_loss": 0.029668707214295865, + "weighted_rkd_loss": 0.032274249009788036 + }, + { + "epoch": 3.302298516147803, + "flow_matching_loss": 0.02963013481348753, + "grad_norm": 0.3137146532535553, + "learning_rate": 2.983732906276851e-05, + "loss": 0.0643, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014403169602155685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00661326521076262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035419604554772374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9781092524528503, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11350, + "weighted_flow_matching_loss": 0.02963013481348753, + "weighted_rkd_loss": 0.03464429061859846 + }, + { + "epoch": 3.305208030258947, + "flow_matching_loss": 0.030497502721846103, + "grad_norm": 0.33877629041671753, + "learning_rate": 2.9836721371775574e-05, + "loss": 0.0667, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014775640238076449, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007415359048172832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03696663938462734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9780709266662597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11360, + "weighted_flow_matching_loss": 0.030497502721846103, + "weighted_rkd_loss": 0.036155995726585385 + }, + { + "epoch": 3.30811754437009, + "flow_matching_loss": 0.029046297632157804, + "grad_norm": 0.3205300569534302, + "learning_rate": 2.983611255403156e-05, + "loss": 0.0601, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012901610136032105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005907174153253436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03171039428561926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9780325710773468, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11370, + "weighted_flow_matching_loss": 0.029046297632157804, + "weighted_rkd_loss": 0.031013792753219603 + }, + { + "epoch": 3.311027058481234, + "flow_matching_loss": 0.027134756930172442, + "grad_norm": 0.3052000105381012, + "learning_rate": 2.9835502609582705e-05, + "loss": 0.059, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013212658371776343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006195204611867666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032620521262288094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9779941856861114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11380, + "weighted_flow_matching_loss": 0.027134756930172442, + "weighted_rkd_loss": 0.03190268538892269 + }, + { + "epoch": 3.313936572592377, + "flow_matching_loss": 0.02656053863465786, + "grad_norm": 0.40611863136291504, + "learning_rate": 2.9834891538475335e-05, + "loss": 0.061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014253298472613096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006664017634466291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03517061490565539, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9779557585716248, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11390, + "weighted_flow_matching_loss": 0.02656053863465786, + "weighted_rkd_loss": 0.034395300410687925 + }, + { + "epoch": 3.3168460867035203, + "flow_matching_loss": 0.02880102451890707, + "grad_norm": 0.30451294779777527, + "learning_rate": 2.983427934075585e-05, + "loss": 0.0613, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01342371916398406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006360246799886227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033207685500383374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9779173016548157, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11400, + "weighted_flow_matching_loss": 0.02880102451890707, + "weighted_rkd_loss": 0.03247438501566648 + }, + { + "epoch": 3.319755600814664, + "flow_matching_loss": 0.030367729999125005, + "grad_norm": 0.38984543085098267, + "learning_rate": 2.9833666016470755e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014433549437671899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006954055372625589, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035821153596043584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9778788149356842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11410, + "weighted_flow_matching_loss": 0.030367729999125005, + "weighted_rkd_loss": 0.035028741136193274 + }, + { + "epoch": 3.3226651149258073, + "flow_matching_loss": 0.028647232800722122, + "grad_norm": 0.29255223274230957, + "learning_rate": 2.9833051565666613e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014017885830253363, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006383377686142921, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03441914934664965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9778402984142304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11420, + "weighted_flow_matching_loss": 0.028647232800722122, + "weighted_rkd_loss": 0.03365643545985222 + }, + { + "epoch": 3.325574629036951, + "flow_matching_loss": 0.03022413104772568, + "grad_norm": 0.36990904808044434, + "learning_rate": 2.9832435988390096e-05, + "loss": 0.0638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013963240664452314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006364486599341035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03429096713662148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9778017401695251, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11430, + "weighted_flow_matching_loss": 0.03022413104772568, + "weighted_rkd_loss": 0.03352977279573679 + }, + { + "epoch": 3.3284841431480943, + "flow_matching_loss": 0.02993905320763588, + "grad_norm": 0.2965964376926422, + "learning_rate": 2.983181928468795e-05, + "loss": 0.0661, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014977121632546187, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070235507097095255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03697779327630997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9777631521224975, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11440, + "weighted_flow_matching_loss": 0.02993905320763588, + "weighted_rkd_loss": 0.03615553472191095 + }, + { + "epoch": 3.3313936572592375, + "flow_matching_loss": 0.02993252296000719, + "grad_norm": 0.32510700821876526, + "learning_rate": 2.9831201454607017e-05, + "loss": 0.0611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013035767432302236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005857109604403376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031928645074367525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9777245342731475, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11450, + "weighted_flow_matching_loss": 0.02993252296000719, + "weighted_rkd_loss": 0.031217430159449576 + }, + { + "epoch": 3.3343031713703812, + "flow_matching_loss": 0.029829706624150278, + "grad_norm": 0.3793618381023407, + "learning_rate": 2.9830582498194204e-05, + "loss": 0.0598, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012422923650592565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00582831222563982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03067415952682495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9776858806610107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11460, + "weighted_flow_matching_loss": 0.029829706624150278, + "weighted_rkd_loss": 0.029989675246179105 + }, + { + "epoch": 3.3372126854815245, + "flow_matching_loss": 0.029286951571702958, + "grad_norm": 0.34367895126342773, + "learning_rate": 2.9829962415496522e-05, + "loss": 0.0616, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013279584143310786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006519877957180143, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03307904601097107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.977647203207016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11470, + "weighted_flow_matching_loss": 0.029286951571702958, + "weighted_rkd_loss": 0.0323396360501647 + }, + { + "epoch": 3.340122199592668, + "flow_matching_loss": 0.029438757337629795, + "grad_norm": 0.33347445726394653, + "learning_rate": 2.982934120656107e-05, + "loss": 0.0631, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014013886824250222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006354068033397197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03438184186816216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9776084721088409, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11480, + "weighted_flow_matching_loss": 0.029438757337629795, + "weighted_rkd_loss": 0.03361197132617235 + }, + { + "epoch": 3.3430317137038115, + "flow_matching_loss": 0.030818380974233152, + "grad_norm": 0.35323795676231384, + "learning_rate": 2.9828718871435018e-05, + "loss": 0.0635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013600124418735504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006208296772092581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03340854551643133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9775697290897369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11490, + "weighted_flow_matching_loss": 0.030818380974233152, + "weighted_rkd_loss": 0.032659162767231464 + }, + { + "epoch": 3.3459412278149547, + "flow_matching_loss": 0.028849440813064575, + "grad_norm": 0.3550246059894562, + "learning_rate": 2.9828095410165628e-05, + "loss": 0.0599, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012798606976866722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006152416113764048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031749630346894264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9775309443473816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11500, + "weighted_flow_matching_loss": 0.028849440813064575, + "weighted_rkd_loss": 0.031036254204809665 + }, + { + "epoch": 3.3488507419260984, + "flow_matching_loss": 0.029620197042822836, + "grad_norm": 0.3423924446105957, + "learning_rate": 2.9827470822800246e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01483826171606779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006943904468789697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036620428785681725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9774921298027038, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11510, + "weighted_flow_matching_loss": 0.029620197042822836, + "weighted_rkd_loss": 0.035796187445521356 + }, + { + "epoch": 3.3517602560372417, + "flow_matching_loss": 0.031370605900883676, + "grad_norm": 0.31723925471305847, + "learning_rate": 2.982684510938631e-05, + "loss": 0.0642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013653564546257258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006248266994953155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03355539608746767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9774532794952393, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11520, + "weighted_flow_matching_loss": 0.031370605900883676, + "weighted_rkd_loss": 0.03279882576316595 + }, + { + "epoch": 3.3546697701483854, + "flow_matching_loss": 0.030946982465684413, + "grad_norm": 0.30764999985694885, + "learning_rate": 2.9826218269971337e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012794023379683495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00603554705157876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031623593717813495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9774143993854523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11530, + "weighted_flow_matching_loss": 0.030946982465684413, + "weighted_rkd_loss": 0.030909379571676256 + }, + { + "epoch": 3.3575792842595287, + "flow_matching_loss": 0.029939335957169532, + "grad_norm": 0.31493180990219116, + "learning_rate": 2.9825590304602925e-05, + "loss": 0.0664, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015010175295174122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007335329661145806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037355680018663406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9773754835128784, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11540, + "weighted_flow_matching_loss": 0.029939335957169532, + "weighted_rkd_loss": 0.036510545015335086 + }, + { + "epoch": 3.360488798370672, + "flow_matching_loss": 0.031565909087657926, + "grad_norm": 0.3345688581466675, + "learning_rate": 2.9824961213328778e-05, + "loss": 0.067, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014765064232051373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006675178837031126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036205307394266126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9773365318775177, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11550, + "weighted_flow_matching_loss": 0.031565909087657926, + "weighted_rkd_loss": 0.03538476936519146 + }, + { + "epoch": 3.3633983124818156, + "flow_matching_loss": 0.029182226210832597, + "grad_norm": 0.30201640725135803, + "learning_rate": 2.982433099619666e-05, + "loss": 0.0637, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014232558012008668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00682203583419323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03528715185821056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.977297556400299, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11560, + "weighted_flow_matching_loss": 0.029182226210832597, + "weighted_rkd_loss": 0.03448604177683592 + }, + { + "epoch": 3.366307826592959, + "flow_matching_loss": 0.028103004954755307, + "grad_norm": 0.34516656398773193, + "learning_rate": 2.9823699653254433e-05, + "loss": 0.0597, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012995837908238172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063483235891908405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03233999963849783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9772585332393646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11570, + "weighted_flow_matching_loss": 0.028103004954755307, + "weighted_rkd_loss": 0.03160451333969831 + }, + { + "epoch": 3.369217340704102, + "flow_matching_loss": 0.02861566636711359, + "grad_norm": 0.3181757628917694, + "learning_rate": 2.9823067184550047e-05, + "loss": 0.0617, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013747017085552215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006364509044215083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033858542889356615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9772194921970367, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11580, + "weighted_flow_matching_loss": 0.02861566636711359, + "weighted_rkd_loss": 0.03308723047375679 + }, + { + "epoch": 3.372126854815246, + "flow_matching_loss": 0.029204029776155948, + "grad_norm": 0.40019720792770386, + "learning_rate": 2.9822433590131532e-05, + "loss": 0.061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01314270794391632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006276733987033367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03256215006113052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9771804213523865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11590, + "weighted_flow_matching_loss": 0.029204029776155948, + "weighted_rkd_loss": 0.03181908205151558 + }, + { + "epoch": 3.375036368926389, + "flow_matching_loss": 0.029521146044135093, + "grad_norm": 0.30137184262275696, + "learning_rate": 2.9821798870047006e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011813634168356657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005535933980718255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029163202084600926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9771413087844849, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11600, + "weighted_flow_matching_loss": 0.029521146044135093, + "weighted_rkd_loss": 0.02849654443562031 + }, + { + "epoch": 3.377945883037533, + "flow_matching_loss": 0.029287055134773254, + "grad_norm": 0.3993358314037323, + "learning_rate": 2.982116302434467e-05, + "loss": 0.0595, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012510671000927686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005857625883072615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0308789674192667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9771021664142608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11610, + "weighted_flow_matching_loss": 0.029287055134773254, + "weighted_rkd_loss": 0.03017190806567669 + }, + { + "epoch": 3.380855397148676, + "flow_matching_loss": 0.029432309418916704, + "grad_norm": 0.31244897842407227, + "learning_rate": 2.9820526053072812e-05, + "loss": 0.0609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01314379069954157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005912612192332745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03220019415020943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.97706298828125, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11620, + "weighted_flow_matching_loss": 0.029432309418916704, + "weighted_rkd_loss": 0.03146161455661058 + }, + { + "epoch": 3.38376491125982, + "flow_matching_loss": 0.028537345491349696, + "grad_norm": 0.2836739420890808, + "learning_rate": 2.981988795627981e-05, + "loss": 0.061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013499346189200878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062047820538282394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03320347461849451, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9770237803459167, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11630, + "weighted_flow_matching_loss": 0.028537345491349696, + "weighted_rkd_loss": 0.032440570183098316 + }, + { + "epoch": 3.386674425370963, + "flow_matching_loss": 0.031021371483802795, + "grad_norm": 0.31137219071388245, + "learning_rate": 2.981924873401412e-05, + "loss": 0.0624, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013019253965467215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00609781569801271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032136323675513265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9769845366477966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11640, + "weighted_flow_matching_loss": 0.031021371483802795, + "weighted_rkd_loss": 0.031396702863276005 + }, + { + "epoch": 3.3895839394821063, + "flow_matching_loss": 0.029199792072176933, + "grad_norm": 0.3254116475582123, + "learning_rate": 2.9818608386324285e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013491436187177897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006486182101070881, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033469054475426675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9769452691078186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11650, + "weighted_flow_matching_loss": 0.029199792072176933, + "weighted_rkd_loss": 0.03269743323326111 + }, + { + "epoch": 3.39249345359325, + "flow_matching_loss": 0.03097305353730917, + "grad_norm": 0.34195923805236816, + "learning_rate": 2.9817966913258942e-05, + "loss": 0.0638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013577145617455244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006442968873307109, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033597260154783724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9769059419631958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11660, + "weighted_flow_matching_loss": 0.03097305353730917, + "weighted_rkd_loss": 0.0328213632106781 + }, + { + "epoch": 3.3954029677043933, + "flow_matching_loss": 0.02891050633043051, + "grad_norm": 0.3165091574192047, + "learning_rate": 2.9817324314866797e-05, + "loss": 0.06, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01302071288228035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005782780237495899, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03182420581579208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.976866602897644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11670, + "weighted_flow_matching_loss": 0.02891050633043051, + "weighted_rkd_loss": 0.0310880308970809 + }, + { + "epoch": 3.3983124818155366, + "flow_matching_loss": 0.030678798072040082, + "grad_norm": 0.3579328954219818, + "learning_rate": 2.9816680591196658e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013276054989546537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006057649943977594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03260975982993841, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9768272519111634, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11680, + "weighted_flow_matching_loss": 0.030678798072040082, + "weighted_rkd_loss": 0.03185406737029552 + }, + { + "epoch": 3.4012219959266803, + "flow_matching_loss": 0.029931216314435006, + "grad_norm": 0.3466469645500183, + "learning_rate": 2.981603574229741e-05, + "loss": 0.0624, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013616527803242207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005995654547587037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03322870973497629, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9767878413200378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11690, + "weighted_flow_matching_loss": 0.029931216314435006, + "weighted_rkd_loss": 0.032457393035292624 + }, + { + "epoch": 3.4041315100378235, + "flow_matching_loss": 0.031092222034931182, + "grad_norm": 0.28228598833084106, + "learning_rate": 2.9815389768218025e-05, + "loss": 0.0621, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012888174317777158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005927112558856607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03170346189290285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.97674840092659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11700, + "weighted_flow_matching_loss": 0.031092222034931182, + "weighted_rkd_loss": 0.030966293625533582 + }, + { + "epoch": 3.4070410241489673, + "flow_matching_loss": 0.029306862130761148, + "grad_norm": 0.3494833707809448, + "learning_rate": 2.981474266900756e-05, + "loss": 0.0609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013078314810991287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006160019664093852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03231664914637804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9767089366912842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11710, + "weighted_flow_matching_loss": 0.029306862130761148, + "weighted_rkd_loss": 0.03156394548714161 + }, + { + "epoch": 3.4099505382601105, + "flow_matching_loss": 0.029940514452755453, + "grad_norm": 0.3116435706615448, + "learning_rate": 2.9814094444715157e-05, + "loss": 0.0606, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012867999449372291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005617048498243094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03135304749011993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.976669430732727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11720, + "weighted_flow_matching_loss": 0.029940514452755453, + "weighted_rkd_loss": 0.030621558800339697 + }, + { + "epoch": 3.4128600523712542, + "flow_matching_loss": 0.029579698480665682, + "grad_norm": 0.4040355384349823, + "learning_rate": 2.9813445095390046e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013600205816328526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0065527093596756455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033753120340406895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9766298949718475, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11730, + "weighted_flow_matching_loss": 0.029579698480665682, + "weighted_rkd_loss": 0.032964270748198035 + }, + { + "epoch": 3.4157695664823975, + "flow_matching_loss": 0.03156264573335647, + "grad_norm": 0.3621487021446228, + "learning_rate": 2.9812794621081542e-05, + "loss": 0.0643, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013681246526539325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062024268787354234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033564919419586656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9765903294086457, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11740, + "weighted_flow_matching_loss": 0.03156264573335647, + "weighted_rkd_loss": 0.03277919925749302 + }, + { + "epoch": 3.4186790805935408, + "flow_matching_loss": 0.029271511361002922, + "grad_norm": 0.3750530779361725, + "learning_rate": 2.981214302183904e-05, + "loss": 0.0627, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013980771880596876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006288214353844523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03424975834786892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9765507280826569, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11750, + "weighted_flow_matching_loss": 0.029271511361002922, + "weighted_rkd_loss": 0.033446648716926576 + }, + { + "epoch": 3.4215885947046845, + "flow_matching_loss": 0.0288599094375968, + "grad_norm": 0.3206844627857208, + "learning_rate": 2.981149029771203e-05, + "loss": 0.0634, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014423989690840244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006475302763283253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035323281586170194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9765110909938812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11760, + "weighted_flow_matching_loss": 0.0288599094375968, + "weighted_rkd_loss": 0.03449360132217407 + }, + { + "epoch": 3.4244981088158277, + "flow_matching_loss": 0.02918738014996052, + "grad_norm": 0.3286077082157135, + "learning_rate": 2.9810836448750074e-05, + "loss": 0.0616, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013642008882015944, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005886084353551269, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03317010290920734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9764714241027832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11770, + "weighted_flow_matching_loss": 0.02918738014996052, + "weighted_rkd_loss": 0.03238964844495058 + }, + { + "epoch": 3.427407622926971, + "flow_matching_loss": 0.028809150867164134, + "grad_norm": 0.3405209481716156, + "learning_rate": 2.9810181475002835e-05, + "loss": 0.0609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013313771784305572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006227953778579831, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03285549767315388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9764317274093628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11780, + "weighted_flow_matching_loss": 0.028809150867164134, + "weighted_rkd_loss": 0.03208115566521883 + }, + { + "epoch": 3.4303171370381147, + "flow_matching_loss": 0.0293285658583045, + "grad_norm": 0.40901392698287964, + "learning_rate": 2.9809525376520058e-05, + "loss": 0.0584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01207842268049717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005635326029732823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029792171716690064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9763919889926911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11790, + "weighted_flow_matching_loss": 0.0293285658583045, + "weighted_rkd_loss": 0.029088815674185753 + }, + { + "epoch": 3.433226651149258, + "flow_matching_loss": 0.033095458149909975, + "grad_norm": 0.26877647638320923, + "learning_rate": 2.9808868153351555e-05, + "loss": 0.0653, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013409404642879963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006133313244208694, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03295212239027023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9763522326946259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11800, + "weighted_flow_matching_loss": 0.033095458149909975, + "weighted_rkd_loss": 0.03217291422188282 + }, + { + "epoch": 3.4361361652604017, + "flow_matching_loss": 0.02763813007622957, + "grad_norm": 0.3334280252456665, + "learning_rate": 2.980820980554725e-05, + "loss": 0.0584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0128324287943542, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00588358654640615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0315484443679452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9763124346733093, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11810, + "weighted_flow_matching_loss": 0.02763813007622957, + "weighted_rkd_loss": 0.03080115243792534 + }, + { + "epoch": 3.439045679371545, + "flow_matching_loss": 0.027799669466912747, + "grad_norm": 0.335406094789505, + "learning_rate": 2.9807550333157136e-05, + "loss": 0.06, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013418263103812934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006159760523587465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032996286451816556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9762726068496704, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11820, + "weighted_flow_matching_loss": 0.027799669466912747, + "weighted_rkd_loss": 0.0322133619338274 + }, + { + "epoch": 3.4419551934826886, + "flow_matching_loss": 0.028908371180295944, + "grad_norm": 0.3539854884147644, + "learning_rate": 2.9806889736231294e-05, + "loss": 0.0621, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013790508732199669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006402977090328932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033983995020389554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9762327432632446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11830, + "weighted_flow_matching_loss": 0.028908371180295944, + "weighted_rkd_loss": 0.03317627627402544 + }, + { + "epoch": 3.444864707593832, + "flow_matching_loss": 0.029630857706069946, + "grad_norm": 0.35846519470214844, + "learning_rate": 2.980622801481989e-05, + "loss": 0.0612, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013156421389430762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006019913358613849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032332755997776984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9761928617954254, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11840, + "weighted_flow_matching_loss": 0.029630857706069946, + "weighted_rkd_loss": 0.031562989391386506 + }, + { + "epoch": 3.447774221704975, + "flow_matching_loss": 0.029504137858748437, + "grad_norm": 0.35813620686531067, + "learning_rate": 2.9805565168973182e-05, + "loss": 0.0607, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012829085625708104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006334433751180768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03199260514229536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9761529266834259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11850, + "weighted_flow_matching_loss": 0.029504137858748437, + "weighted_rkd_loss": 0.03122968152165413 + }, + { + "epoch": 3.450683735816119, + "flow_matching_loss": 0.031148447282612325, + "grad_norm": 0.4156799614429474, + "learning_rate": 2.9804901198741515e-05, + "loss": 0.0632, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013297307398170232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006262357905507087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0328569732606411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.976112961769104, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11860, + "weighted_flow_matching_loss": 0.031148447282612325, + "weighted_rkd_loss": 0.03207213226705789 + }, + { + "epoch": 3.453593249927262, + "flow_matching_loss": 0.027522867918014525, + "grad_norm": 0.34004124999046326, + "learning_rate": 2.9804236104175298e-05, + "loss": 0.0585, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012912956718355417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005922395177185536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031748308800160885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9760729670524597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11870, + "weighted_flow_matching_loss": 0.027522867918014525, + "weighted_rkd_loss": 0.03098864760249853 + }, + { + "epoch": 3.4565027640384054, + "flow_matching_loss": 0.030805515125393867, + "grad_norm": 0.32583901286125183, + "learning_rate": 2.9803569885325053e-05, + "loss": 0.0638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013854818884283304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006078445538878441, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03378808330744505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.976032942533493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11880, + "weighted_flow_matching_loss": 0.030805515125393867, + "weighted_rkd_loss": 0.03297830745577812 + }, + { + "epoch": 3.459412278149549, + "flow_matching_loss": 0.028833813033998013, + "grad_norm": 0.3732094168663025, + "learning_rate": 2.9802902542241366e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014035477489233016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006434096489101648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03450505137443542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.975992888212204, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11890, + "weighted_flow_matching_loss": 0.028833813033998013, + "weighted_rkd_loss": 0.03367670979350805 + }, + { + "epoch": 3.4623217922606924, + "flow_matching_loss": 0.029629402421414854, + "grad_norm": 0.319463849067688, + "learning_rate": 2.9802234074974925e-05, + "loss": 0.0611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013082757219672203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006060409359633923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03222592379897833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9759527981281281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11900, + "weighted_flow_matching_loss": 0.029629402421414854, + "weighted_rkd_loss": 0.03145097475498915 + }, + { + "epoch": 3.465231306371836, + "flow_matching_loss": 0.02790642883628607, + "grad_norm": 0.35851380228996277, + "learning_rate": 2.980156448357649e-05, + "loss": 0.0609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013728508539497853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006377649400383234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03383466601371765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9759126722812652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11910, + "weighted_flow_matching_loss": 0.02790642883628607, + "weighted_rkd_loss": 0.033019672706723215 + }, + { + "epoch": 3.4681408204829793, + "flow_matching_loss": 0.030098992213606835, + "grad_norm": 0.3579767346382141, + "learning_rate": 2.9800893768096916e-05, + "loss": 0.0635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01379102524369955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006647992227226496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034230043180286884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9758725106716156, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11920, + "weighted_flow_matching_loss": 0.030098992213606835, + "weighted_rkd_loss": 0.03340413589030504 + }, + { + "epoch": 3.4710503345941226, + "flow_matching_loss": 0.027521947398781776, + "grad_norm": 0.3430110216140747, + "learning_rate": 2.9800221928587137e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01224647080525756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005604931572452188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03009787332266569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.975832325220108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11930, + "weighted_flow_matching_loss": 0.027521947398781776, + "weighted_rkd_loss": 0.02937045879662037 + }, + { + "epoch": 3.4739598487052663, + "flow_matching_loss": 0.030178902298212053, + "grad_norm": 0.30781418085098267, + "learning_rate": 2.9799548965098176e-05, + "loss": 0.063, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013737802207469941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006167460046708584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033643064089119434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9757921099662781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11940, + "weighted_flow_matching_loss": 0.030178902298212053, + "weighted_rkd_loss": 0.03282864857465029 + }, + { + "epoch": 3.4768693628164096, + "flow_matching_loss": 0.03352406285703182, + "grad_norm": 0.27620309591293335, + "learning_rate": 2.979887487768114e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012101037800312043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005321059678681195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02952313534915447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9757518589496612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11950, + "weighted_flow_matching_loss": 0.03352406285703182, + "weighted_rkd_loss": 0.02880727853626013 + }, + { + "epoch": 3.4797788769275533, + "flow_matching_loss": 0.02772908452898264, + "grad_norm": 0.3255995512008667, + "learning_rate": 2.979819966638722e-05, + "loss": 0.0588, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012962378934025765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005939149716868997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0318639075383544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.975711566209793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11960, + "weighted_flow_matching_loss": 0.02772908452898264, + "weighted_rkd_loss": 0.031089985743165016 + }, + { + "epoch": 3.4826883910386965, + "flow_matching_loss": 0.030214300379157066, + "grad_norm": 0.3298545181751251, + "learning_rate": 2.9797523331267698e-05, + "loss": 0.0609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012709469255059957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006050852267071605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03146979063749313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9756712436676025, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11970, + "weighted_flow_matching_loss": 0.030214300379157066, + "weighted_rkd_loss": 0.030704188533127307 + }, + { + "epoch": 3.48559790514984, + "flow_matching_loss": 0.02790914811193943, + "grad_norm": 0.3029787242412567, + "learning_rate": 2.979684587237393e-05, + "loss": 0.0597, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01326332353055477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006042170664295554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032568817213177684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9756308913230896, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11980, + "weighted_flow_matching_loss": 0.02790914811193943, + "weighted_rkd_loss": 0.03177515901625157 + }, + { + "epoch": 3.4885074192609835, + "flow_matching_loss": 0.027347028814256193, + "grad_norm": 0.297061562538147, + "learning_rate": 2.979616728975737e-05, + "loss": 0.0575, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012434855476021767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00601408276706934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0308837940916419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9755905151367188, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11990, + "weighted_flow_matching_loss": 0.027347028814256193, + "weighted_rkd_loss": 0.03012994062155485 + }, + { + "epoch": 3.4914169333721268, + "flow_matching_loss": 0.02872586213052273, + "grad_norm": 0.32834264636039734, + "learning_rate": 2.979548758346955e-05, + "loss": 0.0629, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014196429774165154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006645545596256852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03503840509802103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9755500912666321, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12000, + "weighted_flow_matching_loss": 0.02872586213052273, + "weighted_rkd_loss": 0.03418172262609005 + }, + { + "epoch": 3.4943264474832705, + "flow_matching_loss": 0.028182243183255195, + "grad_norm": 0.3063358664512634, + "learning_rate": 2.979480675356209e-05, + "loss": 0.0574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012156055774539709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056083434727042915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029920454882085324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.975509649515152, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12010, + "weighted_flow_matching_loss": 0.028182243183255195, + "weighted_rkd_loss": 0.029187703132629396 + }, + { + "epoch": 3.4972359615944137, + "flow_matching_loss": 0.02987075727432966, + "grad_norm": 0.3138520121574402, + "learning_rate": 2.9794124800086694e-05, + "loss": 0.0632, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013803134579211474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006589712854474783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03419598210602999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.975469172000885, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12020, + "weighted_flow_matching_loss": 0.02987075727432966, + "weighted_rkd_loss": 0.03335711602121592 + }, + { + "epoch": 3.500145475705557, + "flow_matching_loss": 0.028605431132018567, + "grad_norm": 0.33674705028533936, + "learning_rate": 2.9793441723095155e-05, + "loss": 0.0583, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012365489918738603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005747217684984207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03047819770872593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9754286408424377, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12030, + "weighted_flow_matching_loss": 0.028605431132018567, + "weighted_rkd_loss": 0.029729314148426056 + }, + { + "epoch": 3.5030549898167007, + "flow_matching_loss": 0.02763616181910038, + "grad_norm": 0.30404940247535706, + "learning_rate": 2.9792757522639345e-05, + "loss": 0.0596, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013403989002108574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005976736359298229, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032784715108573434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9753881096839905, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12040, + "weighted_flow_matching_loss": 0.02763616181910038, + "weighted_rkd_loss": 0.03197783306241035 + }, + { + "epoch": 3.505964503927844, + "flow_matching_loss": 0.028456005081534386, + "grad_norm": 0.3608871102333069, + "learning_rate": 2.9792072198771222e-05, + "loss": 0.0592, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012827877979725599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0058483817148953674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03150413744151592, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9753475248813629, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12050, + "weighted_flow_matching_loss": 0.028456005081534386, + "weighted_rkd_loss": 0.0307274729013443 + }, + { + "epoch": 3.5088740180389877, + "flow_matching_loss": 0.028928382322192192, + "grad_norm": 0.2610683739185333, + "learning_rate": 2.9791385751542837e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013626439031213522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006572957197204232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033825835771858694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9753069043159485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12060, + "weighted_flow_matching_loss": 0.028928382322192192, + "weighted_rkd_loss": 0.03299060668796301 + }, + { + "epoch": 3.511783532150131, + "flow_matching_loss": 0.028959407657384872, + "grad_norm": 0.34352558851242065, + "learning_rate": 2.9790698181006316e-05, + "loss": 0.0608, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013369366060942411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005930566228926182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03266929872334003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9752662658691407, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12070, + "weighted_flow_matching_loss": 0.028959407657384872, + "weighted_rkd_loss": 0.031861234456300735 + }, + { + "epoch": 3.514693046261274, + "flow_matching_loss": 0.03095133788883686, + "grad_norm": 0.31962913274765015, + "learning_rate": 2.979000948721388e-05, + "loss": 0.0613, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01257075872272253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0060277913697063925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031169308722019194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9752255916595459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12080, + "weighted_flow_matching_loss": 0.03095133788883686, + "weighted_rkd_loss": 0.030397111363708974 + }, + { + "epoch": 3.517602560372418, + "flow_matching_loss": 0.02583827041089535, + "grad_norm": 0.3979906439781189, + "learning_rate": 2.978931967021783e-05, + "loss": 0.0568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012722358480095864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006300493981689215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03174521066248417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9751848816871643, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12090, + "weighted_flow_matching_loss": 0.02583827041089535, + "weighted_rkd_loss": 0.030957426503300668 + }, + { + "epoch": 3.520512074483561, + "flow_matching_loss": 0.029840926826000213, + "grad_norm": 0.3445011079311371, + "learning_rate": 2.978862873007055e-05, + "loss": 0.0616, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013260759599506854, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006011549197137356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03253306858241558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9751441299915313, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12100, + "weighted_flow_matching_loss": 0.029840926826000213, + "weighted_rkd_loss": 0.03172442801296711 + }, + { + "epoch": 3.5234215885947044, + "flow_matching_loss": 0.029316752403974532, + "grad_norm": 0.387866735458374, + "learning_rate": 2.9787936666824518e-05, + "loss": 0.0614, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013306576199829579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00628057811409235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03289373070001602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9751033663749695, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12110, + "weighted_flow_matching_loss": 0.029316752403974532, + "weighted_rkd_loss": 0.032074779458343985 + }, + { + "epoch": 3.526331102705848, + "flow_matching_loss": 0.030647489614784716, + "grad_norm": 0.3628469705581665, + "learning_rate": 2.9787243480532283e-05, + "loss": 0.0615, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0129341134801507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057869146578013895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03165514133870602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9750625491142273, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12120, + "weighted_flow_matching_loss": 0.030647489614784716, + "weighted_rkd_loss": 0.030865760147571565 + }, + { + "epoch": 3.5292406168169914, + "flow_matching_loss": 0.029143765941262244, + "grad_norm": 0.29099228978157043, + "learning_rate": 2.9786549171246495e-05, + "loss": 0.0615, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013546316791325808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006115758372470736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033208392187953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9750217139720917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12130, + "weighted_flow_matching_loss": 0.029143765941262244, + "weighted_rkd_loss": 0.03237890973687172 + }, + { + "epoch": 3.532150130928135, + "flow_matching_loss": 0.02939780745655298, + "grad_norm": 0.37975502014160156, + "learning_rate": 2.9785853739019882e-05, + "loss": 0.0604, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012910547852516174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006006885785609484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03182798121124506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9749808371067047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12140, + "weighted_flow_matching_loss": 0.02939780745655298, + "weighted_rkd_loss": 0.031031651608645915 + }, + { + "epoch": 3.5350596450392784, + "flow_matching_loss": 0.02924242541193962, + "grad_norm": 0.32204240560531616, + "learning_rate": 2.9785157183905256e-05, + "loss": 0.0588, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01237469371408224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005555964587256312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03030535243451595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9749399423599243, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12150, + "weighted_flow_matching_loss": 0.02924242541193962, + "weighted_rkd_loss": 0.029545890912413598 + }, + { + "epoch": 3.537969159150422, + "flow_matching_loss": 0.028003623895347118, + "grad_norm": 0.3260674774646759, + "learning_rate": 2.978445950595551e-05, + "loss": 0.0615, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013966958317905665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006396299926564097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03433021679520607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9748989999294281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12160, + "weighted_flow_matching_loss": 0.028003623895347118, + "weighted_rkd_loss": 0.033468523807823655 + }, + { + "epoch": 3.5408786732615654, + "flow_matching_loss": 0.03137469738721847, + "grad_norm": 0.335113525390625, + "learning_rate": 2.978376070522364e-05, + "loss": 0.0631, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01305168503895402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00646536978892982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03256873991340399, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9748580276966095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12170, + "weighted_flow_matching_loss": 0.03137469738721847, + "weighted_rkd_loss": 0.03174986783415079 + }, + { + "epoch": 3.5437881873727086, + "flow_matching_loss": 0.02774461582303047, + "grad_norm": 0.346869558095932, + "learning_rate": 2.9783060781762706e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013598618749529123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006241662707179785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03343890085816383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9748170256614686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12180, + "weighted_flow_matching_loss": 0.02774461582303047, + "weighted_rkd_loss": 0.03259681165218353 + }, + { + "epoch": 3.5466977014838523, + "flow_matching_loss": 0.031073698401451112, + "grad_norm": 0.3169814944267273, + "learning_rate": 2.9782359735625867e-05, + "loss": 0.0635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013649644423276186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006000437587499618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03329972643405199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9747759878635407, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12190, + "weighted_flow_matching_loss": 0.031073698401451112, + "weighted_rkd_loss": 0.03245977833867073 + }, + { + "epoch": 3.5496072155949956, + "flow_matching_loss": 0.02768893875181675, + "grad_norm": 0.3515279293060303, + "learning_rate": 2.9781657566866357e-05, + "loss": 0.0593, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013267457485198975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005847361404448748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032382276654243466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9747349262237549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12200, + "weighted_flow_matching_loss": 0.02768893875181675, + "weighted_rkd_loss": 0.03156410716474056 + }, + { + "epoch": 3.552516729706139, + "flow_matching_loss": 0.028735652193427087, + "grad_norm": 0.34411996603012085, + "learning_rate": 2.978095427553751e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01392271313816309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061544352676719425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03399986177682877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9746938049793243, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12210, + "weighted_flow_matching_loss": 0.028735652193427087, + "weighted_rkd_loss": 0.033139442652463914 + }, + { + "epoch": 3.5554262438172826, + "flow_matching_loss": 0.029663531668484212, + "grad_norm": 0.36926037073135376, + "learning_rate": 2.9780249861692727e-05, + "loss": 0.0597, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012538807466626167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005772887961938977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030850503034889697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9746526777744293, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12220, + "weighted_flow_matching_loss": 0.029663531668484212, + "weighted_rkd_loss": 0.030068526230752467 + }, + { + "epoch": 3.558335757928426, + "flow_matching_loss": 0.028358605317771436, + "grad_norm": 0.30119213461875916, + "learning_rate": 2.977954432538551e-05, + "loss": 0.059, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012827117461711168, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005755790276452899, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03141002468764782, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9746115207672119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12230, + "weighted_flow_matching_loss": 0.028358605317771436, + "weighted_rkd_loss": 0.030612568743526934 + }, + { + "epoch": 3.5612452720395695, + "flow_matching_loss": 0.027771102078258992, + "grad_norm": 0.3420972526073456, + "learning_rate": 2.977883766666944e-05, + "loss": 0.0578, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012507917452603578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005784614314325154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030800448916852474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9745703279972077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12240, + "weighted_flow_matching_loss": 0.027771102078258992, + "weighted_rkd_loss": 0.03001718260347843 + }, + { + "epoch": 3.564154786150713, + "flow_matching_loss": 0.028364061377942562, + "grad_norm": 0.30061107873916626, + "learning_rate": 2.977812988559818e-05, + "loss": 0.0565, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011731430143117904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0054086870048195125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02887154705822468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.974529093503952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12250, + "weighted_flow_matching_loss": 0.028364061377942562, + "weighted_rkd_loss": 0.028136153891682624 + }, + { + "epoch": 3.5670643002618565, + "flow_matching_loss": 0.028838908672332762, + "grad_norm": 0.2801799476146698, + "learning_rate": 2.9777420982225483e-05, + "loss": 0.0596, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012839415017515421, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005900256428867578, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03157908637076616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9744878232479095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12260, + "weighted_flow_matching_loss": 0.028838908672332762, + "weighted_rkd_loss": 0.030773446336388587 + }, + { + "epoch": 3.5699738143729998, + "flow_matching_loss": 0.028084621950984, + "grad_norm": 0.341900497674942, + "learning_rate": 2.977671095660518e-05, + "loss": 0.0596, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013258713204413653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005783198634162545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03230062499642372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9744465351104736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12270, + "weighted_flow_matching_loss": 0.028084621950984, + "weighted_rkd_loss": 0.03147519044578075 + }, + { + "epoch": 3.572883328484143, + "flow_matching_loss": 0.02810999993234873, + "grad_norm": 0.29026395082473755, + "learning_rate": 2.9775999808791203e-05, + "loss": 0.0582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012552943825721741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005796489585191011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030902377143502236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9744052052497864, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12280, + "weighted_flow_matching_loss": 0.02810999993234873, + "weighted_rkd_loss": 0.030111428536474704 + }, + { + "epoch": 3.5757928425952867, + "flow_matching_loss": 0.02676008567214012, + "grad_norm": 0.3276035487651825, + "learning_rate": 2.9775287538837553e-05, + "loss": 0.0585, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01332500008866191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005969515163451433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032619515992701055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9743638455867767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12290, + "weighted_flow_matching_loss": 0.02676008567214012, + "weighted_rkd_loss": 0.03178326040506363 + }, + { + "epoch": 3.57870235670643, + "flow_matching_loss": 0.02924642153084278, + "grad_norm": 0.28673693537712097, + "learning_rate": 2.9774574146798325e-05, + "loss": 0.0606, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013029271084815264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0060774082317948345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03213595077395439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9743224561214447, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12300, + "weighted_flow_matching_loss": 0.02924642153084278, + "weighted_rkd_loss": 0.03131079971790314 + }, + { + "epoch": 3.5816118708175733, + "flow_matching_loss": 0.026547701843082906, + "grad_norm": 0.3119035065174103, + "learning_rate": 2.9773859632727687e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012255672551691531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005348648689687252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02985999397933483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9742810249328613, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12310, + "weighted_flow_matching_loss": 0.026547701843082906, + "weighted_rkd_loss": 0.029092029482126237 + }, + { + "epoch": 3.584521384928717, + "flow_matching_loss": 0.026980428956449032, + "grad_norm": 0.3780948221683502, + "learning_rate": 2.9773143996679914e-05, + "loss": 0.0594, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013595609366893769, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006071357522159815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033262576535344125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.97423956990242, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12320, + "weighted_flow_matching_loss": 0.026980428956449032, + "weighted_rkd_loss": 0.032405678369104865 + }, + { + "epoch": 3.5874308990398602, + "flow_matching_loss": 0.02676228079944849, + "grad_norm": 0.48129749298095703, + "learning_rate": 2.9772427238709348e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012654368486255407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005963709577918053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031272446177899836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9741980731487274, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12330, + "weighted_flow_matching_loss": 0.02676228079944849, + "weighted_rkd_loss": 0.030465543456375598 + }, + { + "epoch": 3.590340413151004, + "flow_matching_loss": 0.029238788038492204, + "grad_norm": 0.34947893023490906, + "learning_rate": 2.9771709358870424e-05, + "loss": 0.0609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013288341090083122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005963017558678984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03253969997167587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9741565465927124, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12340, + "weighted_flow_matching_loss": 0.029238788038492204, + "weighted_rkd_loss": 0.03169879540801048 + }, + { + "epoch": 3.593249927262147, + "flow_matching_loss": 0.029420300759375094, + "grad_norm": 0.3089447617530823, + "learning_rate": 2.9770990357217656e-05, + "loss": 0.0605, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012923184223473072, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006110541708767414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03195691052824259, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.974115002155304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12350, + "weighted_flow_matching_loss": 0.029420300759375094, + "weighted_rkd_loss": 0.031129669025540353 + }, + { + "epoch": 3.596159441373291, + "flow_matching_loss": 0.02946229074150324, + "grad_norm": 0.2748001217842102, + "learning_rate": 2.977027023380565e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012264728546142578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005671370727941394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030200828239321707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9740734040737152, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12360, + "weighted_flow_matching_loss": 0.02946229074150324, + "weighted_rkd_loss": 0.029417833872139455 + }, + { + "epoch": 3.599068955484434, + "flow_matching_loss": 0.028781019896268845, + "grad_norm": 0.2992002069950104, + "learning_rate": 2.9769548988689102e-05, + "loss": 0.0616, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013744790107011795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061920080333948135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033681588247418404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.974031788110733, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12370, + "weighted_flow_matching_loss": 0.028781019896268845, + "weighted_rkd_loss": 0.03280693609267473 + }, + { + "epoch": 3.6019784695955774, + "flow_matching_loss": 0.030490884184837343, + "grad_norm": 0.31955626606941223, + "learning_rate": 2.976882662192277e-05, + "loss": 0.0635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01370526673272252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006442925706505775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033853459358215335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973990136384964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12380, + "weighted_flow_matching_loss": 0.030490884184837343, + "weighted_rkd_loss": 0.032972926273941994 + }, + { + "epoch": 3.604887983706721, + "flow_matching_loss": 0.027762249298393726, + "grad_norm": 0.2978096306324005, + "learning_rate": 2.9768103133561526e-05, + "loss": 0.0568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012180522177368402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005447239940986037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029808284342288972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9739484548568725, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12390, + "weighted_flow_matching_loss": 0.027762249298393726, + "weighted_rkd_loss": 0.02903170809149742 + }, + { + "epoch": 3.6077974978178644, + "flow_matching_loss": 0.031226140819489955, + "grad_norm": 0.3314782977104187, + "learning_rate": 2.976737852366031e-05, + "loss": 0.0618, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01275633331388235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005899055162444711, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03141172174364328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9739067494869232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12400, + "weighted_flow_matching_loss": 0.031226140819489955, + "weighted_rkd_loss": 0.03059211615473032 + }, + { + "epoch": 3.6107070119290077, + "flow_matching_loss": 0.02948796357959509, + "grad_norm": 0.2829451858997345, + "learning_rate": 2.976665279227415e-05, + "loss": 0.0595, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012502895668148994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005850303685292602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030856094695627688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9738649725914001, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12410, + "weighted_flow_matching_loss": 0.02948796357959509, + "weighted_rkd_loss": 0.030049690417945384 + }, + { + "epoch": 3.6136165260401514, + "flow_matching_loss": 0.028224927186965943, + "grad_norm": 0.32270559668540955, + "learning_rate": 2.9765925939458164e-05, + "loss": 0.0584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012684628739953041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056368589866906404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03100611623376608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9738232016563415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12420, + "weighted_flow_matching_loss": 0.028224927186965943, + "weighted_rkd_loss": 0.03019446078687906 + }, + { + "epoch": 3.6165260401512946, + "flow_matching_loss": 0.030354874953627587, + "grad_norm": 0.3197968602180481, + "learning_rate": 2.9765197965267548e-05, + "loss": 0.0618, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013004053942859172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006234483048319817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03224259130656719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9737813830375671, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12430, + "weighted_flow_matching_loss": 0.030354874953627587, + "weighted_rkd_loss": 0.03139723241329193 + }, + { + "epoch": 3.619435554262438, + "flow_matching_loss": 0.02900103460997343, + "grad_norm": 0.3096475303173065, + "learning_rate": 2.976446886975759e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012355629354715347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006153514096513391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0308647733181715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9737395346164703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12440, + "weighted_flow_matching_loss": 0.02900103460997343, + "weighted_rkd_loss": 0.03005425799638033 + }, + { + "epoch": 3.6223450683735816, + "flow_matching_loss": 0.030397335439920424, + "grad_norm": 0.35861602425575256, + "learning_rate": 2.9763738652983657e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01327291177585721, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006221441179513931, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03276726473122835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9736976504325867, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12450, + "weighted_flow_matching_loss": 0.030397335439920424, + "weighted_rkd_loss": 0.03190537914633751 + }, + { + "epoch": 3.6252545824847253, + "flow_matching_loss": 0.0280171450227499, + "grad_norm": 0.33954399824142456, + "learning_rate": 2.9763007315001204e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01254824111238122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005718308128416538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030814790353178977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973655742406845, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12460, + "weighted_flow_matching_loss": 0.0280171450227499, + "weighted_rkd_loss": 0.030002988688647748 + }, + { + "epoch": 3.6281640965958686, + "flow_matching_loss": 0.02976717408746481, + "grad_norm": 0.3328290283679962, + "learning_rate": 2.9762274855865773e-05, + "loss": 0.0622, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013529529236257077, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006261901743710041, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03332096058875322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9736137986183167, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12470, + "weighted_flow_matching_loss": 0.02976717408746481, + "weighted_rkd_loss": 0.03244173415005207 + }, + { + "epoch": 3.631073610707012, + "flow_matching_loss": 0.028219077736139297, + "grad_norm": 0.28524544835090637, + "learning_rate": 2.9761541275632997e-05, + "loss": 0.0601, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013242648169398307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006245214445516467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03273051120340824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9735718190670013, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12480, + "weighted_flow_matching_loss": 0.028219077736139297, + "weighted_rkd_loss": 0.031865506805479525 + }, + { + "epoch": 3.6339831248181556, + "flow_matching_loss": 0.027821382135152818, + "grad_norm": 0.26528486609458923, + "learning_rate": 2.976080657435857e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012552703730762005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0059397091623395685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03104511685669422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9735298156738281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12490, + "weighted_flow_matching_loss": 0.027821382135152818, + "weighted_rkd_loss": 0.03022336270660162 + }, + { + "epoch": 3.636892638929299, + "flow_matching_loss": 0.027168311551213263, + "grad_norm": 0.3456985354423523, + "learning_rate": 2.9760070752098302e-05, + "loss": 0.0595, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013709956780076027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057992749381810425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033219188638031485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9734877586364746, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12500, + "weighted_flow_matching_loss": 0.027168311551213263, + "weighted_rkd_loss": 0.03233846761286259 + }, + { + "epoch": 3.639802153040442, + "flow_matching_loss": 0.029999813251197337, + "grad_norm": 0.30020853877067566, + "learning_rate": 2.9759333808908064e-05, + "loss": 0.062, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013317023497074842, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006260661780834198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03289470858871937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9734456896781921, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12510, + "weighted_flow_matching_loss": 0.029999813251197337, + "weighted_rkd_loss": 0.03202122114598751 + }, + { + "epoch": 3.642711667151586, + "flow_matching_loss": 0.028517670556902887, + "grad_norm": 0.3759399652481079, + "learning_rate": 2.9758595744843822e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013941260799765588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006781721813604235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03466424308717251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9734035730361938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12520, + "weighted_flow_matching_loss": 0.028517670556902887, + "weighted_rkd_loss": 0.03374228831380606 + }, + { + "epoch": 3.645621181262729, + "flow_matching_loss": 0.027860714681446554, + "grad_norm": 0.26313814520835876, + "learning_rate": 2.975785655996164e-05, + "loss": 0.0576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012447878252714872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005681006051599979, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030576762557029725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9733614385128021, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12530, + "weighted_flow_matching_loss": 0.027860714681446554, + "weighted_rkd_loss": 0.029762270115315913 + }, + { + "epoch": 3.6485306953738723, + "flow_matching_loss": 0.029766321927309037, + "grad_norm": 0.2735705077648163, + "learning_rate": 2.9757116254317645e-05, + "loss": 0.0612, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013085651956498624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006132707325741648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03230401147156954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9733192682266235, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12540, + "weighted_flow_matching_loss": 0.029766321927309037, + "weighted_rkd_loss": 0.031442111171782014 + }, + { + "epoch": 3.651440209485016, + "flow_matching_loss": 0.02674099858850241, + "grad_norm": 0.3424938917160034, + "learning_rate": 2.975637482796806e-05, + "loss": 0.057, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012716239783912897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005652158427983523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031084638088941574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9732770621776581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12550, + "weighted_flow_matching_loss": 0.02674099858850241, + "weighted_rkd_loss": 0.030253944918513298 + }, + { + "epoch": 3.6543497235961593, + "flow_matching_loss": 0.029162352345883845, + "grad_norm": 0.30005478858947754, + "learning_rate": 2.975563228096918e-05, + "loss": 0.0618, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013572472147643566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063934938982129095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03353843837976456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9732348263263703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12560, + "weighted_flow_matching_loss": 0.029162352345883845, + "weighted_rkd_loss": 0.03264079466462135 + }, + { + "epoch": 3.657259237707303, + "flow_matching_loss": 0.028394208289682865, + "grad_norm": 0.31136035919189453, + "learning_rate": 2.9754888613377417e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013244416657835245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006278730370104313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03276756312698126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9731925427913666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12570, + "weighted_flow_matching_loss": 0.028394208289682865, + "weighted_rkd_loss": 0.031889164447784425 + }, + { + "epoch": 3.6601687518184463, + "flow_matching_loss": 0.028319074027240276, + "grad_norm": 0.3165707588195801, + "learning_rate": 2.9754143825249236e-05, + "loss": 0.0595, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013098467234522104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005844851164147258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03204178605228662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973150223493576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12580, + "weighted_flow_matching_loss": 0.028319074027240276, + "weighted_rkd_loss": 0.03118146527558565 + }, + { + "epoch": 3.66307826592959, + "flow_matching_loss": 0.02877997178584337, + "grad_norm": 0.2846425771713257, + "learning_rate": 2.9753397916641197e-05, + "loss": 0.0605, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013210999127477408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061889657285064455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03261096421629191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9731079041957855, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12590, + "weighted_flow_matching_loss": 0.02877997178584337, + "weighted_rkd_loss": 0.031734011881053445 + }, + { + "epoch": 3.6659877800407332, + "flow_matching_loss": 0.029081284999847412, + "grad_norm": 0.35304227471351624, + "learning_rate": 2.9752650887609953e-05, + "loss": 0.0604, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013085385039448737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005993910646066069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032164680771529675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9730655491352082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12600, + "weighted_flow_matching_loss": 0.029081284999847412, + "weighted_rkd_loss": 0.03129835743457079 + }, + { + "epoch": 3.6688972941518765, + "flow_matching_loss": 0.027796689234673978, + "grad_norm": 0.34723737835884094, + "learning_rate": 2.975190273821223e-05, + "loss": 0.0564, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011851851176470519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005718499515205621, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02942220214754343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9730231463909149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12610, + "weighted_flow_matching_loss": 0.027796689234673978, + "weighted_rkd_loss": 0.02862846925854683 + }, + { + "epoch": 3.67180680826302, + "flow_matching_loss": 0.02721273507922888, + "grad_norm": 0.3048883378505707, + "learning_rate": 2.9751153468504852e-05, + "loss": 0.0566, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012304337695240974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005565136577934027, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030173812247812747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9729807138442993, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12620, + "weighted_flow_matching_loss": 0.02721273507922888, + "weighted_rkd_loss": 0.02935854308307171 + }, + { + "epoch": 3.6747163223741635, + "flow_matching_loss": 0.02815895210951567, + "grad_norm": 0.3372126519680023, + "learning_rate": 2.9750403078544715e-05, + "loss": 0.059, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01298541547730565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005685734003782273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03165656439960003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9729382514953613, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12630, + "weighted_flow_matching_loss": 0.02815895210951567, + "weighted_rkd_loss": 0.030799887888133527 + }, + { + "epoch": 3.6776258364853067, + "flow_matching_loss": 0.029347079060971736, + "grad_norm": 0.3044270873069763, + "learning_rate": 2.974965156838881e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011931581515818834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005573190143331885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029436353035271166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.972895759344101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12640, + "weighted_flow_matching_loss": 0.029347079060971736, + "weighted_rkd_loss": 0.02863852437585592 + }, + { + "epoch": 3.6805353505964504, + "flow_matching_loss": 0.02739730179309845, + "grad_norm": 0.29947981238365173, + "learning_rate": 2.9748898938094207e-05, + "loss": 0.0584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012956235092133284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005983990943059325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03189646080136299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9728532254695892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12650, + "weighted_flow_matching_loss": 0.02739730179309845, + "weighted_rkd_loss": 0.03103057835251093 + }, + { + "epoch": 3.6834448647075937, + "flow_matching_loss": 0.027357981353998185, + "grad_norm": 0.31362631916999817, + "learning_rate": 2.9748145187718063e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012478640489280223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005717069376260042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03067435063421726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9728106677532196, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12660, + "weighted_flow_matching_loss": 0.027357981353998185, + "weighted_rkd_loss": 0.029840343818068504 + }, + { + "epoch": 3.6863543788187374, + "flow_matching_loss": 0.029452251456677913, + "grad_norm": 0.3703840374946594, + "learning_rate": 2.9747390317317622e-05, + "loss": 0.0571, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011498461663722991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005434682150371373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02843160554766655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9727680683135986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12670, + "weighted_flow_matching_loss": 0.029452251456677913, + "weighted_rkd_loss": 0.027657321095466612 + }, + { + "epoch": 3.6892638929298807, + "flow_matching_loss": 0.02953434083610773, + "grad_norm": 0.35309040546417236, + "learning_rate": 2.9746634326950215e-05, + "loss": 0.0611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013251240085810422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005958304135128855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0324607839807868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9727254450321198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12680, + "weighted_flow_matching_loss": 0.02953434083610773, + "weighted_rkd_loss": 0.03157540392130613 + }, + { + "epoch": 3.6921734070410244, + "flow_matching_loss": 0.027672965452075003, + "grad_norm": 0.29880717396736145, + "learning_rate": 2.9745877216673243e-05, + "loss": 0.0579, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012539282348006964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005979649536311626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031058214232325555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.972682785987854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12690, + "weighted_flow_matching_loss": 0.027672965452075003, + "weighted_rkd_loss": 0.03020978458225727 + }, + { + "epoch": 3.6950829211521676, + "flow_matching_loss": 0.02922078724950552, + "grad_norm": 0.3666744530200958, + "learning_rate": 2.9745118986544214e-05, + "loss": 0.0616, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013518150709569455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006220022030174732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033256324008107185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.972640085220337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12700, + "weighted_flow_matching_loss": 0.02922078724950552, + "weighted_rkd_loss": 0.03234642706811428 + }, + { + "epoch": 3.697992435263311, + "flow_matching_loss": 0.03078927882015705, + "grad_norm": 0.3123747706413269, + "learning_rate": 2.9744359636620707e-05, + "loss": 0.06, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012285112496465444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005506577715277672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03007680270820856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9725973665714264, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12710, + "weighted_flow_matching_loss": 0.03078927882015705, + "weighted_rkd_loss": 0.0292526138946414 + }, + { + "epoch": 3.7009019493744546, + "flow_matching_loss": 0.02757214233279228, + "grad_norm": 0.25938552618026733, + "learning_rate": 2.974359916696039e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011785959731787444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005288485437631607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02886040508747101, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.972554612159729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12720, + "weighted_flow_matching_loss": 0.02757214233279228, + "weighted_rkd_loss": 0.02806829661130905 + }, + { + "epoch": 3.703811463485598, + "flow_matching_loss": 0.02887528371065855, + "grad_norm": 0.32095634937286377, + "learning_rate": 2.9742837577621017e-05, + "loss": 0.0595, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012678071856498718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006141726952046156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031497870571911335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9725118219852448, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12730, + "weighted_flow_matching_loss": 0.02887528371065855, + "weighted_rkd_loss": 0.030632060021162033 + }, + { + "epoch": 3.706720977596741, + "flow_matching_loss": 0.02925197798758745, + "grad_norm": 0.3002657890319824, + "learning_rate": 2.974207486866042e-05, + "loss": 0.0615, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013388910237699747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006379540357738733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03315736148506403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9724690020084381, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12740, + "weighted_flow_matching_loss": 0.02925197798758745, + "weighted_rkd_loss": 0.03224452249705791 + }, + { + "epoch": 3.709630491707885, + "flow_matching_loss": 0.029596076160669327, + "grad_norm": 0.31903815269470215, + "learning_rate": 2.974131104013653e-05, + "loss": 0.0586, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012124892231076956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005564684420824051, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029814468696713448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9724261522293091, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12750, + "weighted_flow_matching_loss": 0.029596076160669327, + "weighted_rkd_loss": 0.028992389142513276 + }, + { + "epoch": 3.712540005819028, + "flow_matching_loss": 0.028680449165403844, + "grad_norm": 0.42099529504776, + "learning_rate": 2.974054609210735e-05, + "loss": 0.0605, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013297203090041875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006156815681606531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03275122195482254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9723832607269287, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12760, + "weighted_flow_matching_loss": 0.028680449165403844, + "weighted_rkd_loss": 0.0318467166274786 + }, + { + "epoch": 3.715449519930172, + "flow_matching_loss": 0.02848835792392492, + "grad_norm": 0.2502017021179199, + "learning_rate": 2.973978002463097e-05, + "loss": 0.0583, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012417866289615631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057947272434830666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030630459636449815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9723403453826904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12770, + "weighted_flow_matching_loss": 0.02848835792392492, + "weighted_rkd_loss": 0.029783269949257375 + }, + { + "epoch": 3.718359034041315, + "flow_matching_loss": 0.026114690862596036, + "grad_norm": 0.28057634830474854, + "learning_rate": 2.9739012837765574e-05, + "loss": 0.0568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012862833123654127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005834718514233827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031560384295880796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9722973942756653, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12780, + "weighted_flow_matching_loss": 0.026114690862596036, + "weighted_rkd_loss": 0.03068609405308962 + }, + { + "epoch": 3.7212685481524588, + "flow_matching_loss": 0.02747528813779354, + "grad_norm": 0.37668514251708984, + "learning_rate": 2.9738244531569424e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012681456003338098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006049009831622243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031411921977996825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9722544193267822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12790, + "weighted_flow_matching_loss": 0.02747528813779354, + "weighted_rkd_loss": 0.03054036609828472 + }, + { + "epoch": 3.724178062263602, + "flow_matching_loss": 0.027601698786020278, + "grad_norm": 0.34934455156326294, + "learning_rate": 2.9737475106100864e-05, + "loss": 0.059, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01309755714610219, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00608319784514606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032278311997652055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9722113966941833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12800, + "weighted_flow_matching_loss": 0.027601698786020278, + "weighted_rkd_loss": 0.03138132579624653 + }, + { + "epoch": 3.7270875763747453, + "flow_matching_loss": 0.029805569164454936, + "grad_norm": 0.26677408814430237, + "learning_rate": 2.9736704561418327e-05, + "loss": 0.0573, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01149360928684473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005263758357614279, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02825097683817148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9721683442592621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12810, + "weighted_flow_matching_loss": 0.029805569164454936, + "weighted_rkd_loss": 0.027464694529771804 + }, + { + "epoch": 3.729997090485889, + "flow_matching_loss": 0.02852000892162323, + "grad_norm": 0.3440515100955963, + "learning_rate": 2.9735932897580335e-05, + "loss": 0.0576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012115611229091883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005681585799902677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029912808164954185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9721252679824829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12820, + "weighted_flow_matching_loss": 0.02852000892162323, + "weighted_rkd_loss": 0.029078984819352626 + }, + { + "epoch": 3.7329066045970323, + "flow_matching_loss": 0.027017262764275074, + "grad_norm": 0.3629791736602783, + "learning_rate": 2.9735160114645486e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011404722835868597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005066408356651664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027875854074954985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9720821619033814, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12830, + "weighted_flow_matching_loss": 0.027017262764275074, + "weighted_rkd_loss": 0.02709762342274189 + }, + { + "epoch": 3.7358161187081755, + "flow_matching_loss": 0.027772304229438306, + "grad_norm": 0.326363742351532, + "learning_rate": 2.973438621267247e-05, + "loss": 0.0574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012427561357617379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005653771990910173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0305088946595788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9720390141010284, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12840, + "weighted_flow_matching_loss": 0.027772304229438306, + "weighted_rkd_loss": 0.02965583633631468 + }, + { + "epoch": 3.7387256328193192, + "flow_matching_loss": 0.02772321756929159, + "grad_norm": 0.3836861550807953, + "learning_rate": 2.973361119172006e-05, + "loss": 0.061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013816813845187426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006623130477964878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03425675798207521, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9719958305358887, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12850, + "weighted_flow_matching_loss": 0.02772321756929159, + "weighted_rkd_loss": 0.033297397568821906 + }, + { + "epoch": 3.7416351469304625, + "flow_matching_loss": 0.029288486018776892, + "grad_norm": 0.3051215410232544, + "learning_rate": 2.9732835051847114e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011886860895901918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005716305458918214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029490027576684952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9719526171684265, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12860, + "weighted_flow_matching_loss": 0.029288486018776892, + "weighted_rkd_loss": 0.02866288721561432 + }, + { + "epoch": 3.744544661041606, + "flow_matching_loss": 0.029026944376528262, + "grad_norm": 0.3168938457965851, + "learning_rate": 2.9732057793112573e-05, + "loss": 0.0584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012283681612461805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005644821049645543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030212183855473994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9719093859195709, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12870, + "weighted_flow_matching_loss": 0.029026944376528262, + "weighted_rkd_loss": 0.02936351057142019 + }, + { + "epoch": 3.7474541751527495, + "flow_matching_loss": 0.02726154439151287, + "grad_norm": 0.37231549620628357, + "learning_rate": 2.973127941557547e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013199865818023682, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006159777054563165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032559508457779884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9718661069869995, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12880, + "weighted_flow_matching_loss": 0.02726154439151287, + "weighted_rkd_loss": 0.03164346292614937 + }, + { + "epoch": 3.750363689263893, + "flow_matching_loss": 0.027955443225800992, + "grad_norm": 0.27892252802848816, + "learning_rate": 2.9730499919294903e-05, + "loss": 0.0565, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01192382127046585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005506654758937657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029354297183454038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9718227982521057, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12890, + "weighted_flow_matching_loss": 0.027955443225800992, + "weighted_rkd_loss": 0.028527174890041352 + }, + { + "epoch": 3.7532732033750364, + "flow_matching_loss": 0.027522642351686954, + "grad_norm": 0.2918701767921448, + "learning_rate": 2.972971930433009e-05, + "loss": 0.0593, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013253802806138993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006147485924884677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032655091397464274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.971779465675354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12900, + "weighted_flow_matching_loss": 0.027522642351686954, + "weighted_rkd_loss": 0.031733544543385506 + }, + { + "epoch": 3.7561827174861797, + "flow_matching_loss": 0.03061317652463913, + "grad_norm": 0.41985905170440674, + "learning_rate": 2.97289375707403e-05, + "loss": 0.0617, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012890897877514362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006208844622597099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03199064042419195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9717360973358155, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12910, + "weighted_flow_matching_loss": 0.03061317652463913, + "weighted_rkd_loss": 0.031086428463459014 + }, + { + "epoch": 3.7590922315973234, + "flow_matching_loss": 0.02852981723845005, + "grad_norm": 0.38188180327415466, + "learning_rate": 2.9728154718584906e-05, + "loss": 0.0597, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012919718772172928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00622339122928679, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03206282872706652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9716926872730255, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12920, + "weighted_flow_matching_loss": 0.02852981723845005, + "weighted_rkd_loss": 0.031155197136104106 + }, + { + "epoch": 3.7620017457084667, + "flow_matching_loss": 0.027213582396507265, + "grad_norm": 0.3159724771976471, + "learning_rate": 2.9727370747923357e-05, + "loss": 0.0593, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013159894198179246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006752124801278114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0330719131976366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9716492533683777, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12930, + "weighted_flow_matching_loss": 0.027213582396507265, + "weighted_rkd_loss": 0.03213430922478437 + }, + { + "epoch": 3.76491125981961, + "flow_matching_loss": 0.02861669771373272, + "grad_norm": 0.2926499843597412, + "learning_rate": 2.972658565881519e-05, + "loss": 0.0607, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013362981379032135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006299335416406393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03302529845386744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.971605795621872, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12940, + "weighted_flow_matching_loss": 0.02861669771373272, + "weighted_rkd_loss": 0.03208759222179651 + }, + { + "epoch": 3.7678207739307537, + "flow_matching_loss": 0.028415670990943907, + "grad_norm": 0.32377493381500244, + "learning_rate": 2.9725799451320032e-05, + "loss": 0.0604, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01337165515869856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006225969782099128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032969280146062374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9715622961521149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12950, + "weighted_flow_matching_loss": 0.028415670990943907, + "weighted_rkd_loss": 0.03203170020133257 + }, + { + "epoch": 3.770730288041897, + "flow_matching_loss": 0.029772800952196123, + "grad_norm": 0.3622610569000244, + "learning_rate": 2.9725012125497584e-05, + "loss": 0.0592, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012314058747142554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005649242270737887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030277359858155252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9715187549591064, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12960, + "weighted_flow_matching_loss": 0.029772800952196123, + "weighted_rkd_loss": 0.029415004514157773 + }, + { + "epoch": 3.77363980215304, + "flow_matching_loss": 0.026488698646426202, + "grad_norm": 0.33562856912612915, + "learning_rate": 2.9724223681407647e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011705145612359047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005338265746831894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028748556785285474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9714751899242401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12970, + "weighted_flow_matching_loss": 0.026488698646426202, + "weighted_rkd_loss": 0.02792850062251091 + }, + { + "epoch": 3.776549316264184, + "flow_matching_loss": 0.02963775545358658, + "grad_norm": 0.2901800274848938, + "learning_rate": 2.972343411911009e-05, + "loss": 0.0582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011934978980571031, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005524318153038621, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029394276067614557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9714315950870513, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12980, + "weighted_flow_matching_loss": 0.02963775545358658, + "weighted_rkd_loss": 0.02855455633252859 + }, + { + "epoch": 3.7794588303753276, + "flow_matching_loss": 0.026318317838013174, + "grad_norm": 0.3282134532928467, + "learning_rate": 2.9722643438664875e-05, + "loss": 0.0549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011938397772610188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005555843934416771, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029432639293372632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9713879585266113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12990, + "weighted_flow_matching_loss": 0.026318317838013174, + "weighted_rkd_loss": 0.02859049439430237 + }, + { + "epoch": 3.782368344486471, + "flow_matching_loss": 0.029308772273361684, + "grad_norm": 0.34888532757759094, + "learning_rate": 2.9721851640132055e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012376157194375991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005695933755487203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030448247864842414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9713442981243133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13000, + "weighted_flow_matching_loss": 0.029308772273361684, + "weighted_rkd_loss": 0.029575729556381703 + }, + { + "epoch": 3.785277858597614, + "flow_matching_loss": 0.027140679769217968, + "grad_norm": 0.33308497071266174, + "learning_rate": 2.9721058723571757e-05, + "loss": 0.0574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012602313421666622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005922570219263434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031127197109162806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9713006019592285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13010, + "weighted_flow_matching_loss": 0.027140679769217968, + "weighted_rkd_loss": 0.03023392055183649 + }, + { + "epoch": 3.788187372708758, + "flow_matching_loss": 0.029638064838945866, + "grad_norm": 0.315049946308136, + "learning_rate": 2.97202646890442e-05, + "loss": 0.0593, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012308203801512718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005928135616704822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030544543266296388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9712568819522858, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13020, + "weighted_flow_matching_loss": 0.029638064838945866, + "weighted_rkd_loss": 0.029666593298316002 + }, + { + "epoch": 3.791096886819901, + "flow_matching_loss": 0.02952105216681957, + "grad_norm": 0.2989894449710846, + "learning_rate": 2.9719469536609688e-05, + "loss": 0.0576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01169700501486659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005496463784947991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028890473768115044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9712131202220917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13030, + "weighted_flow_matching_loss": 0.02952105216681957, + "weighted_rkd_loss": 0.02805883251130581 + }, + { + "epoch": 3.7940064009310444, + "flow_matching_loss": 0.026313252560794354, + "grad_norm": 0.29307645559310913, + "learning_rate": 2.9718673266328602e-05, + "loss": 0.0555, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012086343299597502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005861875554546714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030034562200307847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9711693227291107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13040, + "weighted_flow_matching_loss": 0.026313252560794354, + "weighted_rkd_loss": 0.029168636724352836 + }, + { + "epoch": 3.796915915042188, + "flow_matching_loss": 0.027873578295111657, + "grad_norm": 0.3085886240005493, + "learning_rate": 2.971787587826142e-05, + "loss": 0.056, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011867882683873177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005176763096824288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028912528418004513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9711255013942719, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13050, + "weighted_flow_matching_loss": 0.027873578295111657, + "weighted_rkd_loss": 0.02807769849896431 + }, + { + "epoch": 3.7998254291533313, + "flow_matching_loss": 0.02720769587904215, + "grad_norm": 0.3467690944671631, + "learning_rate": 2.9717077372468692e-05, + "loss": 0.057, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012461039982736111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005790148582309485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030712228268384933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9710816502571106, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13060, + "weighted_flow_matching_loss": 0.02720769587904215, + "weighted_rkd_loss": 0.02982407305389643 + }, + { + "epoch": 3.8027349432644746, + "flow_matching_loss": 0.028672449477016925, + "grad_norm": 0.35341864824295044, + "learning_rate": 2.9716277749011055e-05, + "loss": 0.0586, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012458052020519972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00588259887881577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030798702500760555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.971037757396698, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13070, + "weighted_flow_matching_loss": 0.028672449477016925, + "weighted_rkd_loss": 0.029906699620187282 + }, + { + "epoch": 3.8056444573756183, + "flow_matching_loss": 0.029094273224473, + "grad_norm": 0.3053334653377533, + "learning_rate": 2.9715477007949253e-05, + "loss": 0.0595, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012758217565715314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005843503121286631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031359938159585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9709938406944275, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13080, + "weighted_flow_matching_loss": 0.029094273224473, + "weighted_rkd_loss": 0.030450287461280822 + }, + { + "epoch": 3.8085539714867616, + "flow_matching_loss": 0.028939617238938807, + "grad_norm": 0.3608214557170868, + "learning_rate": 2.971467514934408e-05, + "loss": 0.0605, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013127501867711545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006283466285094619, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03253846950829029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9709498882293701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13090, + "weighted_flow_matching_loss": 0.028939617238938807, + "weighted_rkd_loss": 0.03159322179853916 + }, + { + "epoch": 3.8114634855979053, + "flow_matching_loss": 0.029428137466311455, + "grad_norm": 0.31838199496269226, + "learning_rate": 2.971387217325644e-05, + "loss": 0.0574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011666599754244089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0054952763486653565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02882847562432289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9709059000015259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13100, + "weighted_flow_matching_loss": 0.029428137466311455, + "weighted_rkd_loss": 0.027989740297198296 + }, + { + "epoch": 3.8143729997090485, + "flow_matching_loss": 0.028448431566357613, + "grad_norm": 0.3177170157432556, + "learning_rate": 2.971306807974731e-05, + "loss": 0.0595, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012974227871745824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0060293173883110285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03197777289897204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9708618879318237, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13110, + "weighted_flow_matching_loss": 0.028448431566357613, + "weighted_rkd_loss": 0.031046003662049772 + }, + { + "epoch": 3.8172825138201922, + "flow_matching_loss": 0.027393285743892194, + "grad_norm": 0.3019932210445404, + "learning_rate": 2.9712262868877757e-05, + "loss": 0.0555, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011697236821055413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0055153372697532175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028909811563789845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9708178400993347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13120, + "weighted_flow_matching_loss": 0.027393285743892194, + "weighted_rkd_loss": 0.028066166304051877 + }, + { + "epoch": 3.8201920279313355, + "flow_matching_loss": 0.030691380612552166, + "grad_norm": 0.3048109710216522, + "learning_rate": 2.971145654070893e-05, + "loss": 0.0616, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012839244026690722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006190808536484837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03186929654330015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9707737565040588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13130, + "weighted_flow_matching_loss": 0.030691380612552166, + "weighted_rkd_loss": 0.030937902815639972 + }, + { + "epoch": 3.8231015420424788, + "flow_matching_loss": 0.028472533263266087, + "grad_norm": 0.3223085403442383, + "learning_rate": 2.971064909530207e-05, + "loss": 0.0582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012430530693382024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005724267661571502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030585328862071038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.970729649066925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13140, + "weighted_flow_matching_loss": 0.028472533263266087, + "weighted_rkd_loss": 0.029690083488821982 + }, + { + "epoch": 3.8260110561536225, + "flow_matching_loss": 0.026256032288074493, + "grad_norm": 0.272136926651001, + "learning_rate": 2.9709840532718493e-05, + "loss": 0.0542, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011624693777412176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005491613270714879, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028741000778973104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9706854939460754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13150, + "weighted_flow_matching_loss": 0.026256032288074493, + "weighted_rkd_loss": 0.0278984846547246 + }, + { + "epoch": 3.8289205702647657, + "flow_matching_loss": 0.028690590150654317, + "grad_norm": 0.34398385882377625, + "learning_rate": 2.9709030853019598e-05, + "loss": 0.0582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012459154427051543, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005508105596527457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030426414497196675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9706413149833679, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13160, + "weighted_flow_matching_loss": 0.028690590150654317, + "weighted_rkd_loss": 0.0295331547036767 + }, + { + "epoch": 3.831830084375909, + "flow_matching_loss": 0.029993423819541933, + "grad_norm": 0.2871508300304413, + "learning_rate": 2.9708220056266887e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011773048248142005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005318127339705825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02886422388255596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9705971062183381, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13170, + "weighted_flow_matching_loss": 0.029993423819541933, + "weighted_rkd_loss": 0.02801555208861828 + }, + { + "epoch": 3.8347395984870527, + "flow_matching_loss": 0.02768658343702555, + "grad_norm": 0.3083377480506897, + "learning_rate": 2.9707408142521924e-05, + "loss": 0.0573, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01233891574665904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005824753642082214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03050258457660675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9705528616905212, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13180, + "weighted_flow_matching_loss": 0.02768658343702555, + "weighted_rkd_loss": 0.02960435803979635 + }, + { + "epoch": 3.837649112598196, + "flow_matching_loss": 0.02749551609158516, + "grad_norm": 0.305925190448761, + "learning_rate": 2.9706595111846378e-05, + "loss": 0.0577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012537333462387324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006026217178441584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031100884452462195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.970508587360382, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13190, + "weighted_flow_matching_loss": 0.02749551609158516, + "weighted_rkd_loss": 0.030183624662458896 + }, + { + "epoch": 3.8405586267093397, + "flow_matching_loss": 0.02689399439841509, + "grad_norm": 0.33496779203414917, + "learning_rate": 2.9705780964301987e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011465187463909388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005076539679430425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02800691444426775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9704642772674561, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13200, + "weighted_flow_matching_loss": 0.02689399439841509, + "weighted_rkd_loss": 0.027179684303700925 + }, + { + "epoch": 3.843468140820483, + "flow_matching_loss": 0.028532958403229714, + "grad_norm": 0.30913618206977844, + "learning_rate": 2.970496569995058e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0128254696726799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005831121606752277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03148206062614918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9704199373722077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13210, + "weighted_flow_matching_loss": 0.028532958403229714, + "weighted_rkd_loss": 0.03055085726082325 + }, + { + "epoch": 3.8463776549316266, + "flow_matching_loss": 0.030761534348130226, + "grad_norm": 0.26397693157196045, + "learning_rate": 2.9704149318854074e-05, + "loss": 0.0581, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011486581712961196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00515478472225368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02812794838100672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9703755676746368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13220, + "weighted_flow_matching_loss": 0.030761534348130226, + "weighted_rkd_loss": 0.02729468382894993 + }, + { + "epoch": 3.84928716904277, + "flow_matching_loss": 0.02809046320617199, + "grad_norm": 0.3014857769012451, + "learning_rate": 2.9703331821074463e-05, + "loss": 0.0576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012325708661228419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005773864220827818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030425281450152396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9703311622142792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13230, + "weighted_flow_matching_loss": 0.02809046320617199, + "weighted_rkd_loss": 0.029522551596164702 + }, + { + "epoch": 3.852196683153913, + "flow_matching_loss": 0.026676779985427855, + "grad_norm": 0.32275792956352234, + "learning_rate": 2.9702513206673838e-05, + "loss": 0.0558, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012223855499178172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005558167537674308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030005878768861294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9702867269515991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13240, + "weighted_flow_matching_loss": 0.026676779985427855, + "weighted_rkd_loss": 0.029114308394491674 + }, + { + "epoch": 3.855106197265057, + "flow_matching_loss": 0.029055096954107285, + "grad_norm": 0.23298364877700806, + "learning_rate": 2.970169347571436e-05, + "loss": 0.0571, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011788672767579556, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005286582675762474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028863928467035293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9702422618865967, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13250, + "weighted_flow_matching_loss": 0.029055096954107285, + "weighted_rkd_loss": 0.028005025722086428 + }, + { + "epoch": 3.8580157113762, + "flow_matching_loss": 0.028837499395012855, + "grad_norm": 0.39501047134399414, + "learning_rate": 2.9700872628258288e-05, + "loss": 0.0584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012179034017026424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006138398218899966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03049646597355604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9701977550983429, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13260, + "weighted_flow_matching_loss": 0.028837499395012855, + "weighted_rkd_loss": 0.029587594047188758 + }, + { + "epoch": 3.8609252254873434, + "flow_matching_loss": 0.029100535064935686, + "grad_norm": 0.33542776107788086, + "learning_rate": 2.9700050664367955e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012041309662163258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005741483392193913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02982410341501236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9701532244682312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13270, + "weighted_flow_matching_loss": 0.029100535064935686, + "weighted_rkd_loss": 0.028933926299214364 + }, + { + "epoch": 3.863834739598487, + "flow_matching_loss": 0.025517894886434078, + "grad_norm": 0.3504047691822052, + "learning_rate": 2.9699227584105785e-05, + "loss": 0.0542, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012093326728790998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00537875471636653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029565408267080782, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9701086580753326, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13280, + "weighted_flow_matching_loss": 0.025517894886434078, + "weighted_rkd_loss": 0.02868164535611868 + }, + { + "epoch": 3.8667442537096304, + "flow_matching_loss": 0.029343554563820363, + "grad_norm": 0.3754524290561676, + "learning_rate": 2.969840338753429e-05, + "loss": 0.0566, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011409223172813654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052645992487668995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02808304559439421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9700640559196472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13290, + "weighted_flow_matching_loss": 0.029343554563820363, + "weighted_rkd_loss": 0.027242347970604895 + }, + { + "epoch": 3.869653767820774, + "flow_matching_loss": 0.02845729235559702, + "grad_norm": 0.2964904308319092, + "learning_rate": 2.9697578074716056e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011955087073147297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005670392489992082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02958056703209877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9700194299221039, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13300, + "weighted_flow_matching_loss": 0.02845729235559702, + "weighted_rkd_loss": 0.02869373857975006 + }, + { + "epoch": 3.8725632819319173, + "flow_matching_loss": 0.026108592003583907, + "grad_norm": 0.3182048499584198, + "learning_rate": 2.969675164571376e-05, + "loss": 0.0554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012270430102944374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005628998996689916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030169859156012536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9699747800827027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13310, + "weighted_flow_matching_loss": 0.026108592003583907, + "weighted_rkd_loss": 0.02926402259618044 + }, + { + "epoch": 3.875472796043061, + "flow_matching_loss": 0.027400464937090873, + "grad_norm": 0.3313969671726227, + "learning_rate": 2.969592410059017e-05, + "loss": 0.0555, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011842466425150633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005278481519781053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028963414393365382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9699300825595856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13320, + "weighted_flow_matching_loss": 0.027400464937090873, + "weighted_rkd_loss": 0.028092454001307488 + }, + { + "epoch": 3.8783823101542043, + "flow_matching_loss": 0.02873739618808031, + "grad_norm": 0.34516656398773193, + "learning_rate": 2.9695095439408132e-05, + "loss": 0.0575, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01207502130419016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005500092008151114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02965013459324837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9698853492736816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13330, + "weighted_flow_matching_loss": 0.02873739618808031, + "weighted_rkd_loss": 0.028757226094603537 + }, + { + "epoch": 3.8812918242653476, + "flow_matching_loss": 0.02793668080121279, + "grad_norm": 0.36019352078437805, + "learning_rate": 2.9694265662230567e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012280004657804966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005590641172602773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03015065025538206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9698405861854553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13340, + "weighted_flow_matching_loss": 0.02793668080121279, + "weighted_rkd_loss": 0.029241287708282472 + }, + { + "epoch": 3.8842013383764913, + "flow_matching_loss": 0.02752911262214184, + "grad_norm": 0.27551525831222534, + "learning_rate": 2.9693434769120508e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011750299111008644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005448607658036053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028949205949902534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9697957992553711, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13350, + "weighted_flow_matching_loss": 0.02752911262214184, + "weighted_rkd_loss": 0.028074833378195763 + }, + { + "epoch": 3.8871108524876345, + "flow_matching_loss": 0.028644006699323654, + "grad_norm": 0.32776808738708496, + "learning_rate": 2.969260276014104e-05, + "loss": 0.0604, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01321016438305378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006333843385800719, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03275417257100344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9697509706020355, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13360, + "weighted_flow_matching_loss": 0.028644006699323654, + "weighted_rkd_loss": 0.031763351149857046 + }, + { + "epoch": 3.890020366598778, + "flow_matching_loss": 0.0293565072119236, + "grad_norm": 0.2974604666233063, + "learning_rate": 2.9691769635355356e-05, + "loss": 0.059, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012540385033935309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005497666518203914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030578436329960823, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.969706118106842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13370, + "weighted_flow_matching_loss": 0.0293565072119236, + "weighted_rkd_loss": 0.029652091674506664 + }, + { + "epoch": 3.8929298807099215, + "flow_matching_loss": 0.02873695157468319, + "grad_norm": 0.35797417163848877, + "learning_rate": 2.9690935394826725e-05, + "loss": 0.0599, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012932087387889623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006315107131376863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03217928167432547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9696612358093262, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13380, + "weighted_flow_matching_loss": 0.02873695157468319, + "weighted_rkd_loss": 0.031203008256852627 + }, + { + "epoch": 3.895839394821065, + "flow_matching_loss": 0.027215837873518466, + "grad_norm": 0.30458977818489075, + "learning_rate": 2.9690100038618508e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0113750952295959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052514973562210795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028001687489449978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.969616311788559, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13390, + "weighted_flow_matching_loss": 0.027215837873518466, + "weighted_rkd_loss": 0.027150871977210046 + }, + { + "epoch": 3.898748908932208, + "flow_matching_loss": 0.02829542476683855, + "grad_norm": 0.25127974152565, + "learning_rate": 2.9689263566794132e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01207472663372755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005611614463850856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0297610679641366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9695713639259338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13400, + "weighted_flow_matching_loss": 0.02829542476683855, + "weighted_rkd_loss": 0.028855489380657674 + }, + { + "epoch": 3.9016584230433518, + "flow_matching_loss": 0.028470135852694513, + "grad_norm": 0.3253576457500458, + "learning_rate": 2.9688425979417136e-05, + "loss": 0.0562, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01159041551873088, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005456179892644286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02863701116293669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9695263803005219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13410, + "weighted_flow_matching_loss": 0.028470135852694513, + "weighted_rkd_loss": 0.027764338254928588 + }, + { + "epoch": 3.9045679371544955, + "flow_matching_loss": 0.027646046318113803, + "grad_norm": 0.3083587884902954, + "learning_rate": 2.968758727655112e-05, + "loss": 0.0565, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012097280845046044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005583651550114155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02977821361273527, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.969481360912323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13420, + "weighted_flow_matching_loss": 0.027646046318113803, + "weighted_rkd_loss": 0.028869439288973807 + }, + { + "epoch": 3.9074774512656387, + "flow_matching_loss": 0.02696374021470547, + "grad_norm": 0.24960088729858398, + "learning_rate": 2.9686747458259776e-05, + "loss": 0.056, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01219355259090662, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005590076977387071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029977181926369667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9694363117218018, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13430, + "weighted_flow_matching_loss": 0.02696374021470547, + "weighted_rkd_loss": 0.02906100433319807 + }, + { + "epoch": 3.910386965376782, + "flow_matching_loss": 0.027002386935055255, + "grad_norm": 0.2984280288219452, + "learning_rate": 2.9685906524606887e-05, + "loss": 0.0586, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013310576789081097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005991473933681846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03261262737214565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9693912327289581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13440, + "weighted_flow_matching_loss": 0.027002386935055255, + "weighted_rkd_loss": 0.03161443471908569 + }, + { + "epoch": 3.9132964794879257, + "flow_matching_loss": 0.026484834402799605, + "grad_norm": 0.28723615407943726, + "learning_rate": 2.9685064475656315e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012186187505722045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005665511172264814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03003788609057665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9693461120128631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13450, + "weighted_flow_matching_loss": 0.026484834402799605, + "weighted_rkd_loss": 0.02911710273474455 + }, + { + "epoch": 3.916205993599069, + "flow_matching_loss": 0.02742168288677931, + "grad_norm": 0.2849438488483429, + "learning_rate": 2.9684221311472013e-05, + "loss": 0.0576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012711993604898452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005673697357997299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031097684428095816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9693009734153748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13460, + "weighted_flow_matching_loss": 0.02742168288677931, + "weighted_rkd_loss": 0.030143053270876408 + }, + { + "epoch": 3.919115507710212, + "flow_matching_loss": 0.027907604910433294, + "grad_norm": 0.276831716299057, + "learning_rate": 2.968337703211801e-05, + "loss": 0.0568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012128074187785387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005601780582219362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02985792886465788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.969255793094635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13470, + "weighted_flow_matching_loss": 0.027907604910433294, + "weighted_rkd_loss": 0.028939989767968653 + }, + { + "epoch": 3.922025021821356, + "flow_matching_loss": 0.026979203894734384, + "grad_norm": 0.3089790940284729, + "learning_rate": 2.968253163765842e-05, + "loss": 0.0546, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011608925368636846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005256494344212115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028474345430731775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9692105889320374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13480, + "weighted_flow_matching_loss": 0.026979203894734384, + "weighted_rkd_loss": 0.027597656659781932 + }, + { + "epoch": 3.924934535932499, + "flow_matching_loss": 0.027800914458930494, + "grad_norm": 0.33659759163856506, + "learning_rate": 2.968168512815745e-05, + "loss": 0.0553, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011515422258526087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005323608405888081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02835445310920477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9691653430461884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13490, + "weighted_flow_matching_loss": 0.027800914458930494, + "weighted_rkd_loss": 0.02748013585805893 + }, + { + "epoch": 3.9278440500436425, + "flow_matching_loss": 0.02860066741704941, + "grad_norm": 0.30492064356803894, + "learning_rate": 2.9680837503679387e-05, + "loss": 0.0582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012376479897648096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005779396556317806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03053235597908497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9691200852394104, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13500, + "weighted_flow_matching_loss": 0.02860066741704941, + "weighted_rkd_loss": 0.02958955932408571 + }, + { + "epoch": 3.930753564154786, + "flow_matching_loss": 0.02688055206090212, + "grad_norm": 0.27667859196662903, + "learning_rate": 2.96799887642886e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011522779148072005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00508731473237276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028132872842252254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9690747797489166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13510, + "weighted_flow_matching_loss": 0.02688055206090212, + "weighted_rkd_loss": 0.0272628640756011 + }, + { + "epoch": 3.9336630782659294, + "flow_matching_loss": 0.026484046876430512, + "grad_norm": 0.3209742605686188, + "learning_rate": 2.9679138910049545e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011159981228411197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049219105392694475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02724187299609184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.969029426574707, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13520, + "weighted_flow_matching_loss": 0.026484046876430512, + "weighted_rkd_loss": 0.02639819011092186 + }, + { + "epoch": 3.936572592377073, + "flow_matching_loss": 0.027726302668452264, + "grad_norm": 0.3613860309123993, + "learning_rate": 2.9678287941026765e-05, + "loss": 0.0543, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011180265340954065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005092409625649452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027452940121293068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9689840495586395, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13530, + "weighted_flow_matching_loss": 0.027726302668452264, + "weighted_rkd_loss": 0.02660145703703165 + }, + { + "epoch": 3.9394821064882164, + "flow_matching_loss": 0.025475494749844074, + "grad_norm": 0.3730306625366211, + "learning_rate": 2.967743585728488e-05, + "loss": 0.0557, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012724741920828819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005709640821442008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031159124709665774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9689386546611786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13540, + "weighted_flow_matching_loss": 0.025475494749844074, + "weighted_rkd_loss": 0.030191263556480406 + }, + { + "epoch": 3.94239162059936, + "flow_matching_loss": 0.02584183420985937, + "grad_norm": 0.30518630146980286, + "learning_rate": 2.9676582658888607e-05, + "loss": 0.0531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011498576123267412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005095082591287792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02809223476797342, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9688932180404664, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13550, + "weighted_flow_matching_loss": 0.02584183420985937, + "weighted_rkd_loss": 0.02721839491277933 + }, + { + "epoch": 3.9453011347105034, + "flow_matching_loss": 0.027070369385182858, + "grad_norm": 0.2914401888847351, + "learning_rate": 2.967572834590274e-05, + "loss": 0.0543, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011464963294565677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005151507677510381, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028081434406340124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9688477575778961, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13560, + "weighted_flow_matching_loss": 0.027070369385182858, + "weighted_rkd_loss": 0.02720663920044899 + }, + { + "epoch": 3.9482106488216466, + "flow_matching_loss": 0.028373962640762328, + "grad_norm": 0.4178442358970642, + "learning_rate": 2.9674872918392152e-05, + "loss": 0.0567, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01192135140299797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005380065785720945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029222768545150758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9688022553920745, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13570, + "weighted_flow_matching_loss": 0.028373962640762328, + "weighted_rkd_loss": 0.028311064280569554 + }, + { + "epoch": 3.9511201629327903, + "flow_matching_loss": 0.02653410565108061, + "grad_norm": 0.3171394467353821, + "learning_rate": 2.9674016376421816e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011565461196005345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005333482194691897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028464404679834843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9687567234039307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13580, + "weighted_flow_matching_loss": 0.02653410565108061, + "weighted_rkd_loss": 0.02757506724447012 + }, + { + "epoch": 3.9540296770439336, + "flow_matching_loss": 0.026210622861981392, + "grad_norm": 0.2715652883052826, + "learning_rate": 2.9673158720056773e-05, + "loss": 0.0538, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011552453972399235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005327933328226208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02843284122645855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9687111616134644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13590, + "weighted_flow_matching_loss": 0.026210622861981392, + "weighted_rkd_loss": 0.027543205581605435 + }, + { + "epoch": 3.956939191155077, + "flow_matching_loss": 0.027928980253636836, + "grad_norm": 0.3071509003639221, + "learning_rate": 2.9672299949362165e-05, + "loss": 0.0574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012355733197182417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005664457520470023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030375923961400986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9686655700206757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13600, + "weighted_flow_matching_loss": 0.027928980253636836, + "weighted_rkd_loss": 0.029424085281789303 + }, + { + "epoch": 3.9598487052662206, + "flow_matching_loss": 0.02631175983697176, + "grad_norm": 0.2940013110637665, + "learning_rate": 2.96714400644032e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01178077757358551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00544478832744062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029006343521177768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9686199426651001, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13610, + "weighted_flow_matching_loss": 0.02631175983697176, + "weighted_rkd_loss": 0.028096117451786994 + }, + { + "epoch": 3.962758219377364, + "flow_matching_loss": 0.028148905374109745, + "grad_norm": 0.34479257464408875, + "learning_rate": 2.9670579065245185e-05, + "loss": 0.0562, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011764401011168957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005382970860227943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028911772556602953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9685742855072021, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13620, + "weighted_flow_matching_loss": 0.028148905374109745, + "weighted_rkd_loss": 0.02800317984074354 + }, + { + "epoch": 3.9656677334885075, + "flow_matching_loss": 0.027169151976704598, + "grad_norm": 0.2883748412132263, + "learning_rate": 2.966971695195351e-05, + "loss": 0.054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011257469095289708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005183638166636228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027698576264083386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9685285866260529, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13630, + "weighted_flow_matching_loss": 0.027169151976704598, + "weighted_rkd_loss": 0.02682684324681759 + }, + { + "epoch": 3.968577247599651, + "flow_matching_loss": 0.026612108573317528, + "grad_norm": 0.33865785598754883, + "learning_rate": 2.966885372459364e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011346493940800428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005188350239768625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02788133770227432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9684828698635102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13640, + "weighted_flow_matching_loss": 0.026612108573317528, + "weighted_rkd_loss": 0.027002582512795924 + }, + { + "epoch": 3.9714867617107945, + "flow_matching_loss": 0.025792904384434222, + "grad_norm": 0.27468371391296387, + "learning_rate": 2.966798938323114e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011642538849264384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005391951417550445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028677029348909856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.968437111377716, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13650, + "weighted_flow_matching_loss": 0.025792904384434222, + "weighted_rkd_loss": 0.027771885320544244 + }, + { + "epoch": 3.9743962758219378, + "flow_matching_loss": 0.027055289037525652, + "grad_norm": 0.31402313709259033, + "learning_rate": 2.9667123927931643e-05, + "loss": 0.0543, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011485479399561882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005130928475409746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028101887181401254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9683913230895996, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13660, + "weighted_flow_matching_loss": 0.027055289037525652, + "weighted_rkd_loss": 0.02721363361924887 + }, + { + "epoch": 3.977305789933081, + "flow_matching_loss": 0.027526299469172955, + "grad_norm": 0.33828282356262207, + "learning_rate": 2.966625735876088e-05, + "loss": 0.0561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011955473106354475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005642485711723566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029553431645035745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9683455049991607, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13670, + "weighted_flow_matching_loss": 0.027526299469172955, + "weighted_rkd_loss": 0.028617927618324757 + }, + { + "epoch": 3.9802153040442247, + "flow_matching_loss": 0.02651790380477905, + "grad_norm": 0.35361674427986145, + "learning_rate": 2.966538967578466e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011583634745329618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005329084978438914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02849635425955057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9682996571063995, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13680, + "weighted_flow_matching_loss": 0.02651790380477905, + "weighted_rkd_loss": 0.027592979185283185 + }, + { + "epoch": 3.983124818155368, + "flow_matching_loss": 0.02935158796608448, + "grad_norm": 0.28968217968940735, + "learning_rate": 2.9664520879068873e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01155039481818676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005322228884324431, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02842301856726408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9682537615299225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13690, + "weighted_flow_matching_loss": 0.02935158796608448, + "weighted_rkd_loss": 0.027520694024860858 + }, + { + "epoch": 3.9860343322665113, + "flow_matching_loss": 0.026871115528047086, + "grad_norm": 0.3260597884654999, + "learning_rate": 2.966365096867951e-05, + "loss": 0.055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011818033549934626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005451181088574231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02908724844455719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.968207859992981, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13700, + "weighted_flow_matching_loss": 0.026871115528047086, + "weighted_rkd_loss": 0.028162537142634392 + }, + { + "epoch": 3.988943846377655, + "flow_matching_loss": 0.028406647592782976, + "grad_norm": 0.29474198818206787, + "learning_rate": 2.9662779944682625e-05, + "loss": 0.0562, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011595504172146321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005538500240072608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028729508630931377, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9681618988513947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13710, + "weighted_flow_matching_loss": 0.028406647592782976, + "weighted_rkd_loss": 0.027814795821905137 + }, + { + "epoch": 3.9918533604887982, + "flow_matching_loss": 0.025602718256413935, + "grad_norm": 0.37420570850372314, + "learning_rate": 2.9661907807144367e-05, + "loss": 0.0554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012488312553614378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005758225126191974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030734850838780404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9681159257888794, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13720, + "weighted_flow_matching_loss": 0.025602718256413935, + "weighted_rkd_loss": 0.029754899255931378 + }, + { + "epoch": 3.994762874599942, + "flow_matching_loss": 0.026708690263330937, + "grad_norm": 0.3079204857349396, + "learning_rate": 2.9661034556130973e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011587250791490079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0054556784685701135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0286301800981164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9680699050426483, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13730, + "weighted_flow_matching_loss": 0.026708690263330937, + "weighted_rkd_loss": 0.027716006338596343 + }, + { + "epoch": 3.997672388711085, + "flow_matching_loss": 0.027852177247405054, + "grad_norm": 0.26884445548057556, + "learning_rate": 2.966016019170876e-05, + "loss": 0.0539, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010920695681124925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005051988735795021, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026893380098044872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9680238664150238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13740, + "weighted_flow_matching_loss": 0.027852177247405054, + "weighted_rkd_loss": 0.026033436693251134 + }, + { + "epoch": 4.000581902822229, + "flow_matching_loss": 0.027403277531266212, + "grad_norm": 0.2831171452999115, + "learning_rate": 2.9659284713944127e-05, + "loss": 0.0561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01231577293947339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004991774959489703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029623320512473583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9679777801036835, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13750, + "weighted_flow_matching_loss": 0.027403277531266212, + "weighted_rkd_loss": 0.028674724511802197 + }, + { + "epoch": 4.003491416933372, + "flow_matching_loss": 0.02631546091288328, + "grad_norm": 0.3270809054374695, + "learning_rate": 2.9658408122903565e-05, + "loss": 0.057, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012955874763429165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00583185413852334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03174360375851393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9679316818714142, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13760, + "weighted_flow_matching_loss": 0.02631546091288328, + "weighted_rkd_loss": 0.030725622177124025 + }, + { + "epoch": 4.0064009310445154, + "flow_matching_loss": 0.027328317798674107, + "grad_norm": 0.31635379791259766, + "learning_rate": 2.9657530418653638e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011994542367756367, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005222455924376845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029211540706455708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9678855419158936, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13770, + "weighted_flow_matching_loss": 0.027328317798674107, + "weighted_rkd_loss": 0.02827340252697468 + }, + { + "epoch": 4.009310445155659, + "flow_matching_loss": 0.027755224518477915, + "grad_norm": 0.2896119952201843, + "learning_rate": 2.9656651601261014e-05, + "loss": 0.0539, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011032100860029459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004939793515950442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027003995142877102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9678393602371216, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13780, + "weighted_flow_matching_loss": 0.027755224518477915, + "weighted_rkd_loss": 0.026135510206222533 + }, + { + "epoch": 4.012219959266803, + "flow_matching_loss": 0.028147760219871996, + "grad_norm": 0.2638031840324402, + "learning_rate": 2.965577167079242e-05, + "loss": 0.0576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012329083308577538, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005782657396048307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030440823920071126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9677931666374207, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13790, + "weighted_flow_matching_loss": 0.028147760219871996, + "weighted_rkd_loss": 0.029460435174405573 + }, + { + "epoch": 4.015129473377946, + "flow_matching_loss": 0.027045803144574165, + "grad_norm": 0.3638290464878082, + "learning_rate": 2.965489062731469e-05, + "loss": 0.0555, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011997565627098083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005371363391168416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02936649452894926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9677469253540039, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13800, + "weighted_flow_matching_loss": 0.027045803144574165, + "weighted_rkd_loss": 0.02841933649033308 + }, + { + "epoch": 4.018038987489089, + "flow_matching_loss": 0.027890866808593272, + "grad_norm": 0.2754981815814972, + "learning_rate": 2.9654008470894733e-05, + "loss": 0.0573, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012336217053234576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005693343840539456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030365778133273125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9677006483078003, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13810, + "weighted_flow_matching_loss": 0.027890866808593272, + "weighted_rkd_loss": 0.02938502114266157 + }, + { + "epoch": 4.020948501600233, + "flow_matching_loss": 0.026930526457726957, + "grad_norm": 0.3033384382724762, + "learning_rate": 2.9653125201599538e-05, + "loss": 0.0553, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011802252568304538, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005700867855921388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02930537275969982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9676543533802032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13820, + "weighted_flow_matching_loss": 0.026930526457726957, + "weighted_rkd_loss": 0.0283574253320694 + }, + { + "epoch": 4.023858015711376, + "flow_matching_loss": 0.02941749934107065, + "grad_norm": 0.33227404952049255, + "learning_rate": 2.9652240819496184e-05, + "loss": 0.0563, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011178700439631938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005466939625330269, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027824340388178825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9676080226898194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13830, + "weighted_flow_matching_loss": 0.02941749934107065, + "weighted_rkd_loss": 0.026923032104969026 + }, + { + "epoch": 4.02676752982252, + "flow_matching_loss": 0.027703217417001724, + "grad_norm": 0.35549071431159973, + "learning_rate": 2.965135532465184e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0112154858186841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004847387922927738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02727835960686207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9675616562366486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13840, + "weighted_flow_matching_loss": 0.027703217417001724, + "weighted_rkd_loss": 0.02639350574463606 + }, + { + "epoch": 4.029677043933663, + "flow_matching_loss": 0.027299553528428076, + "grad_norm": 0.3149397075176239, + "learning_rate": 2.965046871713375e-05, + "loss": 0.0554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011714735440909863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005620340327732265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029049811139702796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9675152599811554, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13850, + "weighted_flow_matching_loss": 0.027299553528428076, + "weighted_rkd_loss": 0.028106130473315716 + }, + { + "epoch": 4.032586558044806, + "flow_matching_loss": 0.027756514959037304, + "grad_norm": 0.353768914937973, + "learning_rate": 2.964958099700924e-05, + "loss": 0.0574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012386212032288313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005912042735144496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03068446684628725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9674688339233398, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13860, + "weighted_flow_matching_loss": 0.027756514959037304, + "weighted_rkd_loss": 0.029686239548027515 + }, + { + "epoch": 4.03549607215595, + "flow_matching_loss": 0.027269625291228294, + "grad_norm": 0.3262771666049957, + "learning_rate": 2.9648692164345738e-05, + "loss": 0.0543, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011441448982805014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005104170367121697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02798706851899624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.967422366142273, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13870, + "weighted_flow_matching_loss": 0.027269625291228294, + "weighted_rkd_loss": 0.02707528751343489 + }, + { + "epoch": 4.038405586267094, + "flow_matching_loss": 0.029598343744874002, + "grad_norm": 0.3198982775211334, + "learning_rate": 2.9647802219210737e-05, + "loss": 0.0582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01202372694388032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005535813048481941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029583267122507095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9673758745193481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13880, + "weighted_flow_matching_loss": 0.029598343744874002, + "weighted_rkd_loss": 0.028618137538433074 + }, + { + "epoch": 4.041315100378237, + "flow_matching_loss": 0.02576069552451372, + "grad_norm": 0.3558005094528198, + "learning_rate": 2.9646911161671826e-05, + "loss": 0.0548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01209552977234125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005825161887332797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0300162211060524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.967329353094101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13890, + "weighted_flow_matching_loss": 0.02576069552451372, + "weighted_rkd_loss": 0.029035562090575694 + }, + { + "epoch": 4.04422461448938, + "flow_matching_loss": 0.024586107395589352, + "grad_norm": 0.2820574939250946, + "learning_rate": 2.964601899179667e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012031870242208243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051630036905407906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029226743802428245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9672827959060669, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13900, + "weighted_flow_matching_loss": 0.024586107395589352, + "weighted_rkd_loss": 0.028270531818270682 + }, + { + "epoch": 4.047134128600524, + "flow_matching_loss": 0.027776043489575387, + "grad_norm": 0.33329907059669495, + "learning_rate": 2.964512570965303e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011618449445813894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005511103849858046, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02874800283461809, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9672362029552459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13910, + "weighted_flow_matching_loss": 0.027776043489575387, + "weighted_rkd_loss": 0.02780609969049692 + }, + { + "epoch": 4.0500436427116675, + "flow_matching_loss": 0.029716173931956292, + "grad_norm": 0.2840656340122223, + "learning_rate": 2.964423131530874e-05, + "loss": 0.0566, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011317463219165802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00513675450347364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027771680802106857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9671895802021027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13920, + "weighted_flow_matching_loss": 0.029716173931956292, + "weighted_rkd_loss": 0.02686049398034811 + }, + { + "epoch": 4.05295315682281, + "flow_matching_loss": 0.0295438090339303, + "grad_norm": 0.35761067271232605, + "learning_rate": 2.9643335808831728e-05, + "loss": 0.0571, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011520214844495058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005411365767940879, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02845179531723261, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.967142927646637, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13930, + "weighted_flow_matching_loss": 0.0295438090339303, + "weighted_rkd_loss": 0.027516941353678704 + }, + { + "epoch": 4.055862670933954, + "flow_matching_loss": 0.026255127973854542, + "grad_norm": 0.2927597463130951, + "learning_rate": 2.9642439190289996e-05, + "loss": 0.0566, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012858283426612615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005702077457681298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03141864463686943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9670962393283844, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13940, + "weighted_flow_matching_loss": 0.026255127973854542, + "weighted_rkd_loss": 0.030384847335517405 + }, + { + "epoch": 4.058772185045098, + "flow_matching_loss": 0.028306414000689985, + "grad_norm": 0.2773386538028717, + "learning_rate": 2.964154145975164e-05, + "loss": 0.0553, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011314106080681086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053294899873435496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027957702241837978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.967049527168274, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13950, + "weighted_flow_matching_loss": 0.028306414000689985, + "weighted_rkd_loss": 0.027036487869918347 + }, + { + "epoch": 4.0616816991562406, + "flow_matching_loss": 0.026018293015658854, + "grad_norm": 0.2474839836359024, + "learning_rate": 2.9640642617284836e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011204146686941386, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00524829535279423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027656588703393936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9670027792453766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13960, + "weighted_flow_matching_loss": 0.026018293015658854, + "weighted_rkd_loss": 0.026743998005986213 + }, + { + "epoch": 4.064591213267384, + "flow_matching_loss": 0.02918685171753168, + "grad_norm": 0.2685488164424896, + "learning_rate": 2.9639742662957846e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011742178350687027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00519165052101016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028676006942987442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9669559895992279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13970, + "weighted_flow_matching_loss": 0.02918685171753168, + "weighted_rkd_loss": 0.027728460542857648 + }, + { + "epoch": 4.067500727378528, + "flow_matching_loss": 0.02734316010028124, + "grad_norm": 0.26238229870796204, + "learning_rate": 2.9638841596839012e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010932857263833284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005198179697617889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02706389408558607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9669091880321503, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13980, + "weighted_flow_matching_loss": 0.02734316010028124, + "weighted_rkd_loss": 0.02616835031658411 + }, + { + "epoch": 4.070410241489671, + "flow_matching_loss": 0.027082542330026625, + "grad_norm": 0.2883733808994293, + "learning_rate": 2.9637939418996766e-05, + "loss": 0.0574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012708168104290962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005923939938656986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03134027663618326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9668623328208923, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13990, + "weighted_flow_matching_loss": 0.027082542330026625, + "weighted_rkd_loss": 0.030301734618842602 + }, + { + "epoch": 4.0733197556008145, + "flow_matching_loss": 0.028996407613158227, + "grad_norm": 0.2819787561893463, + "learning_rate": 2.9637036129499628e-05, + "loss": 0.0564, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01150456266477704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005352892726659775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028362017683684827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9668154656887055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14000, + "weighted_flow_matching_loss": 0.028996407613158227, + "weighted_rkd_loss": 0.0274208327755332 + }, + { + "epoch": 4.076229269711958, + "flow_matching_loss": 0.026728216931223868, + "grad_norm": 0.3104046583175659, + "learning_rate": 2.963613172841619e-05, + "loss": 0.0538, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011393610667437316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005224404484033584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028011626191437246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9667685508728028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14010, + "weighted_flow_matching_loss": 0.026728216931223868, + "weighted_rkd_loss": 0.02708076424896717 + }, + { + "epoch": 4.079138783823102, + "flow_matching_loss": 0.02841180358082056, + "grad_norm": 0.2873464822769165, + "learning_rate": 2.9635226215815133e-05, + "loss": 0.056, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01171634728088975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005138337286189199, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028571032173931598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9667216122150422, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14020, + "weighted_flow_matching_loss": 0.02841180358082056, + "weighted_rkd_loss": 0.027620247192680837 + }, + { + "epoch": 4.082048297934245, + "flow_matching_loss": 0.02735273875296116, + "grad_norm": 0.30505871772766113, + "learning_rate": 2.9634319591765237e-05, + "loss": 0.0551, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011763904243707657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005206756922416389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02873456533998251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9666746318340301, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14030, + "weighted_flow_matching_loss": 0.02735273875296116, + "weighted_rkd_loss": 0.02777697555720806 + }, + { + "epoch": 4.084957812045388, + "flow_matching_loss": 0.02879700195044279, + "grad_norm": 0.2910770773887634, + "learning_rate": 2.963341185633534e-05, + "loss": 0.0577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012136902194470167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005589583097025752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029863387532532214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9666276335716247, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14040, + "weighted_flow_matching_loss": 0.02879700195044279, + "weighted_rkd_loss": 0.028866758197546007 + }, + { + "epoch": 4.087867326156532, + "flow_matching_loss": 0.027659283950924873, + "grad_norm": 0.3852434754371643, + "learning_rate": 2.9632503009594386e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010896367114037275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00495207984931767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026744814217090608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9665805995464325, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14050, + "weighted_flow_matching_loss": 0.027659283950924873, + "weighted_rkd_loss": 0.025850982405245304 + }, + { + "epoch": 4.090776840267675, + "flow_matching_loss": 0.027362342551350595, + "grad_norm": 0.3611926734447479, + "learning_rate": 2.9631593051611395e-05, + "loss": 0.0543, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011378256697207689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005127509031444788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02788402233272791, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9665335118770599, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14060, + "weighted_flow_matching_loss": 0.027362342551350595, + "weighted_rkd_loss": 0.026950832828879355 + }, + { + "epoch": 4.093686354378819, + "flow_matching_loss": 0.024609065614640713, + "grad_norm": 0.26198479533195496, + "learning_rate": 2.9630681982455472e-05, + "loss": 0.0493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010333806835114956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004898980120196938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02556659374386072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9664864242076874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14070, + "weighted_flow_matching_loss": 0.024609065614640713, + "weighted_rkd_loss": 0.02470978759229183 + }, + { + "epoch": 4.096595868489962, + "flow_matching_loss": 0.026508463360369206, + "grad_norm": 0.31518319249153137, + "learning_rate": 2.962976980219581e-05, + "loss": 0.0551, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01202211445197463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005541061703115702, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029585291258990766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.966439288854599, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14080, + "weighted_flow_matching_loss": 0.026508463360369206, + "weighted_rkd_loss": 0.028592334873974324 + }, + { + "epoch": 4.099505382601105, + "flow_matching_loss": 0.02719938289374113, + "grad_norm": 0.33367013931274414, + "learning_rate": 2.962885651090167e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011459946259856224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005286652548238635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028206545300781725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9663921236991883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14090, + "weighted_flow_matching_loss": 0.02719938289374113, + "weighted_rkd_loss": 0.02725856564939022 + }, + { + "epoch": 4.102414896712249, + "flow_matching_loss": 0.029397854581475258, + "grad_norm": 0.27289852499961853, + "learning_rate": 2.9627942108642427e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01054010195657611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004904323955997825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025984528101980688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9663449287414551, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14100, + "weighted_flow_matching_loss": 0.029397854581475258, + "weighted_rkd_loss": 0.02511001322418451 + }, + { + "epoch": 4.105324410823393, + "flow_matching_loss": 0.027050002850592135, + "grad_norm": 0.3087314963340759, + "learning_rate": 2.962702659548752e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011417829245328904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051174575695767995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027953116036951543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.966297698020935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14110, + "weighted_flow_matching_loss": 0.027050002850592135, + "weighted_rkd_loss": 0.027011049166321754 + }, + { + "epoch": 4.108233924934536, + "flow_matching_loss": 0.025886940024793147, + "grad_norm": 0.3451271057128906, + "learning_rate": 2.962610997150647e-05, + "loss": 0.0531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0114713404327631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052653011865913864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02820798195898533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9662504374980927, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14120, + "weighted_flow_matching_loss": 0.025886940024793147, + "weighted_rkd_loss": 0.02725598532706499 + }, + { + "epoch": 4.111143439045679, + "flow_matching_loss": 0.027715365029871465, + "grad_norm": 0.2958535850048065, + "learning_rate": 2.9625192236768884e-05, + "loss": 0.057, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012377758510410786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005536482064053416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030291999317705633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9662031471729279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14130, + "weighted_flow_matching_loss": 0.027715365029871465, + "weighted_rkd_loss": 0.029268219135701656 + }, + { + "epoch": 4.114052953156823, + "flow_matching_loss": 0.02664227783679962, + "grad_norm": 0.24854977428913116, + "learning_rate": 2.9624273391344475e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011403701826930046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004991503152996302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02779890689998865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9661558151245118, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14140, + "weighted_flow_matching_loss": 0.02664227783679962, + "weighted_rkd_loss": 0.026858095824718476 + }, + { + "epoch": 4.1169624672679666, + "flow_matching_loss": 0.026779770851135254, + "grad_norm": 0.32285958528518677, + "learning_rate": 2.9623353435303015e-05, + "loss": 0.0548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011795760318636894, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005461640516296029, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02905316147953272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9661084711551666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14150, + "weighted_flow_matching_loss": 0.026779770851135254, + "weighted_rkd_loss": 0.02806851714849472 + }, + { + "epoch": 4.119871981379109, + "flow_matching_loss": 0.026913177408277988, + "grad_norm": 0.31633198261260986, + "learning_rate": 2.962243236871436e-05, + "loss": 0.0551, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011865466367453337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0054976946208626035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029228627867996692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9660610735416413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14160, + "weighted_flow_matching_loss": 0.026913177408277988, + "weighted_rkd_loss": 0.028236629255115987 + }, + { + "epoch": 4.122781495490253, + "flow_matching_loss": 0.02666933201253414, + "grad_norm": 0.2782754600048065, + "learning_rate": 2.962151019164847e-05, + "loss": 0.0553, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011958492640405893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005670689349062741, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029587674513459207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9660136580467225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14170, + "weighted_flow_matching_loss": 0.02666933201253414, + "weighted_rkd_loss": 0.0285821046680212 + }, + { + "epoch": 4.125691009601397, + "flow_matching_loss": 0.024849380552768707, + "grad_norm": 0.247942715883255, + "learning_rate": 2.9620586904175376e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011620660312473774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005547817260958254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028789138048887254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9659661948680878, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14180, + "weighted_flow_matching_loss": 0.024849380552768707, + "weighted_rkd_loss": 0.027809350937604903 + }, + { + "epoch": 4.12860052371254, + "flow_matching_loss": 0.026064147241413593, + "grad_norm": 0.28841596841812134, + "learning_rate": 2.961966250636519e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011497267335653306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005410139425657689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02840467430651188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9659187078475953, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14190, + "weighted_flow_matching_loss": 0.026064147241413593, + "weighted_rkd_loss": 0.02743660341948271 + }, + { + "epoch": 4.131510037823683, + "flow_matching_loss": 0.02643804531544447, + "grad_norm": 0.2693885862827301, + "learning_rate": 2.9618736998288122e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011129588447511196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005285607394762337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02754478417336941, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9658712029457093, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14200, + "weighted_flow_matching_loss": 0.02643804531544447, + "weighted_rkd_loss": 0.02660475578159094 + }, + { + "epoch": 4.134419551934827, + "flow_matching_loss": 0.02783497180789709, + "grad_norm": 0.26536738872528076, + "learning_rate": 2.961781038001445e-05, + "loss": 0.0568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012210060190409423, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005603341921232641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030023462325334548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.965823644399643, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14210, + "weighted_flow_matching_loss": 0.02783497180789709, + "weighted_rkd_loss": 0.028997422195971012 + }, + { + "epoch": 4.137329066045971, + "flow_matching_loss": 0.025732751935720444, + "grad_norm": 0.2764129638671875, + "learning_rate": 2.961688265161455e-05, + "loss": 0.0524, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01135898008942604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004890632070600987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027608592063188553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9657760679721832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14220, + "weighted_flow_matching_loss": 0.025732751935720444, + "weighted_rkd_loss": 0.026663716696202756 + }, + { + "epoch": 4.1402385801571135, + "flow_matching_loss": 0.02496427670121193, + "grad_norm": 0.25703859329223633, + "learning_rate": 2.9615953813158877e-05, + "loss": 0.052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011446239240467549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005082550784572959, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027975029684603213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9657284557819367, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14230, + "weighted_flow_matching_loss": 0.02496427670121193, + "weighted_rkd_loss": 0.027016281522810458 + }, + { + "epoch": 4.143148094268257, + "flow_matching_loss": 0.026159568503499032, + "grad_norm": 0.2934229075908661, + "learning_rate": 2.961502386471797e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010916740261018276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004856374068185687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026689854636788368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9656808197498321, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14240, + "weighted_flow_matching_loss": 0.026159568503499032, + "weighted_rkd_loss": 0.025773898512125016 + }, + { + "epoch": 4.146057608379401, + "flow_matching_loss": 0.02822255976498127, + "grad_norm": 0.29231908917427063, + "learning_rate": 2.961409280636245e-05, + "loss": 0.0576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01229364862665534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005866923788562417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03045422099530697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9656331539154053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14250, + "weighted_flow_matching_loss": 0.02822255976498127, + "weighted_rkd_loss": 0.029407611675560476 + }, + { + "epoch": 4.148967122490544, + "flow_matching_loss": 0.027069106698036194, + "grad_norm": 0.3076808452606201, + "learning_rate": 2.9613160638163022e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010564194526523352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005009378190152347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026137767359614372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9655854165554046, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14260, + "weighted_flow_matching_loss": 0.027069106698036194, + "weighted_rkd_loss": 0.02523821499198675 + }, + { + "epoch": 4.1518766366016875, + "flow_matching_loss": 0.02551446259021759, + "grad_norm": 0.29658448696136475, + "learning_rate": 2.9612227360190486e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011200267076492309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004977844259701669, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027378378435969353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.965537691116333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14270, + "weighted_flow_matching_loss": 0.02551446259021759, + "weighted_rkd_loss": 0.026434843055903913 + }, + { + "epoch": 4.154786150712831, + "flow_matching_loss": 0.026999876461923124, + "grad_norm": 0.3025854825973511, + "learning_rate": 2.9611292972515715e-05, + "loss": 0.0539, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011374691687524318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051137658301740885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0278631491586566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9654899120330811, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14280, + "weighted_flow_matching_loss": 0.026999876461923124, + "weighted_rkd_loss": 0.026901589520275593 + }, + { + "epoch": 4.157695664823974, + "flow_matching_loss": 0.026486547850072383, + "grad_norm": 0.3414483964443207, + "learning_rate": 2.9610357475209667e-05, + "loss": 0.0561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012515387311577798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005641630617901683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03067240510135889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9654421091079712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14290, + "weighted_flow_matching_loss": 0.026486547850072383, + "weighted_rkd_loss": 0.02961242701858282 + }, + { + "epoch": 4.160605178935118, + "flow_matching_loss": 0.02698430959135294, + "grad_norm": 0.2871284484863281, + "learning_rate": 2.960942086834339e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01099037816748023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004977821838110685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026958578452467918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.965394276380539, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14300, + "weighted_flow_matching_loss": 0.02698430959135294, + "weighted_rkd_loss": 0.026025643572211266 + }, + { + "epoch": 4.163514693046261, + "flow_matching_loss": 0.026514748670160772, + "grad_norm": 0.3275584280490875, + "learning_rate": 2.960848315198801e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011214249767363071, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004906225600279868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027334725111722948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9653464078903198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14310, + "weighted_flow_matching_loss": 0.026514748670160772, + "weighted_rkd_loss": 0.02638746127486229 + }, + { + "epoch": 4.166424207157405, + "flow_matching_loss": 0.02670150585472584, + "grad_norm": 0.33155229687690735, + "learning_rate": 2.9607544326214747e-05, + "loss": 0.0547, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011808437574654818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005338321020826697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028955196402966975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9652985095977783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14320, + "weighted_flow_matching_loss": 0.02670150585472584, + "weighted_rkd_loss": 0.027950406633317472 + }, + { + "epoch": 4.169333721268548, + "flow_matching_loss": 0.025347601622343063, + "grad_norm": 0.3189598023891449, + "learning_rate": 2.9606604391094888e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011137898359447718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00486823390237987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027144030295312406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.96525057554245, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14330, + "weighted_flow_matching_loss": 0.025347601622343063, + "weighted_rkd_loss": 0.026200783252716065 + }, + { + "epoch": 4.172243235379692, + "flow_matching_loss": 0.02446103561669588, + "grad_norm": 0.31474626064300537, + "learning_rate": 2.9605663346699824e-05, + "loss": 0.0499, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010767466016113758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004788813483901322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026323745772242547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9652025997638702, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14340, + "weighted_flow_matching_loss": 0.02446103561669588, + "weighted_rkd_loss": 0.025407738238573074 + }, + { + "epoch": 4.175152749490835, + "flow_matching_loss": 0.029096063412725925, + "grad_norm": 0.27966293692588806, + "learning_rate": 2.9604721193101017e-05, + "loss": 0.0546, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010758940689265728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004892958630807698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026410840451717377, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.965154618024826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14350, + "weighted_flow_matching_loss": 0.029096063412725925, + "weighted_rkd_loss": 0.025490562058985235 + }, + { + "epoch": 4.178062263601978, + "flow_matching_loss": 0.02773665152490139, + "grad_norm": 0.2828695476055145, + "learning_rate": 2.9603777930370018e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010945858340710402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00481302528642118, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02670474201440811, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.965106588602066, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14360, + "weighted_flow_matching_loss": 0.02773665152490139, + "weighted_rkd_loss": 0.025772925838828085 + }, + { + "epoch": 4.180971777713122, + "flow_matching_loss": 0.026555696874856947, + "grad_norm": 0.3020399808883667, + "learning_rate": 2.960283355857846e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011104435101151467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005048537347465754, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027257407270371913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9650585234165192, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14370, + "weighted_flow_matching_loss": 0.026555696874856947, + "weighted_rkd_loss": 0.02630499191582203 + }, + { + "epoch": 4.183881291824266, + "flow_matching_loss": 0.027676854468882083, + "grad_norm": 0.26945817470550537, + "learning_rate": 2.9601888077798064e-05, + "loss": 0.0547, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011334632989019156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053231422789394855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027992408163845538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9650104403495788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14380, + "weighted_flow_matching_loss": 0.027676854468882083, + "weighted_rkd_loss": 0.02701298613101244 + }, + { + "epoch": 4.186790805935408, + "flow_matching_loss": 0.027200613170862198, + "grad_norm": 0.2898646593093872, + "learning_rate": 2.960094148810064e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011173942778259515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005006824154406786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02735470961779356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9649623215198517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14390, + "weighted_flow_matching_loss": 0.027200613170862198, + "weighted_rkd_loss": 0.02639626245945692 + }, + { + "epoch": 4.189700320046552, + "flow_matching_loss": 0.02853945717215538, + "grad_norm": 0.3554503619670868, + "learning_rate": 2.9599993789558057e-05, + "loss": 0.0575, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012082726415246724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005838827369734645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030004280619323254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9649141609668732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14400, + "weighted_flow_matching_loss": 0.02853945717215538, + "weighted_rkd_loss": 0.028951520845294 + }, + { + "epoch": 4.192609834157696, + "flow_matching_loss": 0.026955915428698064, + "grad_norm": 0.34406307339668274, + "learning_rate": 2.9599044982242302e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01089825453236699, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049459774047136305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026742486469447612, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9648659765720368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14410, + "weighted_flow_matching_loss": 0.026955915428698064, + "weighted_rkd_loss": 0.025802887231111526 + }, + { + "epoch": 4.195519348268839, + "flow_matching_loss": 0.028701861202716828, + "grad_norm": 0.28443700075149536, + "learning_rate": 2.959809506622543e-05, + "loss": 0.0562, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011626360192894935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005260644340887665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02851336430758238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9648177564144135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14420, + "weighted_flow_matching_loss": 0.028701861202716828, + "weighted_rkd_loss": 0.027510226145386694 + }, + { + "epoch": 4.198428862379982, + "flow_matching_loss": 0.026790831983089448, + "grad_norm": 0.3279111087322235, + "learning_rate": 2.959714404157957e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011628590896725655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005316671053878963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02857385277748108, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9647695124149323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14430, + "weighted_flow_matching_loss": 0.026790831983089448, + "weighted_rkd_loss": 0.027567157335579396 + }, + { + "epoch": 4.201338376491126, + "flow_matching_loss": 0.027205392718315125, + "grad_norm": 0.2642003893852234, + "learning_rate": 2.9596191908376955e-05, + "loss": 0.0539, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011235264223068952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005181284202262759, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027651812881231308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9647212326526642, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14440, + "weighted_flow_matching_loss": 0.027205392718315125, + "weighted_rkd_loss": 0.026676321402192116 + }, + { + "epoch": 4.20424789060227, + "flow_matching_loss": 0.029346407763659953, + "grad_norm": 0.23066875338554382, + "learning_rate": 2.9595238666689893e-05, + "loss": 0.0561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011278460267931223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052008026978001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027757723443210126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9646729111671448, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14450, + "weighted_flow_matching_loss": 0.029346407763659953, + "weighted_rkd_loss": 0.0267771540209651 + }, + { + "epoch": 4.207157404713413, + "flow_matching_loss": 0.0257515624165535, + "grad_norm": 0.31522682309150696, + "learning_rate": 2.9594284316590773e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010847465135157108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004784683976322412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02647961415350437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.964624571800232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14460, + "weighted_flow_matching_loss": 0.0257515624165535, + "weighted_rkd_loss": 0.025542858988046646 + }, + { + "epoch": 4.210066918824556, + "flow_matching_loss": 0.026285459101200104, + "grad_norm": 0.32128822803497314, + "learning_rate": 2.959332885815207e-05, + "loss": 0.0547, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01195746213197708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005585454637184739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029500378668308257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9645761847496033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14470, + "weighted_flow_matching_loss": 0.026285459101200104, + "weighted_rkd_loss": 0.028455368988215924 + }, + { + "epoch": 4.2129764329357, + "flow_matching_loss": 0.025945229083299638, + "grad_norm": 0.2610456347465515, + "learning_rate": 2.9592372291446353e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010449751000851392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004631996317766607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025531498529016972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9645277738571167, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14480, + "weighted_flow_matching_loss": 0.025945229083299638, + "weighted_rkd_loss": 0.024625853449106217 + }, + { + "epoch": 4.215885947046843, + "flow_matching_loss": 0.026198641024529935, + "grad_norm": 0.2657729983329773, + "learning_rate": 2.959141461654626e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009874164592474699, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043739672284573315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024122296646237374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9644793391227722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14490, + "weighted_flow_matching_loss": 0.026198641024529935, + "weighted_rkd_loss": 0.023265472799539565 + }, + { + "epoch": 4.2187954611579865, + "flow_matching_loss": 0.027336325868964194, + "grad_norm": 0.2887745201587677, + "learning_rate": 2.959045583352452e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010798579268157481, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050415641628205774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0266387227922678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9644308686256409, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14500, + "weighted_flow_matching_loss": 0.027336325868964194, + "weighted_rkd_loss": 0.025691216439008714 + }, + { + "epoch": 4.22170497526913, + "flow_matching_loss": 0.026113887876272203, + "grad_norm": 0.31214314699172974, + "learning_rate": 2.958949594245395e-05, + "loss": 0.0525, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01113194515928626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005075241485610604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027339131757616996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9643823564052582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14510, + "weighted_flow_matching_loss": 0.026113887876272203, + "weighted_rkd_loss": 0.02636538092046976 + }, + { + "epoch": 4.224614489380274, + "flow_matching_loss": 0.026763302646577357, + "grad_norm": 0.35016196966171265, + "learning_rate": 2.9588534943407444e-05, + "loss": 0.0546, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011673388443887234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005557626811787486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028904403559863567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9643338203430176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14520, + "weighted_flow_matching_loss": 0.026763302646577357, + "weighted_rkd_loss": 0.02787348423153162 + }, + { + "epoch": 4.227524003491417, + "flow_matching_loss": 0.028675445914268495, + "grad_norm": 0.2403384894132614, + "learning_rate": 2.958757283645799e-05, + "loss": 0.0562, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011593129485845566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005337669583968818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02852392867207527, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9642852544784546, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14530, + "weighted_flow_matching_loss": 0.028675445914268495, + "weighted_rkd_loss": 0.02750520408153534 + }, + { + "epoch": 4.2304335176025605, + "flow_matching_loss": 0.02925674393773079, + "grad_norm": 0.3283948600292206, + "learning_rate": 2.9586609621678643e-05, + "loss": 0.055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010857981629669666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004948237212374806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026664200611412524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9642366528511047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14540, + "weighted_flow_matching_loss": 0.02925674393773079, + "weighted_rkd_loss": 0.025710569694638252 + }, + { + "epoch": 4.233343031713704, + "flow_matching_loss": 0.026300404965877534, + "grad_norm": 0.27325311303138733, + "learning_rate": 2.958564529914256e-05, + "loss": 0.0508, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010375523753464223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004634823743253946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02538587134331465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.964188015460968, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14550, + "weighted_flow_matching_loss": 0.026300404965877534, + "weighted_rkd_loss": 0.02447674572467804 + }, + { + "epoch": 4.236252545824847, + "flow_matching_loss": 0.02717321999371052, + "grad_norm": 0.35668259859085083, + "learning_rate": 2.9584679868922973e-05, + "loss": 0.0538, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011215517483651637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005224743019789457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027655778080224992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9641393542289733, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14560, + "weighted_flow_matching_loss": 0.02717321999371052, + "weighted_rkd_loss": 0.02666399348527193 + }, + { + "epoch": 4.239162059935991, + "flow_matching_loss": 0.024972752667963505, + "grad_norm": 0.2596094012260437, + "learning_rate": 2.95837133310932e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011901643220335245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056843028869479895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02948758900165558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9640906572341919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14570, + "weighted_flow_matching_loss": 0.024972752667963505, + "weighted_rkd_loss": 0.028428738377988337 + }, + { + "epoch": 4.242071574047134, + "flow_matching_loss": 0.026826751232147217, + "grad_norm": 0.29089275002479553, + "learning_rate": 2.9582745685726646e-05, + "loss": 0.0553, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01194680305197835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00563663721550256, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02953024338930845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.964041930437088, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14580, + "weighted_flow_matching_loss": 0.026826751232147217, + "weighted_rkd_loss": 0.02846839986741543 + }, + { + "epoch": 4.244981088158277, + "flow_matching_loss": 0.02879106756299734, + "grad_norm": 0.2717079818248749, + "learning_rate": 2.9581776932896792e-05, + "loss": 0.0562, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011571511626243591, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005310759018175304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028453782387077808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9639931678771972, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14590, + "weighted_flow_matching_loss": 0.02879106756299734, + "weighted_rkd_loss": 0.027429243177175523 + }, + { + "epoch": 4.247890602269421, + "flow_matching_loss": 0.02766853403300047, + "grad_norm": 0.2725946605205536, + "learning_rate": 2.9580807072677215e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010012738686054946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044764465186744925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02450192403048277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9639443814754486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14600, + "weighted_flow_matching_loss": 0.02766853403300047, + "weighted_rkd_loss": 0.023618486523628236 + }, + { + "epoch": 4.250800116380565, + "flow_matching_loss": 0.02629689872264862, + "grad_norm": 0.25376105308532715, + "learning_rate": 2.9579836105141562e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010592331318184734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004808516753837466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02599317915737629, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9638955593109131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14610, + "weighted_flow_matching_loss": 0.02629689872264862, + "weighted_rkd_loss": 0.025054766051471233 + }, + { + "epoch": 4.2537096304917075, + "flow_matching_loss": 0.02677214574068785, + "grad_norm": 0.27557095885276794, + "learning_rate": 2.9578864030363577e-05, + "loss": 0.0543, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01159424353390932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005370032275095582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02855851948261261, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9638466835021973, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14620, + "weighted_flow_matching_loss": 0.02677214574068785, + "weighted_rkd_loss": 0.02752606347203255 + }, + { + "epoch": 4.256619144602851, + "flow_matching_loss": 0.026232253946363925, + "grad_norm": 0.24928483366966248, + "learning_rate": 2.9577890848417083e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010741356667131186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005029771523550153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02651248462498188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9637978076934814, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14630, + "weighted_flow_matching_loss": 0.026232253946363925, + "weighted_rkd_loss": 0.025552668049931525 + }, + { + "epoch": 4.259528658713995, + "flow_matching_loss": 0.027335985749959945, + "grad_norm": 0.3173639178276062, + "learning_rate": 2.9576916559375983e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011547633819282055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005104554304853082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028199822083115576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9637488961219788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14640, + "weighted_flow_matching_loss": 0.027335985749959945, + "weighted_rkd_loss": 0.027177521772682667 + }, + { + "epoch": 4.262438172825139, + "flow_matching_loss": 0.027400927245616914, + "grad_norm": 0.2374511957168579, + "learning_rate": 2.957594116331427e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01097676893696189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005002927267923951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026956465281546115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9636999487876892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14650, + "weighted_flow_matching_loss": 0.027400927245616914, + "weighted_rkd_loss": 0.02597792837768793 + }, + { + "epoch": 4.265347686936281, + "flow_matching_loss": 0.026766899041831493, + "grad_norm": 0.3229216933250427, + "learning_rate": 2.9574964660306017e-05, + "loss": 0.0542, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01152240987867117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005464081512764096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02850890103727579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9636509656906128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14660, + "weighted_flow_matching_loss": 0.026766899041831493, + "weighted_rkd_loss": 0.027472622506320477 + }, + { + "epoch": 4.268257201047425, + "flow_matching_loss": 0.02663260269910097, + "grad_norm": 0.29998984932899475, + "learning_rate": 2.9573987050425385e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011083187814801932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048878195229917765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02705419510602951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9636019468307495, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14670, + "weighted_flow_matching_loss": 0.02663260269910097, + "weighted_rkd_loss": 0.026069485954940318 + }, + { + "epoch": 4.271166715158569, + "flow_matching_loss": 0.028376254253089428, + "grad_norm": 0.2870756983757019, + "learning_rate": 2.957300833374662e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010579607263207436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049036913085728885, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026062905974686145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9635529041290283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14680, + "weighted_flow_matching_loss": 0.028376254253089428, + "weighted_rkd_loss": 0.025112972036004065 + }, + { + "epoch": 4.274076229269712, + "flow_matching_loss": 0.028183619678020477, + "grad_norm": 0.26489463448524475, + "learning_rate": 2.957202851034404e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010228023305535317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004581594979390502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025037641637027264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9635038256645203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14690, + "weighted_flow_matching_loss": 0.028183619678020477, + "weighted_rkd_loss": 0.02412385679781437 + }, + { + "epoch": 4.276985743380855, + "flow_matching_loss": 0.027819396741688252, + "grad_norm": 0.2418234795331955, + "learning_rate": 2.957104758029206e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011179860401898622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005303086643107235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027662807330489158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9634547173976898, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14700, + "weighted_flow_matching_loss": 0.027819396741688252, + "weighted_rkd_loss": 0.026651888340711593 + }, + { + "epoch": 4.279895257491999, + "flow_matching_loss": 0.026696300134062767, + "grad_norm": 0.31918591260910034, + "learning_rate": 2.9570065543665184e-05, + "loss": 0.0542, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011668379418551922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052626708522439, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02859942987561226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.963405579328537, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14710, + "weighted_flow_matching_loss": 0.026696300134062767, + "weighted_rkd_loss": 0.02755284421145916 + }, + { + "epoch": 4.282804771603142, + "flow_matching_loss": 0.02642238661646843, + "grad_norm": 0.24906331300735474, + "learning_rate": 2.9569082400537983e-05, + "loss": 0.054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011687055509537459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00520319992210716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028577310778200627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9633564054965973, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14720, + "weighted_flow_matching_loss": 0.02642238661646843, + "weighted_rkd_loss": 0.027530172280967234 + }, + { + "epoch": 4.285714285714286, + "flow_matching_loss": 0.02814547307789326, + "grad_norm": 0.33725789189338684, + "learning_rate": 2.9568098150985114e-05, + "loss": 0.0559, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011701717134565115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053844907321035865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028787925280630587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9633072018623352, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14730, + "weighted_flow_matching_loss": 0.02814547307789326, + "weighted_rkd_loss": 0.02773162480443716 + }, + { + "epoch": 4.288623799825429, + "flow_matching_loss": 0.026153484359383583, + "grad_norm": 0.33277735114097595, + "learning_rate": 2.9567112795081336e-05, + "loss": 0.0511, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010497273504734039, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004941995535045862, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0259365426376462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9632579684257507, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14740, + "weighted_flow_matching_loss": 0.026153484359383583, + "weighted_rkd_loss": 0.024983553774654867 + }, + { + "epoch": 4.291533313936573, + "flow_matching_loss": 0.02692895829677582, + "grad_norm": 0.3319658637046814, + "learning_rate": 2.956612633290148e-05, + "loss": 0.0511, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010255208797752857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046202272176742555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025130644999444486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.963208693265915, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14750, + "weighted_flow_matching_loss": 0.02692895829677582, + "weighted_rkd_loss": 0.024206057749688627 + }, + { + "epoch": 4.294442828047716, + "flow_matching_loss": 0.02847529426217079, + "grad_norm": 0.3026524484157562, + "learning_rate": 2.9565138764520452e-05, + "loss": 0.0552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01129600228741765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051197406835854055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02771174516528845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9631594002246857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14760, + "weighted_flow_matching_loss": 0.02847529426217079, + "weighted_rkd_loss": 0.026690849848091603 + }, + { + "epoch": 4.2973523421588595, + "flow_matching_loss": 0.02541551347821951, + "grad_norm": 0.29438549280166626, + "learning_rate": 2.956415009001326e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011201353650540114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00496673493180424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027369442209601403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9631100654602051, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14770, + "weighted_flow_matching_loss": 0.02541551347821951, + "weighted_rkd_loss": 0.026359772495925426 + }, + { + "epoch": 4.300261856270003, + "flow_matching_loss": 0.02758054919540882, + "grad_norm": 0.30816423892974854, + "learning_rate": 2.956316030945498e-05, + "loss": 0.0538, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011089953454211354, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005094660236500204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027274567261338235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9630607008934021, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14780, + "weighted_flow_matching_loss": 0.02758054919540882, + "weighted_rkd_loss": 0.026267068833112715 + }, + { + "epoch": 4.303171370381146, + "flow_matching_loss": 0.026803189516067506, + "grad_norm": 0.2686375677585602, + "learning_rate": 2.9562169422920784e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010376554448157549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00462283156812191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025375940650701524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9630113065242767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14790, + "weighted_flow_matching_loss": 0.026803189516067506, + "weighted_rkd_loss": 0.02443730216473341 + }, + { + "epoch": 4.30608088449229, + "flow_matching_loss": 0.02608464676886797, + "grad_norm": 0.2615009844303131, + "learning_rate": 2.9561177430485926e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010744779650121928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004800669034011662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026290228590369225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.962961882352829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14800, + "weighted_flow_matching_loss": 0.02608464676886797, + "weighted_rkd_loss": 0.025316494144499303 + }, + { + "epoch": 4.3089903986034335, + "flow_matching_loss": 0.02700482904911041, + "grad_norm": 0.31706804037094116, + "learning_rate": 2.9560184332225734e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011445732135325671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005562229501083493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02845369353890419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9629124104976654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14810, + "weighted_flow_matching_loss": 0.02700482904911041, + "weighted_rkd_loss": 0.027398349717259407 + }, + { + "epoch": 4.311899912714576, + "flow_matching_loss": 0.027286323346197607, + "grad_norm": 0.3041384816169739, + "learning_rate": 2.9559190128215627e-05, + "loss": 0.0549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011587313562631606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005522562935948372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02869719062000513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9628629386425018, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14820, + "weighted_flow_matching_loss": 0.027286323346197607, + "weighted_rkd_loss": 0.02763143554329872 + }, + { + "epoch": 4.31480942682572, + "flow_matching_loss": 0.02508534025400877, + "grad_norm": 0.28477194905281067, + "learning_rate": 2.9558194818531117e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010961357038468122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004846944101154804, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02676965855062008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.962813401222229, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14830, + "weighted_flow_matching_loss": 0.02508534025400877, + "weighted_rkd_loss": 0.025774205476045607 + }, + { + "epoch": 4.317718940936864, + "flow_matching_loss": 0.025394799932837487, + "grad_norm": 0.2823173701763153, + "learning_rate": 2.955719840324779e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011426384374499322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005356603395193815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028209372237324715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9627638518810272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14840, + "weighted_flow_matching_loss": 0.025394799932837487, + "weighted_rkd_loss": 0.027158970944583416 + }, + { + "epoch": 4.320628455048007, + "flow_matching_loss": 0.027871724404394626, + "grad_norm": 0.29748600721359253, + "learning_rate": 2.9556200882441303e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010730252135545015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050884657073765995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026548969745635986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9627142667770385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14850, + "weighted_flow_matching_loss": 0.027871724404394626, + "weighted_rkd_loss": 0.0255590720102191 + }, + { + "epoch": 4.32353796915915, + "flow_matching_loss": 0.026138755679130554, + "grad_norm": 0.2836916148662567, + "learning_rate": 2.9555202256187432e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010235004965215922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004819163284264505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025289173051714898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.962664645910263, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14860, + "weighted_flow_matching_loss": 0.026138755679130554, + "weighted_rkd_loss": 0.024344976618885995 + }, + { + "epoch": 4.326447483270294, + "flow_matching_loss": 0.027508435770869255, + "grad_norm": 0.2914380431175232, + "learning_rate": 2.9554202524562e-05, + "loss": 0.0537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011043735593557358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00516874932218343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027256220392882825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9626150071620941, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14870, + "weighted_flow_matching_loss": 0.027508435770869255, + "weighted_rkd_loss": 0.026237252168357372 + }, + { + "epoch": 4.329356997381438, + "flow_matching_loss": 0.026335588097572325, + "grad_norm": 0.24894122779369354, + "learning_rate": 2.955320168764094e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011842072010040283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0054382467176765205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0291223905980587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9625653266906739, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14880, + "weighted_flow_matching_loss": 0.026335588097572325, + "weighted_rkd_loss": 0.02803221307694912 + }, + { + "epoch": 4.3322665114925805, + "flow_matching_loss": 0.026073152758181097, + "grad_norm": 0.3211744427680969, + "learning_rate": 2.9552199745500252e-05, + "loss": 0.0522, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011049814149737358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005003654072061181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027103282138705255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9625156104564667, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14890, + "weighted_flow_matching_loss": 0.026073152758181097, + "weighted_rkd_loss": 0.026087306998670103 + }, + { + "epoch": 4.335176025603724, + "flow_matching_loss": 0.027000742219388484, + "grad_norm": 0.2805027365684509, + "learning_rate": 2.9551196698216026e-05, + "loss": 0.052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010642271395772696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047236901940777894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026008233055472375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9624658703804017, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14900, + "weighted_flow_matching_loss": 0.027000742219388484, + "weighted_rkd_loss": 0.025032032653689383 + }, + { + "epoch": 4.338085539714868, + "flow_matching_loss": 0.02691067010164261, + "grad_norm": 0.31932976841926575, + "learning_rate": 2.9550192545864442e-05, + "loss": 0.052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010670160967856645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004715728247538209, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026056050136685372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9624160885810852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14910, + "weighted_flow_matching_loss": 0.02691067010164261, + "weighted_rkd_loss": 0.025076756626367568 + }, + { + "epoch": 4.340995053826011, + "flow_matching_loss": 0.02645995244383812, + "grad_norm": 0.25846195220947266, + "learning_rate": 2.954918728852176e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0096623701043427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004273164155893028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023597904480993748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9623662829399109, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14920, + "weighted_flow_matching_loss": 0.02645995244383812, + "weighted_rkd_loss": 0.02270984072238207 + }, + { + "epoch": 4.343904567937154, + "flow_matching_loss": 0.026136834919452668, + "grad_norm": 0.2437972128391266, + "learning_rate": 2.954818092626432e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010249299462884665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004524612100794911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025023211166262625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9623164415359498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14930, + "weighted_flow_matching_loss": 0.026136834919452668, + "weighted_rkd_loss": 0.024080246686935425 + }, + { + "epoch": 4.346814082048298, + "flow_matching_loss": 0.02655438520014286, + "grad_norm": 0.28289666771888733, + "learning_rate": 2.954717345916854e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01110637690871954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052585761528462175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027471329644322397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9622665762901306, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14940, + "weighted_flow_matching_loss": 0.02655438520014286, + "weighted_rkd_loss": 0.02643476501107216 + }, + { + "epoch": 4.349723596159441, + "flow_matching_loss": 0.026566199027001858, + "grad_norm": 0.31926438212394714, + "learning_rate": 2.954616488731095e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010627064108848571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00474708192050457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026001210138201714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9622166752815247, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14950, + "weighted_flow_matching_loss": 0.026566199027001858, + "weighted_rkd_loss": 0.025018766149878503 + }, + { + "epoch": 4.352633110270585, + "flow_matching_loss": 0.025043147802352905, + "grad_norm": 0.2560657262802124, + "learning_rate": 2.9545155210768126e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011044481489807367, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049055512761697175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026994513906538488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9621667385101318, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14960, + "weighted_flow_matching_loss": 0.025043147802352905, + "weighted_rkd_loss": 0.025973188132047652 + }, + { + "epoch": 4.355542624381728, + "flow_matching_loss": 0.02579595111310482, + "grad_norm": 0.2625141739845276, + "learning_rate": 2.954414442961676e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011359348893165588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005543804261833429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028262502141296865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9621167719364166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14970, + "weighted_flow_matching_loss": 0.02579595111310482, + "weighted_rkd_loss": 0.027191852033138276 + }, + { + "epoch": 4.358452138492872, + "flow_matching_loss": 0.028461663424968718, + "grad_norm": 0.3048444986343384, + "learning_rate": 2.95431325439336e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010417049750685692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005040853447280824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02587495297193527, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9620667695999146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14980, + "weighted_flow_matching_loss": 0.028461663424968718, + "weighted_rkd_loss": 0.024893447384238244 + }, + { + "epoch": 4.361361652604015, + "flow_matching_loss": 0.025488576479256154, + "grad_norm": 0.3030805289745331, + "learning_rate": 2.9542119553795503e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010797670483589173, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004936800268478691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02653214130550623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.962016761302948, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14990, + "weighted_flow_matching_loss": 0.025488576479256154, + "weighted_rkd_loss": 0.02552436161786318 + }, + { + "epoch": 4.364271166715159, + "flow_matching_loss": 0.025693446584045886, + "grad_norm": 0.28827565908432007, + "learning_rate": 2.9541105459279395e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010171756520867348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004565509594976902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024909022636711596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9619666934013367, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15000, + "weighted_flow_matching_loss": 0.025693446584045886, + "weighted_rkd_loss": 0.023961652256548404 + }, + { + "epoch": 4.367180680826302, + "flow_matching_loss": 0.02559107095003128, + "grad_norm": 0.2922222316265106, + "learning_rate": 2.954009026046229e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010895997937768697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048067259136587385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02659872192889452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.961916595697403, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15010, + "weighted_flow_matching_loss": 0.02559107095003128, + "weighted_rkd_loss": 0.025585759431123734 + }, + { + "epoch": 4.370090194937445, + "flow_matching_loss": 0.028992803208529948, + "grad_norm": 0.34054329991340637, + "learning_rate": 2.9539073957421285e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010434361360967159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00461624360177666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025484966486692427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9618664681911469, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15020, + "weighted_flow_matching_loss": 0.028992803208529948, + "weighted_rkd_loss": 0.02451313380151987 + }, + { + "epoch": 4.372999709048589, + "flow_matching_loss": 0.027260355465114117, + "grad_norm": 0.26756182312965393, + "learning_rate": 2.953805655023357e-05, + "loss": 0.0531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010889590065926314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005094612739048899, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026873792894184588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9618163108825684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15030, + "weighted_flow_matching_loss": 0.027260355465114117, + "weighted_rkd_loss": 0.02584766112267971 + }, + { + "epoch": 4.3759092231597325, + "flow_matching_loss": 0.02735651209950447, + "grad_norm": 0.2547847628593445, + "learning_rate": 2.9537038038976393e-05, + "loss": 0.0542, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011454905197024346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00501887509599328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027928685396909715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9617661237716675, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15040, + "weighted_flow_matching_loss": 0.02735651209950447, + "weighted_rkd_loss": 0.026860851049423217 + }, + { + "epoch": 4.378818737270876, + "flow_matching_loss": 0.02640341278165579, + "grad_norm": 0.23191581666469574, + "learning_rate": 2.9536018423727116e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010536041762679815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004838929534889758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025911012664437294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9617159068584442, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15050, + "weighted_flow_matching_loss": 0.02640341278165579, + "weighted_rkd_loss": 0.024919041618704795 + }, + { + "epoch": 4.381728251382019, + "flow_matching_loss": 0.02503595259040594, + "grad_norm": 0.26005202531814575, + "learning_rate": 2.953499770456317e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011670152842998504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005274112755432725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028614418767392635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9616656601428986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15060, + "weighted_flow_matching_loss": 0.02503595259040594, + "weighted_rkd_loss": 0.02751749213784933 + }, + { + "epoch": 4.384637765493163, + "flow_matching_loss": 0.025814596563577652, + "grad_norm": 0.3282230496406555, + "learning_rate": 2.9533975881562073e-05, + "loss": 0.0554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012527551129460334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005711491266265511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03076659347862005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9616153717041016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15070, + "weighted_flow_matching_loss": 0.025814596563577652, + "weighted_rkd_loss": 0.029585647396743298 + }, + { + "epoch": 4.3875472796043065, + "flow_matching_loss": 0.02555778045207262, + "grad_norm": 0.2950439155101776, + "learning_rate": 2.9532952954801422e-05, + "loss": 0.0499, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01028407271951437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004797224584035575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025365369953215124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9615650653839112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15080, + "weighted_flow_matching_loss": 0.02555778045207262, + "weighted_rkd_loss": 0.024390440434217453 + }, + { + "epoch": 4.390456793715449, + "flow_matching_loss": 0.02718770634382963, + "grad_norm": 0.2985314428806305, + "learning_rate": 2.95319289243589e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01019010441377759, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004295087466016412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024675296060740946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9615147113800049, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15090, + "weighted_flow_matching_loss": 0.02718770634382963, + "weighted_rkd_loss": 0.023725672252476217 + }, + { + "epoch": 4.393366307826593, + "flow_matching_loss": 0.02541099414229393, + "grad_norm": 0.2963924705982208, + "learning_rate": 2.9530903790312284e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01199378790333867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005153193254955113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02914076913148165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9614643394947052, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15100, + "weighted_flow_matching_loss": 0.02541099414229393, + "weighted_rkd_loss": 0.028017793036997318 + }, + { + "epoch": 4.396275821937737, + "flow_matching_loss": 0.02624759152531624, + "grad_norm": 0.28951674699783325, + "learning_rate": 2.9529877552739414e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010715469159185887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004970738012343645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026401676051318646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9614139258861542, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15110, + "weighted_flow_matching_loss": 0.02624759152531624, + "weighted_rkd_loss": 0.02538294605910778 + }, + { + "epoch": 4.3991853360488795, + "flow_matching_loss": 0.025506841577589513, + "grad_norm": 0.26573699712753296, + "learning_rate": 2.952885021171824e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009885718021541834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004411251284182072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024182686768472194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9613634943962097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15120, + "weighted_flow_matching_loss": 0.025506841577589513, + "weighted_rkd_loss": 0.02324836440384388 + }, + { + "epoch": 4.402094850160023, + "flow_matching_loss": 0.026350581273436546, + "grad_norm": 0.20488083362579346, + "learning_rate": 2.9527821767326774e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009838483855128288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004573960462585092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02425092812627554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9613130152225494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15130, + "weighted_flow_matching_loss": 0.026350581273436546, + "weighted_rkd_loss": 0.023312772624194622 + }, + { + "epoch": 4.405004364271167, + "flow_matching_loss": 0.028429919481277467, + "grad_norm": 0.23866085708141327, + "learning_rate": 2.9526792219643112e-05, + "loss": 0.0532, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010439709853380918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004934381949715317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025813801772892477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9612625062465667, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15140, + "weighted_flow_matching_loss": 0.028429919481277467, + "weighted_rkd_loss": 0.024813848733901977 + }, + { + "epoch": 4.40791387838231, + "flow_matching_loss": 0.02539002411067486, + "grad_norm": 0.27519822120666504, + "learning_rate": 2.952576156874546e-05, + "loss": 0.0499, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010413554310798646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004723909450694919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025551017932593824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9612119793891907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15150, + "weighted_flow_matching_loss": 0.02539002411067486, + "weighted_rkd_loss": 0.02455995436757803 + }, + { + "epoch": 4.4108233924934535, + "flow_matching_loss": 0.026088137179613113, + "grad_norm": 0.23038989305496216, + "learning_rate": 2.952472981471207e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009957819385454059, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047532770782709125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024668916128575802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9611614108085632, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15160, + "weighted_flow_matching_loss": 0.026088137179613113, + "weighted_rkd_loss": 0.023710820823907852 + }, + { + "epoch": 4.413732906604597, + "flow_matching_loss": 0.02697605360299349, + "grad_norm": 0.2671060562133789, + "learning_rate": 2.952369695762131e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011009801644831896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004857702739536762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02687730547040701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9611108064651489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15170, + "weighted_flow_matching_loss": 0.02697605360299349, + "weighted_rkd_loss": 0.025832116045057772 + }, + { + "epoch": 4.416642420715741, + "flow_matching_loss": 0.027469544671475887, + "grad_norm": 0.23221498727798462, + "learning_rate": 2.9522662997551614e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010747040435671807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004890504223294556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026384585350751878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9610601961612701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15180, + "weighted_flow_matching_loss": 0.027469544671475887, + "weighted_rkd_loss": 0.02535723503679037 + }, + { + "epoch": 4.419551934826884, + "flow_matching_loss": 0.024461352080106736, + "grad_norm": 0.33974841237068176, + "learning_rate": 2.9521627934581503e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010659135039895773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004793601552955806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02611187193542719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9610095322132111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15190, + "weighted_flow_matching_loss": 0.024461352080106736, + "weighted_rkd_loss": 0.0250937320291996 + }, + { + "epoch": 4.422461448938027, + "flow_matching_loss": 0.024998693726956844, + "grad_norm": 0.28665831685066223, + "learning_rate": 2.9520591768789585e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010794989392161369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004587030969560146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026177009753882884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9609588205814361, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15200, + "weighted_flow_matching_loss": 0.024998693726956844, + "weighted_rkd_loss": 0.025155031122267246 + }, + { + "epoch": 4.425370963049171, + "flow_matching_loss": 0.026079923287034033, + "grad_norm": 0.3394307792186737, + "learning_rate": 2.951955450025455e-05, + "loss": 0.0524, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011185329360887408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005002045631408691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02737270463258028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9609081029891968, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15210, + "weighted_flow_matching_loss": 0.026079923287034033, + "weighted_rkd_loss": 0.026302674040198325 + }, + { + "epoch": 4.428280477160314, + "flow_matching_loss": 0.02327432408928871, + "grad_norm": 0.33063358068466187, + "learning_rate": 2.9518516129055172e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011094063147902488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004833369329571724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0270214956253767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.960857343673706, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15220, + "weighted_flow_matching_loss": 0.02327432408928871, + "weighted_rkd_loss": 0.02596381399780512 + }, + { + "epoch": 4.431189991271458, + "flow_matching_loss": 0.02601812556385994, + "grad_norm": 0.30820921063423157, + "learning_rate": 2.9517476655270306e-05, + "loss": 0.0521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01113034812733531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00485813154373318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027118827775120737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9608065605163574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15230, + "weighted_flow_matching_loss": 0.02601812556385994, + "weighted_rkd_loss": 0.026055950298905374 + }, + { + "epoch": 4.434099505382601, + "flow_matching_loss": 0.027146417647600174, + "grad_norm": 0.3025723993778229, + "learning_rate": 2.9516436078978893e-05, + "loss": 0.0539, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011390624567866325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050699755549430845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02785122487694025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9607557356357574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15240, + "weighted_flow_matching_loss": 0.027146417647600174, + "weighted_rkd_loss": 0.026758250221610068 + }, + { + "epoch": 4.437009019493744, + "flow_matching_loss": 0.026373401656746863, + "grad_norm": 0.38666024804115295, + "learning_rate": 2.9515394400259967e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011220530327409505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049711090046912435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02741216979920864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9607048809528351, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15250, + "weighted_flow_matching_loss": 0.026373401656746863, + "weighted_rkd_loss": 0.026335006579756737 + }, + { + "epoch": 4.439918533604888, + "flow_matching_loss": 0.025273259542882443, + "grad_norm": 0.2942643165588379, + "learning_rate": 2.9514351619192627e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0110141696408391, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005107351276092231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027135690674185753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9606540024280548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15260, + "weighted_flow_matching_loss": 0.025273259542882443, + "weighted_rkd_loss": 0.026067986711859704 + }, + { + "epoch": 4.442828047716032, + "flow_matching_loss": 0.027114429324865342, + "grad_norm": 0.28066545724868774, + "learning_rate": 2.9513307735856065e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011148848105221987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005174346477724611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02747204266488552, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9606030881404877, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15270, + "weighted_flow_matching_loss": 0.027114429324865342, + "weighted_rkd_loss": 0.026389720663428307 + }, + { + "epoch": 4.445737561827175, + "flow_matching_loss": 0.02528537381440401, + "grad_norm": 0.28429463505744934, + "learning_rate": 2.9512262750329562e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010994342900812627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005076246103271842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02706493213772774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9605521559715271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15280, + "weighted_flow_matching_loss": 0.02528537381440401, + "weighted_rkd_loss": 0.025997281633317472 + }, + { + "epoch": 4.448647075938318, + "flow_matching_loss": 0.024725789576768874, + "grad_norm": 0.29194194078445435, + "learning_rate": 2.9511216662692476e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011481076758354903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005025897664017976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027988051250576972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9605011641979218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15290, + "weighted_flow_matching_loss": 0.024725789576768874, + "weighted_rkd_loss": 0.02688259966671467 + }, + { + "epoch": 4.451556590049462, + "flow_matching_loss": 0.024473156221210955, + "grad_norm": 0.3470372259616852, + "learning_rate": 2.9510169473024248e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010841885209083557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005043702106922865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02672747243195772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9604501545429229, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15300, + "weighted_flow_matching_loss": 0.024473156221210955, + "weighted_rkd_loss": 0.02567035350948572 + }, + { + "epoch": 4.4544661041606055, + "flow_matching_loss": 0.026735103316605092, + "grad_norm": 0.320624977350235, + "learning_rate": 2.9509121181404414e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01097162002697587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005065091513097286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027008331939578058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9603991210460663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15310, + "weighted_flow_matching_loss": 0.026735103316605092, + "weighted_rkd_loss": 0.02593876626342535 + }, + { + "epoch": 4.457375618271748, + "flow_matching_loss": 0.025476276502013208, + "grad_norm": 0.27576392889022827, + "learning_rate": 2.950807178791257e-05, + "loss": 0.0522, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011455370299518108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00488340430893004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027794145047664642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9603480458259582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15320, + "weighted_flow_matching_loss": 0.025476276502013208, + "weighted_rkd_loss": 0.026692065037786962 + }, + { + "epoch": 4.460285132382892, + "flow_matching_loss": 0.027183234319090845, + "grad_norm": 0.2776589095592499, + "learning_rate": 2.950702129262842e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011148070078343153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005176942003890872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027473081834614277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9602969408035278, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15330, + "weighted_flow_matching_loss": 0.027183234319090845, + "weighted_rkd_loss": 0.02638230100274086 + }, + { + "epoch": 4.463194646494036, + "flow_matching_loss": 0.027288996055722235, + "grad_norm": 0.3258931636810303, + "learning_rate": 2.9505969695631744e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010252745682373643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004839592566713691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025345084071159363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9602458059787751, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15340, + "weighted_flow_matching_loss": 0.027288996055722235, + "weighted_rkd_loss": 0.02433752827346325 + }, + { + "epoch": 4.466104160605179, + "flow_matching_loss": 0.025365937501192093, + "grad_norm": 0.30189618468284607, + "learning_rate": 2.95049169970024e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011027637589722871, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051170675549656154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027172342501580717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9601946353912354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15350, + "weighted_flow_matching_loss": 0.025365937501192093, + "weighted_rkd_loss": 0.02609073407948017 + }, + { + "epoch": 4.469013674716322, + "flow_matching_loss": 0.02644382882863283, + "grad_norm": 0.3281659781932831, + "learning_rate": 2.9503863196820332e-05, + "loss": 0.0504, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010116038471460342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004716272628866136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024948349595069884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9601434409618378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15360, + "weighted_flow_matching_loss": 0.02644382882863283, + "weighted_rkd_loss": 0.023953962326049804 + }, + { + "epoch": 4.471923188827466, + "flow_matching_loss": 0.025176119059324265, + "grad_norm": 0.3028617203235626, + "learning_rate": 2.9502808295165575e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011244086362421513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004863743577152491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027351916767656803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9600921869277954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15370, + "weighted_flow_matching_loss": 0.025176119059324265, + "weighted_rkd_loss": 0.026260350830852985 + }, + { + "epoch": 4.47483270293861, + "flow_matching_loss": 0.02713526077568531, + "grad_norm": 0.2797560393810272, + "learning_rate": 2.9501752292118236e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01034505721181631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00490170563571155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02559181973338127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9600409269332886, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15380, + "weighted_flow_matching_loss": 0.02713526077568531, + "weighted_rkd_loss": 0.024569215066730977 + }, + { + "epoch": 4.4777422170497525, + "flow_matching_loss": 0.023546452261507513, + "grad_norm": 0.30936357378959656, + "learning_rate": 2.9500695187758514e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010856041219085455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005057326378300786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026769408769905568, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9599896550178528, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15390, + "weighted_flow_matching_loss": 0.023546452261507513, + "weighted_rkd_loss": 0.025698342733085154 + }, + { + "epoch": 4.480651731160896, + "flow_matching_loss": 0.025681373476982117, + "grad_norm": 0.2959950268268585, + "learning_rate": 2.949963698216669e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01071741571649909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004923956235870719, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02635878771543503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9599383294582366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15400, + "weighted_flow_matching_loss": 0.025681373476982117, + "weighted_rkd_loss": 0.025302793271839617 + }, + { + "epoch": 4.48356124527204, + "flow_matching_loss": 0.0258448239415884, + "grad_norm": 0.2959859073162079, + "learning_rate": 2.9498577675423123e-05, + "loss": 0.0533, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011538548022508621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005485990038141608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028563085570931436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9598869740962982, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15410, + "weighted_flow_matching_loss": 0.0258448239415884, + "weighted_rkd_loss": 0.02741732057183981 + }, + { + "epoch": 4.486470759383183, + "flow_matching_loss": 0.028344218619167805, + "grad_norm": 0.26491329073905945, + "learning_rate": 2.9497517267608262e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010907476767897606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005065451748669148, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026880405098199844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9598355889320374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15420, + "weighted_flow_matching_loss": 0.028344218619167805, + "weighted_rkd_loss": 0.025800774246454238 + }, + { + "epoch": 4.4893802734943264, + "flow_matching_loss": 0.02491443008184433, + "grad_norm": 0.3065016567707062, + "learning_rate": 2.9496455758802646e-05, + "loss": 0.0508, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010921238083392381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005164397461339831, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027006873674690724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9597841680049897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15430, + "weighted_flow_matching_loss": 0.02491443008184433, + "weighted_rkd_loss": 0.025920776650309563 + }, + { + "epoch": 4.49228978760547, + "flow_matching_loss": 0.02549192067235708, + "grad_norm": 0.37373870611190796, + "learning_rate": 2.9495393149086882e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010605289926752448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00483968851622194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026050267927348613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9597327232360839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15440, + "weighted_flow_matching_loss": 0.02549192067235708, + "weighted_rkd_loss": 0.025001301988959314 + }, + { + "epoch": 4.495199301716613, + "flow_matching_loss": 0.02697901502251625, + "grad_norm": 0.2772229015827179, + "learning_rate": 2.949432943854167e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011596686020493508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053506689611822365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028544041328132153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.959681236743927, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15450, + "weighted_flow_matching_loss": 0.02697901502251625, + "weighted_rkd_loss": 0.027393182367086412 + }, + { + "epoch": 4.498108815827757, + "flow_matching_loss": 0.024730242416262628, + "grad_norm": 0.42336708307266235, + "learning_rate": 2.949326462724779e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01133175464347005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005233735241927206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027897244319319725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9596297204494476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15460, + "weighted_flow_matching_loss": 0.024730242416262628, + "weighted_rkd_loss": 0.02677098345011473 + }, + { + "epoch": 4.5010183299389, + "flow_matching_loss": 0.025980211049318313, + "grad_norm": 0.2795815169811249, + "learning_rate": 2.949219871528611e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010605028178542853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004833551496267319, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026043608039617538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9595781862735748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15470, + "weighted_flow_matching_loss": 0.025980211049318313, + "weighted_rkd_loss": 0.024990873783826827 + }, + { + "epoch": 4.503927844050043, + "flow_matching_loss": 0.026897961832582952, + "grad_norm": 0.2949189245700836, + "learning_rate": 2.9491131702737572e-05, + "loss": 0.0532, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011129014799371362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051811460172757505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02743917554616928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9595265924930573, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15480, + "weighted_flow_matching_loss": 0.026897961832582952, + "weighted_rkd_loss": 0.02632864937186241 + }, + { + "epoch": 4.506837358161187, + "flow_matching_loss": 0.026012043841183184, + "grad_norm": 0.2494649440050125, + "learning_rate": 2.9490063589683223e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010827623680233956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004954510601237416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026609758101403713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9594749867916107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15490, + "weighted_flow_matching_loss": 0.026012043841183184, + "weighted_rkd_loss": 0.025531398132443428 + }, + { + "epoch": 4.509746872272331, + "flow_matching_loss": 0.026873834431171417, + "grad_norm": 0.2941317856311798, + "learning_rate": 2.9488994376204166e-05, + "loss": 0.0511, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010148279089480638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004910439485684037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025206997618079187, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9594233453273773, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15500, + "weighted_flow_matching_loss": 0.026873834431171417, + "weighted_rkd_loss": 0.024184187315404414 + }, + { + "epoch": 4.512656386383474, + "flow_matching_loss": 0.02564951702952385, + "grad_norm": 0.2468189001083374, + "learning_rate": 2.9487924062381603e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010788000840693712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005313410703092813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026889412291347982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.959371680021286, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15510, + "weighted_flow_matching_loss": 0.02564951702952385, + "weighted_rkd_loss": 0.0257969506084919 + }, + { + "epoch": 4.515565900494617, + "flow_matching_loss": 0.026240949518978596, + "grad_norm": 0.3079967796802521, + "learning_rate": 2.9486852648296825e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010716667911037803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048310566693544384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026264392025768756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9593199789524078, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15520, + "weighted_flow_matching_loss": 0.026240949518978596, + "weighted_rkd_loss": 0.02519595269113779 + }, + { + "epoch": 4.518475414605761, + "flow_matching_loss": 0.026332886330783366, + "grad_norm": 0.2559441328048706, + "learning_rate": 2.9485780134031187e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011231792066246271, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052920761052519085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02775566019117832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9592682421207428, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15530, + "weighted_flow_matching_loss": 0.026332886330783366, + "weighted_rkd_loss": 0.026625166833400726 + }, + { + "epoch": 4.521384928716905, + "flow_matching_loss": 0.02529022265225649, + "grad_norm": 0.2613259553909302, + "learning_rate": 2.948470651966615e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00977450916543603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004483319912105798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024032338336110116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9592164814472198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15540, + "weighted_flow_matching_loss": 0.02529022265225649, + "weighted_rkd_loss": 0.02305223699659109 + }, + { + "epoch": 4.524294442828047, + "flow_matching_loss": 0.025037405267357826, + "grad_norm": 0.31957298517227173, + "learning_rate": 2.9483631805283244e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010435598623007535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004766581277363002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02563777845352888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.95916468501091, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15550, + "weighted_flow_matching_loss": 0.025037405267357826, + "weighted_rkd_loss": 0.024590842239558697 + }, + { + "epoch": 4.527203956939191, + "flow_matching_loss": 0.026500866934657096, + "grad_norm": 0.2736507058143616, + "learning_rate": 2.9482555990964084e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009649439062923194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004248572210781276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02354745026677847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9591128528118134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15560, + "weighted_flow_matching_loss": 0.026500866934657096, + "weighted_rkd_loss": 0.022584656067192555 + }, + { + "epoch": 4.530113471050335, + "flow_matching_loss": 0.026210481487214565, + "grad_norm": 0.2603415250778198, + "learning_rate": 2.948147907679037e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010843472927808762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048229342326521875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026509879902005196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9590609967708588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15570, + "weighted_flow_matching_loss": 0.026210481487214565, + "weighted_rkd_loss": 0.025424604862928392 + }, + { + "epoch": 4.5330229851614785, + "flow_matching_loss": 0.02592820767313242, + "grad_norm": 0.2769428491592407, + "learning_rate": 2.9480401062843894e-05, + "loss": 0.0504, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010421732300892472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004693929385393858, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025537393800914288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9590090990066529, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15580, + "weighted_flow_matching_loss": 0.02592820767313242, + "weighted_rkd_loss": 0.0244906036183238 + }, + { + "epoch": 4.535932499272621, + "flow_matching_loss": 0.023042229004204273, + "grad_norm": 0.29994598031044006, + "learning_rate": 2.947932194920652e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010434780828654766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00459570528473705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025465266779065132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9589571833610535, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15590, + "weighted_flow_matching_loss": 0.023042229004204273, + "weighted_rkd_loss": 0.024420098587870597 + }, + { + "epoch": 4.538842013383765, + "flow_matching_loss": 0.02588649969547987, + "grad_norm": 0.26937827467918396, + "learning_rate": 2.9478241735960196e-05, + "loss": 0.0508, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010597233613952994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004788181209005415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025982648693025114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9589052200317383, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15600, + "weighted_flow_matching_loss": 0.02588649969547987, + "weighted_rkd_loss": 0.024914898537099363 + }, + { + "epoch": 4.541751527494909, + "flow_matching_loss": 0.025728535652160645, + "grad_norm": 0.2946641147136688, + "learning_rate": 2.9477160423186964e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010304382722824811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004343778500333429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024952543899416924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9588532388210297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15610, + "weighted_flow_matching_loss": 0.025728535652160645, + "weighted_rkd_loss": 0.023925792053341867 + }, + { + "epoch": 4.5446610416060516, + "flow_matching_loss": 0.024887623637914656, + "grad_norm": 0.3037256598472595, + "learning_rate": 2.9476078010968938e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010507770534604787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004661931190639734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02567747253924608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9588012218475341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15620, + "weighted_flow_matching_loss": 0.024887623637914656, + "weighted_rkd_loss": 0.02461959086358547 + }, + { + "epoch": 4.547570555717195, + "flow_matching_loss": 0.026025267504155635, + "grad_norm": 0.29927924275398254, + "learning_rate": 2.947499449938832e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011205444671213627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005287370644509792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027698259986937045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9587491750717163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15630, + "weighted_flow_matching_loss": 0.026025267504155635, + "weighted_rkd_loss": 0.026555723510682583 + }, + { + "epoch": 4.550480069828339, + "flow_matching_loss": 0.027915033511817455, + "grad_norm": 0.3076566755771637, + "learning_rate": 2.9473909888527394e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010404142551124096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004660930251702666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025469215400516988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9586970925331115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15640, + "weighted_flow_matching_loss": 0.027915033511817455, + "weighted_rkd_loss": 0.02441727090626955 + }, + { + "epoch": 4.553389583939482, + "flow_matching_loss": 0.024963701702654363, + "grad_norm": 0.3178195357322693, + "learning_rate": 2.9472824178468533e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01047347392886877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004708195920102298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025655143707990647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9586449801921845, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15650, + "weighted_flow_matching_loss": 0.024963701702654363, + "weighted_rkd_loss": 0.024594176560640335 + }, + { + "epoch": 4.5562990980506255, + "flow_matching_loss": 0.025926905497908593, + "grad_norm": 0.24329763650894165, + "learning_rate": 2.947173736929419e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011377224791795015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005408611567690969, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0281630614772439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9585928320884705, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15660, + "weighted_flow_matching_loss": 0.025926905497908593, + "weighted_rkd_loss": 0.02699693590402603 + }, + { + "epoch": 4.559208612161769, + "flow_matching_loss": 0.024931035935878754, + "grad_norm": 0.2740671634674072, + "learning_rate": 2.9470649461086897e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010104196704924106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004675830830819905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024884224869310856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9585406661033631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15670, + "weighted_flow_matching_loss": 0.024931035935878754, + "weighted_rkd_loss": 0.02385255992412567 + }, + { + "epoch": 4.562118126272912, + "flow_matching_loss": 0.025631052255630494, + "grad_norm": 0.29117435216903687, + "learning_rate": 2.9469560453929277e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009924246137961746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046333424281328915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02448183484375477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9584884583950043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15680, + "weighted_flow_matching_loss": 0.025631052255630494, + "weighted_rkd_loss": 0.023465533554553986 + }, + { + "epoch": 4.565027640384056, + "flow_matching_loss": 0.02545277327299118, + "grad_norm": 0.3138757646083832, + "learning_rate": 2.946847034790403e-05, + "loss": 0.0511, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010731677804142237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005309821013361216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026773176342248916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9584362208843231, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15690, + "weighted_flow_matching_loss": 0.02545277327299118, + "weighted_rkd_loss": 0.025660387054085732 + }, + { + "epoch": 4.567937154495199, + "flow_matching_loss": 0.0249929241836071, + "grad_norm": 0.31982630491256714, + "learning_rate": 2.9467379143093942e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011011785361915827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050281349569559096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02705170549452305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9583839535713196, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15700, + "weighted_flow_matching_loss": 0.0249929241836071, + "weighted_rkd_loss": 0.02592591233551502 + }, + { + "epoch": 4.570846668606343, + "flow_matching_loss": 0.025784294307231902, + "grad_norm": 0.25285202264785767, + "learning_rate": 2.9466286839581888e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011110297404229641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004753663949668407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026974258385598658, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9583316504955292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15710, + "weighted_flow_matching_loss": 0.025784294307231902, + "weighted_rkd_loss": 0.02585028614848852 + }, + { + "epoch": 4.573756182717486, + "flow_matching_loss": 0.025321221724152566, + "grad_norm": 0.2561205327510834, + "learning_rate": 2.9465193437450817e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011173745431005955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005014031194150448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02736152205616236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9582793235778808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15720, + "weighted_flow_matching_loss": 0.025321221724152566, + "weighted_rkd_loss": 0.026220013573765754 + }, + { + "epoch": 4.57666569682863, + "flow_matching_loss": 0.027116508781909944, + "grad_norm": 0.3143163025379181, + "learning_rate": 2.9464098936783765e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011156274611130357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005323897022753954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02763644643127918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9582269549369812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15730, + "weighted_flow_matching_loss": 0.027116508781909944, + "weighted_rkd_loss": 0.02648200523108244 + }, + { + "epoch": 4.579575210939773, + "flow_matching_loss": 0.024270927906036376, + "grad_norm": 0.27197232842445374, + "learning_rate": 2.946300333766385e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009859938267618417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004431473952718079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024151350557804107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9581745684146881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15740, + "weighted_flow_matching_loss": 0.024270927906036376, + "weighted_rkd_loss": 0.02314122524112463 + }, + { + "epoch": 4.582484725050916, + "flow_matching_loss": 0.025255798175930978, + "grad_norm": 0.3232405185699463, + "learning_rate": 2.9461906640174287e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010097329877316952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046957027399912475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024890362471342086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9581221342086792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15750, + "weighted_flow_matching_loss": 0.025255798175930978, + "weighted_rkd_loss": 0.02384800147265196 + }, + { + "epoch": 4.58539423916206, + "flow_matching_loss": 0.023146138340234757, + "grad_norm": 0.285480797290802, + "learning_rate": 2.9460808844398352e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010332823637872935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004522365401498973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02518801260739565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9580696821212769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15760, + "weighted_flow_matching_loss": 0.023146138340234757, + "weighted_rkd_loss": 0.024131873436272145 + }, + { + "epoch": 4.588303753273204, + "flow_matching_loss": 0.02230694191530347, + "grad_norm": 0.2588960826396942, + "learning_rate": 2.9459709950419417e-05, + "loss": 0.0453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009721979964524508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004588532214984298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024032492376863956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9580171883106232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15770, + "weighted_flow_matching_loss": 0.02230694191530347, + "weighted_rkd_loss": 0.023023544624447823 + }, + { + "epoch": 4.591213267384346, + "flow_matching_loss": 0.02627000380307436, + "grad_norm": 0.29434531927108765, + "learning_rate": 2.9458609958320938e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009445283701643348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041465276386588815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023037095181643964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9579646766185761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15780, + "weighted_flow_matching_loss": 0.02627000380307436, + "weighted_rkd_loss": 0.02206872496753931 + }, + { + "epoch": 4.59412278149549, + "flow_matching_loss": 0.02646268643438816, + "grad_norm": 0.3370545506477356, + "learning_rate": 2.945750886818645e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01063308296725154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004538534535095095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025804700516164302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9579121172428131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15790, + "weighted_flow_matching_loss": 0.02646268643438816, + "weighted_rkd_loss": 0.024718635715544224 + }, + { + "epoch": 4.597032295606634, + "flow_matching_loss": 0.023923393711447714, + "grad_norm": 0.29927921295166016, + "learning_rate": 2.9456406680099577e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010588153172284365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004692280315794051, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025868586823344232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9578595280647277, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15800, + "weighted_flow_matching_loss": 0.023923393711447714, + "weighted_rkd_loss": 0.024778442084789278 + }, + { + "epoch": 4.599941809717777, + "flow_matching_loss": 0.02565068770200014, + "grad_norm": 0.2506061792373657, + "learning_rate": 2.9455303394144014e-05, + "loss": 0.0482, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009652621671557426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004254084639251232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02355932854115963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.957806921005249, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15810, + "weighted_flow_matching_loss": 0.02565068770200014, + "weighted_rkd_loss": 0.02256527915596962 + }, + { + "epoch": 4.60285132382892, + "flow_matching_loss": 0.025053959898650648, + "grad_norm": 0.33116573095321655, + "learning_rate": 2.9454199010403564e-05, + "loss": 0.0499, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010569989494979382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004844020470045507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02598399929702282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9577542841434479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15820, + "weighted_flow_matching_loss": 0.025053959898650648, + "weighted_rkd_loss": 0.0248862911015749 + }, + { + "epoch": 4.605760837940064, + "flow_matching_loss": 0.023554767668247222, + "grad_norm": 0.2741450071334839, + "learning_rate": 2.9453093528962082e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010002433136105538, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00456116502173245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02456603180617094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9577015995979309, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15830, + "weighted_flow_matching_loss": 0.023554767668247222, + "weighted_rkd_loss": 0.02352691814303398 + }, + { + "epoch": 4.608670352051208, + "flow_matching_loss": 0.026753751933574675, + "grad_norm": 0.28504812717437744, + "learning_rate": 2.945198694990353e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011116931680589914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005114653590135276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027348517067730428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9576488912105561, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15840, + "weighted_flow_matching_loss": 0.026753751933574675, + "weighted_rkd_loss": 0.02619025968015194 + }, + { + "epoch": 4.611579866162351, + "flow_matching_loss": 0.024103691056370734, + "grad_norm": 0.274064838886261, + "learning_rate": 2.9450879273311947e-05, + "loss": 0.0493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010644595604389906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005040330323390662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02632952146232128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9575961530208588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15850, + "weighted_flow_matching_loss": 0.024103691056370734, + "weighted_rkd_loss": 0.02521302867680788 + }, + { + "epoch": 4.614489380273494, + "flow_matching_loss": 0.0273085443302989, + "grad_norm": 0.2976329028606415, + "learning_rate": 2.9449770499271446e-05, + "loss": 0.0521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010660584270954131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004519395157694816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02584056369960308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9575433850288391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15860, + "weighted_flow_matching_loss": 0.0273085443302989, + "weighted_rkd_loss": 0.02474344987422228 + }, + { + "epoch": 4.617398894384638, + "flow_matching_loss": 0.02430966217070818, + "grad_norm": 0.2765113413333893, + "learning_rate": 2.944866062786624e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010232770349830389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004610839951783419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025076380744576455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9574905753135681, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15870, + "weighted_flow_matching_loss": 0.02430966217070818, + "weighted_rkd_loss": 0.024010361358523368 + }, + { + "epoch": 4.620308408495781, + "flow_matching_loss": 0.02693540006875992, + "grad_norm": 0.27707967162132263, + "learning_rate": 2.9447549659180612e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010347480000928045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048101508058607575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025505110435187817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9574377536773682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15880, + "weighted_flow_matching_loss": 0.02693540006875992, + "weighted_rkd_loss": 0.024419588968157768 + }, + { + "epoch": 4.6232179226069245, + "flow_matching_loss": 0.02670214679092169, + "grad_norm": 0.2400524914264679, + "learning_rate": 2.944643759329893e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010130248125642538, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004652099497616291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024912595748901367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9573848783969879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15890, + "weighted_flow_matching_loss": 0.02670214679092169, + "weighted_rkd_loss": 0.02385091558098793 + }, + { + "epoch": 4.626127436718068, + "flow_matching_loss": 0.02654460184276104, + "grad_norm": 0.34773963689804077, + "learning_rate": 2.9445324430305656e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010078268777579068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004595407680608332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024751945212483407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9573319852352142, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15900, + "weighted_flow_matching_loss": 0.02654460184276104, + "weighted_rkd_loss": 0.023695814609527587 + }, + { + "epoch": 4.629036950829212, + "flow_matching_loss": 0.026099560782313347, + "grad_norm": 0.2608988285064697, + "learning_rate": 2.9444210170285317e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010860165394842625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004713333980180323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026433664374053478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9572790563106537, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15910, + "weighted_flow_matching_loss": 0.026099560782313347, + "weighted_rkd_loss": 0.025304413586854934 + }, + { + "epoch": 4.631946464940355, + "flow_matching_loss": 0.0255017651244998, + "grad_norm": 0.23869988322257996, + "learning_rate": 2.9443094813322543e-05, + "loss": 0.0485, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00982910329475999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004358388343825936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024016594886779784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9572260975837708, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15920, + "weighted_flow_matching_loss": 0.0255017651244998, + "weighted_rkd_loss": 0.022989307716488838 + }, + { + "epoch": 4.6348559790514985, + "flow_matching_loss": 0.024589270167052747, + "grad_norm": 0.2720361053943634, + "learning_rate": 2.944197835950203e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010422848910093308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004746063146740198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02559176106005907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9571731090545654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15930, + "weighted_flow_matching_loss": 0.024589270167052747, + "weighted_rkd_loss": 0.024495754577219486 + }, + { + "epoch": 4.637765493162642, + "flow_matching_loss": 0.026693664491176605, + "grad_norm": 0.344223290681839, + "learning_rate": 2.944086080890857e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010269747301936149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004819564754143357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02535905912518501, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9571200907230377, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15940, + "weighted_flow_matching_loss": 0.026693664491176605, + "weighted_rkd_loss": 0.024271602742373943 + }, + { + "epoch": 4.640675007273785, + "flow_matching_loss": 0.026468516141176224, + "grad_norm": 0.2658344507217407, + "learning_rate": 2.9439742161627036e-05, + "loss": 0.0499, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009991102200001479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004519007331691682, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024501211754977704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9570670425891876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15950, + "weighted_flow_matching_loss": 0.026468516141176224, + "weighted_rkd_loss": 0.02344929352402687 + }, + { + "epoch": 4.643584521384929, + "flow_matching_loss": 0.027744814939796924, + "grad_norm": 0.2488730251789093, + "learning_rate": 2.9438622417742375e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010867951530963182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005170074291527271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026905977353453635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9570139467716217, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15960, + "weighted_flow_matching_loss": 0.027744814939796924, + "weighted_rkd_loss": 0.0257494131103158 + }, + { + "epoch": 4.646494035496072, + "flow_matching_loss": 0.025098148733377457, + "grad_norm": 0.3762645721435547, + "learning_rate": 2.9437501577339632e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010909959115087986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005181003618054092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027000921592116357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9569608330726623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15970, + "weighted_flow_matching_loss": 0.025098148733377457, + "weighted_rkd_loss": 0.02583878394216299 + }, + { + "epoch": 4.649403549607215, + "flow_matching_loss": 0.028471183590590955, + "grad_norm": 0.24219217896461487, + "learning_rate": 2.9436379640503923e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010379802342504263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004490079660899937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025249684043228627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9569076895713806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15980, + "weighted_flow_matching_loss": 0.028471183590590955, + "weighted_rkd_loss": 0.024161605350673198 + }, + { + "epoch": 4.652313063718359, + "flow_matching_loss": 0.026258095167577267, + "grad_norm": 0.31075116991996765, + "learning_rate": 2.943525660732045e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009952051937580109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004667556588537991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024571660533547403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9568545043468475, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15990, + "weighted_flow_matching_loss": 0.026258095167577267, + "weighted_rkd_loss": 0.023511520214378835 + }, + { + "epoch": 4.655222577829503, + "flow_matching_loss": 0.024826092831790447, + "grad_norm": 0.2605994641780853, + "learning_rate": 2.94341324778745e-05, + "loss": 0.0494, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010554155521094798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004547221236862242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02565553244203329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9568012952804565, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16000, + "weighted_flow_matching_loss": 0.024826092831790447, + "weighted_rkd_loss": 0.02454721983522177 + }, + { + "epoch": 4.6581320919406455, + "flow_matching_loss": 0.02395418509840965, + "grad_norm": 0.20905011892318726, + "learning_rate": 2.943300725225145e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009851002041250468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00443349180277437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024135495908558367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9567480504512786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16010, + "weighted_flow_matching_loss": 0.02395418509840965, + "weighted_rkd_loss": 0.023091642744839193 + }, + { + "epoch": 4.661041606051789, + "flow_matching_loss": 0.026014064066112043, + "grad_norm": 0.2956032156944275, + "learning_rate": 2.943188093053675e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010296665225178003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004577002534642815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025170333124697207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9566947877407074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16020, + "weighted_flow_matching_loss": 0.026014064066112043, + "weighted_rkd_loss": 0.024080315604805946 + }, + { + "epoch": 4.663951120162933, + "flow_matching_loss": 0.02653152197599411, + "grad_norm": 0.3170897364616394, + "learning_rate": 2.9430753512815926e-05, + "loss": 0.0508, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010320007614791394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004700944456271827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025340959802269934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9566414773464202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16030, + "weighted_flow_matching_loss": 0.02653152197599411, + "weighted_rkd_loss": 0.02424218226224184 + }, + { + "epoch": 4.666860634274077, + "flow_matching_loss": 0.024982247687876224, + "grad_norm": 0.22512121498584747, + "learning_rate": 2.942962499917462e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00979337263852358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004428763850592076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024015508778393268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9565881490707397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16040, + "weighted_flow_matching_loss": 0.024982247687876224, + "weighted_rkd_loss": 0.022972950339317323 + }, + { + "epoch": 4.669770148385219, + "flow_matching_loss": 0.02420241739600897, + "grad_norm": 0.2537977397441864, + "learning_rate": 2.9428495389698515e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010580621100962162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004970135749317706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026131377927958964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9565347790718078, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16050, + "weighted_flow_matching_loss": 0.02420241739600897, + "weighted_rkd_loss": 0.024995556846261025 + }, + { + "epoch": 4.672679662496363, + "flow_matching_loss": 0.024609540589153767, + "grad_norm": 0.23506441712379456, + "learning_rate": 2.9427364684473403e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009836670802906155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004438830842263997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024112172424793243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9564813792705535, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16060, + "weighted_flow_matching_loss": 0.024609540589153767, + "weighted_rkd_loss": 0.023062840662896635 + }, + { + "epoch": 4.675589176607507, + "flow_matching_loss": 0.025072935968637466, + "grad_norm": 0.31600600481033325, + "learning_rate": 2.942623288358516e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00975250662304461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004240756249055266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023745769262313844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9564279615879059, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16070, + "weighted_flow_matching_loss": 0.025072935968637466, + "weighted_rkd_loss": 0.022711095027625562 + }, + { + "epoch": 4.67849869071865, + "flow_matching_loss": 0.026234292052686216, + "grad_norm": 0.3016827404499054, + "learning_rate": 2.9425099987119733e-05, + "loss": 0.0493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00986759182997048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004351808573119342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02408699207007885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9563744902610779, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16080, + "weighted_flow_matching_loss": 0.026234292052686216, + "weighted_rkd_loss": 0.02303616050630808 + }, + { + "epoch": 4.681408204829793, + "flow_matching_loss": 0.025975288823246956, + "grad_norm": 0.3197636902332306, + "learning_rate": 2.942396599516316e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009561976185068488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00436493456363678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023488887026906012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9563210010528564, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16090, + "weighted_flow_matching_loss": 0.025975288823246956, + "weighted_rkd_loss": 0.02246288750320673 + }, + { + "epoch": 4.684317718940937, + "flow_matching_loss": 0.025739990174770355, + "grad_norm": 0.3076944649219513, + "learning_rate": 2.942283090780156e-05, + "loss": 0.0511, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010852484498172998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004833088628947735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026538057997822762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9562674760818481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16100, + "weighted_flow_matching_loss": 0.025739990174770355, + "weighted_rkd_loss": 0.025377508625388144 + }, + { + "epoch": 4.687227233052081, + "flow_matching_loss": 0.024349621310830117, + "grad_norm": 0.26995041966438293, + "learning_rate": 2.9421694725121132e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010693519841879606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049804886104539035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026367528177797795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9562139272689819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16110, + "weighted_flow_matching_loss": 0.024349621310830117, + "weighted_rkd_loss": 0.025212997570633888 + }, + { + "epoch": 4.690136747163224, + "flow_matching_loss": 0.02684119250625372, + "grad_norm": 0.30064505338668823, + "learning_rate": 2.9420557447208165e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009890896454453468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004502529907040298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02428432274609804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9561603426933288, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16120, + "weighted_flow_matching_loss": 0.02684119250625372, + "weighted_rkd_loss": 0.023219664208590986 + }, + { + "epoch": 4.693046261274367, + "flow_matching_loss": 0.023918944969773294, + "grad_norm": 0.28241226077079773, + "learning_rate": 2.9419419074149027e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008950183633714914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004062049789354205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021962417103350163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9561067223548889, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16130, + "weighted_flow_matching_loss": 0.023918944969773294, + "weighted_rkd_loss": 0.02099840100854635 + }, + { + "epoch": 4.69595577538551, + "flow_matching_loss": 0.026407055929303168, + "grad_norm": 0.24119921028614044, + "learning_rate": 2.9418279606030175e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009189164731651545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004225851828232407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022604181431233883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9560530781745911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16140, + "weighted_flow_matching_loss": 0.026407055929303168, + "weighted_rkd_loss": 0.021610809303820134 + }, + { + "epoch": 4.698865289496654, + "flow_matching_loss": 0.024948425590991974, + "grad_norm": 0.2919919192790985, + "learning_rate": 2.9417139042938136e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009968395717442036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004473353014327585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02441014423966408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9559993982315064, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16150, + "weighted_flow_matching_loss": 0.024948425590991974, + "weighted_rkd_loss": 0.023336102068424226 + }, + { + "epoch": 4.7017748036077975, + "flow_matching_loss": 0.02524760365486145, + "grad_norm": 0.2957918345928192, + "learning_rate": 2.941599738495953e-05, + "loss": 0.0479, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009630474075675011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004435513354837895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023696461319923402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9559456884860993, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16160, + "weighted_flow_matching_loss": 0.02524760365486145, + "weighted_rkd_loss": 0.02265253644436598 + }, + { + "epoch": 4.704684317718941, + "flow_matching_loss": 0.024899196438491346, + "grad_norm": 0.32055574655532837, + "learning_rate": 2.9414854632181064e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010818489640951157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004964726977050304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026601706631481647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9558919429779053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16170, + "weighted_flow_matching_loss": 0.024899196438491346, + "weighted_rkd_loss": 0.025428365729749203 + }, + { + "epoch": 4.707593831830084, + "flow_matching_loss": 0.026728720776736735, + "grad_norm": 0.24748216569423676, + "learning_rate": 2.9413710784689516e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009853263478726148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004389528627507388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024096055328845976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9558381736278534, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16180, + "weighted_flow_matching_loss": 0.026728720776736735, + "weighted_rkd_loss": 0.023031947016716004 + }, + { + "epoch": 4.710503345941228, + "flow_matching_loss": 0.02559161316603422, + "grad_norm": 0.28664469718933105, + "learning_rate": 2.9412565842571758e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010147153120487928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00436994805932045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024664253927767277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9557843685150147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16190, + "weighted_flow_matching_loss": 0.02559161316603422, + "weighted_rkd_loss": 0.023573706671595574 + }, + { + "epoch": 4.7134128600523715, + "flow_matching_loss": 0.024818651005625726, + "grad_norm": 0.2588174045085907, + "learning_rate": 2.9411419805914735e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01020197095349431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004479640140198171, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02488358225673437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.955730527639389, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16200, + "weighted_flow_matching_loss": 0.024818651005625726, + "weighted_rkd_loss": 0.02378201987594366 + }, + { + "epoch": 4.716322374163514, + "flow_matching_loss": 0.024727688916027547, + "grad_norm": 0.31249088048934937, + "learning_rate": 2.9410272674805486e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010809545684605838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004925120016559958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026544211618602275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9556766629219056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16210, + "weighted_flow_matching_loss": 0.024727688916027547, + "weighted_rkd_loss": 0.0253677099943161 + }, + { + "epoch": 4.719231888274658, + "flow_matching_loss": 0.026724160835146903, + "grad_norm": 0.29539766907691956, + "learning_rate": 2.9409124449331126e-05, + "loss": 0.0511, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01039591645821929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046939478488639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02548578064888716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9556227684020996, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16220, + "weighted_flow_matching_loss": 0.026724160835146903, + "weighted_rkd_loss": 0.024354812502861024 + }, + { + "epoch": 4.722141402385802, + "flow_matching_loss": 0.027106250077486037, + "grad_norm": 0.2568419873714447, + "learning_rate": 2.940797512957886e-05, + "loss": 0.0511, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01019025631248951, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004740225151181221, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02512073777616024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9555688500404358, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16230, + "weighted_flow_matching_loss": 0.027106250077486037, + "weighted_rkd_loss": 0.024004605039954184 + }, + { + "epoch": 4.725050916496945, + "flow_matching_loss": 0.025373581796884537, + "grad_norm": 0.25619903206825256, + "learning_rate": 2.9406824715635965e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010434186924248934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004659656365402043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02552803009748459, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9555148780345917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16240, + "weighted_flow_matching_loss": 0.025373581796884537, + "weighted_rkd_loss": 0.024392406456172466 + }, + { + "epoch": 4.727960430608088, + "flow_matching_loss": 0.02776153329759836, + "grad_norm": 0.26409146189689636, + "learning_rate": 2.940567320758981e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010687981639057398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004875447973608971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026251411251723766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9554608881473541, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16250, + "weighted_flow_matching_loss": 0.02776153329759836, + "weighted_rkd_loss": 0.025082228519022465 + }, + { + "epoch": 4.730869944719232, + "flow_matching_loss": 0.025835301727056503, + "grad_norm": 0.37358999252319336, + "learning_rate": 2.9404520605527843e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010183892399072646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044726826250553135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024840467609465122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9554068624973298, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16260, + "weighted_flow_matching_loss": 0.025835301727056503, + "weighted_rkd_loss": 0.023732753098011018 + }, + { + "epoch": 4.733779458830376, + "flow_matching_loss": 0.024843456223607063, + "grad_norm": 0.29595643281936646, + "learning_rate": 2.94033669095376e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009840844478458167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042360760271549225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023917765356600285, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9553528130054474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16270, + "weighted_flow_matching_loss": 0.024843456223607063, + "weighted_rkd_loss": 0.022849874570965767 + }, + { + "epoch": 4.7366889729415185, + "flow_matching_loss": 0.026412265561521053, + "grad_norm": 0.22742243111133575, + "learning_rate": 2.940221211970669e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008636434655636549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003928531636483967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021201400831341744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9552987217903137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16280, + "weighted_flow_matching_loss": 0.026412265561521053, + "weighted_rkd_loss": 0.020253690984100105 + }, + { + "epoch": 4.739598487052662, + "flow_matching_loss": 0.024029985070228577, + "grad_norm": 0.31627488136291504, + "learning_rate": 2.940105623612282e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01046809828840196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004762691399082542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02569888811558485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9552446067333221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16290, + "weighted_flow_matching_loss": 0.024029985070228577, + "weighted_rkd_loss": 0.024548684991896153 + }, + { + "epoch": 4.742508001163806, + "flow_matching_loss": 0.024803357012569905, + "grad_norm": 0.278072714805603, + "learning_rate": 2.9399899258873767e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010186612512916327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004535920405760408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024909145571291445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9551904678344727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16300, + "weighted_flow_matching_loss": 0.024803357012569905, + "weighted_rkd_loss": 0.023792975954711436 + }, + { + "epoch": 4.745417515274949, + "flow_matching_loss": 0.025485927239060402, + "grad_norm": 0.30212846398353577, + "learning_rate": 2.9398741188047395e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009919500816613436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004536397638730705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024375399202108385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9551362812519073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16310, + "weighted_flow_matching_loss": 0.025485927239060402, + "weighted_rkd_loss": 0.02328182999044657 + }, + { + "epoch": 4.748327029386092, + "flow_matching_loss": 0.02729117013514042, + "grad_norm": 0.2562486529350281, + "learning_rate": 2.939758202373165e-05, + "loss": 0.0532, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01101627480238676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005107016139663756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027139565907418728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9550820887088776, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16320, + "weighted_flow_matching_loss": 0.02729117013514042, + "weighted_rkd_loss": 0.025920544192194937 + }, + { + "epoch": 4.751236543497236, + "flow_matching_loss": 0.025415817089378834, + "grad_norm": 0.2836042046546936, + "learning_rate": 2.9396421766014574e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009868416003882885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004633133648894727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024369965866208078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.955027848482132, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16330, + "weighted_flow_matching_loss": 0.025415817089378834, + "weighted_rkd_loss": 0.023273996263742446 + }, + { + "epoch": 4.754146057608379, + "flow_matching_loss": 0.02665398344397545, + "grad_norm": 0.25913137197494507, + "learning_rate": 2.9395260414984267e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010331018501892686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004421904124319554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025083941221237183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.954973554611206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16340, + "weighted_flow_matching_loss": 0.02665398344397545, + "weighted_rkd_loss": 0.023954474553465843 + }, + { + "epoch": 4.757055571719523, + "flow_matching_loss": 0.027423860877752303, + "grad_norm": 0.2770973742008209, + "learning_rate": 2.9394097970728935e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010062965657562017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004913662141188979, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025039593689143658, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9549192607402801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16350, + "weighted_flow_matching_loss": 0.027423860877752303, + "weighted_rkd_loss": 0.023910770565271376 + }, + { + "epoch": 4.759965085830666, + "flow_matching_loss": 0.027128028497099877, + "grad_norm": 0.2852018177509308, + "learning_rate": 2.9392934433336852e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009492580033838748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004140327032655477, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02312548700720072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9548649251461029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16360, + "weighted_flow_matching_loss": 0.027128028497099877, + "weighted_rkd_loss": 0.022081713005900384 + }, + { + "epoch": 4.76287459994181, + "flow_matching_loss": 0.025962813384830953, + "grad_norm": 0.28343337774276733, + "learning_rate": 2.9391769802896385e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010033739730715752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004856249573640526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02492372877895832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9548105597496033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16370, + "weighted_flow_matching_loss": 0.025962813384830953, + "weighted_rkd_loss": 0.02379743251949549 + }, + { + "epoch": 4.765784114052953, + "flow_matching_loss": 0.023748760297894477, + "grad_norm": 0.2701338529586792, + "learning_rate": 2.939060407949598e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010378238651901484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047369523672387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025493429601192476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9547561645507813, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16380, + "weighted_flow_matching_loss": 0.023748760297894477, + "weighted_rkd_loss": 0.02433999963104725 + }, + { + "epoch": 4.768693628164097, + "flow_matching_loss": 0.025692971423268318, + "grad_norm": 0.35167232155799866, + "learning_rate": 2.9389437263224163e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010946380672976375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048010325990617275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026693793758749963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9547017335891723, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16390, + "weighted_flow_matching_loss": 0.025692971423268318, + "weighted_rkd_loss": 0.025484582781791686 + }, + { + "epoch": 4.77160314227524, + "flow_matching_loss": 0.024253463745117186, + "grad_norm": 0.25478899478912354, + "learning_rate": 2.938826935416955e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008844619803130626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038674807641655208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021556720696389677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9546472728252411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16400, + "weighted_flow_matching_loss": 0.024253463745117186, + "weighted_rkd_loss": 0.02057909108698368 + }, + { + "epoch": 4.774512656386383, + "flow_matching_loss": 0.02516526523977518, + "grad_norm": 0.2880135178565979, + "learning_rate": 2.9387100352420833e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010354848019778729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004330229829065502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025039925798773767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9545927882194519, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16410, + "weighted_flow_matching_loss": 0.02516526523977518, + "weighted_rkd_loss": 0.02390292417258024 + }, + { + "epoch": 4.777422170497527, + "flow_matching_loss": 0.02431030310690403, + "grad_norm": 0.25855913758277893, + "learning_rate": 2.938593025806679e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010511049628257751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00463051563128829, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025652614794671534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9545382559299469, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16420, + "weighted_flow_matching_loss": 0.02431030310690403, + "weighted_rkd_loss": 0.024486407823860645 + }, + { + "epoch": 4.7803316846086705, + "flow_matching_loss": 0.023486164025962354, + "grad_norm": 0.22640055418014526, + "learning_rate": 2.9384759071196282e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009536770405247808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044260963564738635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023499637097120284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9544837057590485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16430, + "weighted_flow_matching_loss": 0.023486164025962354, + "weighted_rkd_loss": 0.022430053912103176 + }, + { + "epoch": 4.783241198719814, + "flow_matching_loss": 0.023839531280100346, + "grad_norm": 0.24880094826221466, + "learning_rate": 2.9383586791898254e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009895962430164218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004481222154572606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024273147247731684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9544291257858276, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16440, + "weighted_flow_matching_loss": 0.023839531280100346, + "weighted_rkd_loss": 0.02316700778901577 + }, + { + "epoch": 4.786150712830957, + "flow_matching_loss": 0.025378895923495293, + "grad_norm": 0.2894848883152008, + "learning_rate": 2.9382413420261733e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010534996632486582, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004783727740868926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025853720866143702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9543745160102844, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16450, + "weighted_flow_matching_loss": 0.025378895923495293, + "weighted_rkd_loss": 0.02467415537685156 + }, + { + "epoch": 4.789060226942101, + "flow_matching_loss": 0.022660620883107184, + "grad_norm": 0.3207176923751831, + "learning_rate": 2.938123895637583e-05, + "loss": 0.0458, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00997171108610928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004333191690966487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024276614002883435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9543198645114899, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16460, + "weighted_flow_matching_loss": 0.022660620883107184, + "weighted_rkd_loss": 0.023167601227760314 + }, + { + "epoch": 4.7919697410532445, + "flow_matching_loss": 0.024636637605726718, + "grad_norm": 0.27787473797798157, + "learning_rate": 2.9380063400329735e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009715256746858358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004369003116153181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023799516819417478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9542651891708374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16470, + "weighted_flow_matching_loss": 0.024636637605726718, + "weighted_rkd_loss": 0.022711088135838508 + }, + { + "epoch": 4.794879255164387, + "flow_matching_loss": 0.025708685256540777, + "grad_norm": 0.23179759085178375, + "learning_rate": 2.9378886752212728e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010034653637558223, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004538674512878061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02460798155516386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9542104780673981, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16480, + "weighted_flow_matching_loss": 0.025708685256540777, + "weighted_rkd_loss": 0.023481228854507208 + }, + { + "epoch": 4.797788769275531, + "flow_matching_loss": 0.023152987100183965, + "grad_norm": 0.25697463750839233, + "learning_rate": 2.9377709012114163e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010369378933683037, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004739910108037293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02547866813838482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9541557371616364, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16490, + "weighted_flow_matching_loss": 0.023152987100183965, + "weighted_rkd_loss": 0.024310632981359957 + }, + { + "epoch": 4.800698283386675, + "flow_matching_loss": 0.024357030540704726, + "grad_norm": 0.31989747285842896, + "learning_rate": 2.937653018012348e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010123671172186733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046740163583308455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02492135837674141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9541009664535522, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16500, + "weighted_flow_matching_loss": 0.024357030540704726, + "weighted_rkd_loss": 0.023777486570179463 + }, + { + "epoch": 4.8036077974978175, + "flow_matching_loss": 0.026243143901228904, + "grad_norm": 0.2779523432254791, + "learning_rate": 2.9375350256330208e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010330234188586473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004712935537099838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025373403541743757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9540461838245392, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16510, + "weighted_flow_matching_loss": 0.026243143901228904, + "weighted_rkd_loss": 0.024207398854196073 + }, + { + "epoch": 4.806517311608961, + "flow_matching_loss": 0.02399100698530674, + "grad_norm": 0.2718713879585266, + "learning_rate": 2.937416924082395e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010504300612956286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004849117156118155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025857718475162984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9539913535118103, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16520, + "weighted_flow_matching_loss": 0.02399100698530674, + "weighted_rkd_loss": 0.02466802913695574 + }, + { + "epoch": 4.809426825720105, + "flow_matching_loss": 0.02271990440785885, + "grad_norm": 0.24199046194553375, + "learning_rate": 2.937298713369441e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010382396727800369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004452744289301336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025217537954449653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9539364635944366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16530, + "weighted_flow_matching_loss": 0.02271990440785885, + "weighted_rkd_loss": 0.024055957235395907 + }, + { + "epoch": 4.812336339831248, + "flow_matching_loss": 0.025566176697611808, + "grad_norm": 0.48558852076530457, + "learning_rate": 2.9371803935031343e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010377374757081271, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004929154412820935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025683904066681862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9538815796375275, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16540, + "weighted_flow_matching_loss": 0.025566176697611808, + "weighted_rkd_loss": 0.02449938151985407 + }, + { + "epoch": 4.8152458539423915, + "flow_matching_loss": 0.02335724253207445, + "grad_norm": 0.24837267398834229, + "learning_rate": 2.937061964492461e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010057160723954439, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004675282258540392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024789603799581526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.953826642036438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16550, + "weighted_flow_matching_loss": 0.02335724253207445, + "weighted_rkd_loss": 0.023644992522895335 + }, + { + "epoch": 4.818155368053535, + "flow_matching_loss": 0.024758398160338403, + "grad_norm": 0.24991901218891144, + "learning_rate": 2.936943426346416e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009757104143500329, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004491059482097626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024005267955362796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9537716925144195, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16560, + "weighted_flow_matching_loss": 0.024758398160338403, + "weighted_rkd_loss": 0.02289553340524435 + }, + { + "epoch": 4.821064882164679, + "flow_matching_loss": 0.026505802758038044, + "grad_norm": 0.24864721298217773, + "learning_rate": 2.9368247790740005e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00947855031117797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004429198522120714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02338629886507988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9537167012691498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16570, + "weighted_flow_matching_loss": 0.026505802758038044, + "weighted_rkd_loss": 0.022303901240229608 + }, + { + "epoch": 4.823974396275822, + "flow_matching_loss": 0.024754415452480315, + "grad_norm": 0.2656085789203644, + "learning_rate": 2.9367060226842246e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009621510421857238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004559255996719002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02380227707326412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9536616921424865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16580, + "weighted_flow_matching_loss": 0.024754415452480315, + "weighted_rkd_loss": 0.02269929591566324 + }, + { + "epoch": 4.826883910386965, + "flow_matching_loss": 0.026891641318798065, + "grad_norm": 0.2889483869075775, + "learning_rate": 2.9365871571861088e-05, + "loss": 0.0499, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009957477077841758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042070576688274745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02412201203405857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9536066353321075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16590, + "weighted_flow_matching_loss": 0.026891641318798065, + "weighted_rkd_loss": 0.02300288751721382 + }, + { + "epoch": 4.829793424498109, + "flow_matching_loss": 0.025514380447566508, + "grad_norm": 0.2957097291946411, + "learning_rate": 2.936468182588678e-05, + "loss": 0.049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010100333672016858, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004457825841382146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02465849332511425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9535515487194062, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16600, + "weighted_flow_matching_loss": 0.025514380447566508, + "weighted_rkd_loss": 0.02351313903927803 + }, + { + "epoch": 4.832702938609252, + "flow_matching_loss": 0.02410937212407589, + "grad_norm": 0.22779136896133423, + "learning_rate": 2.9363490989009693e-05, + "loss": 0.0493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01060196771286428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005259063746780157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026462999545037745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9534964263439178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16610, + "weighted_flow_matching_loss": 0.02410937212407589, + "weighted_rkd_loss": 0.025232398323714732 + }, + { + "epoch": 4.835612452720396, + "flow_matching_loss": 0.025390291400253773, + "grad_norm": 0.2779054045677185, + "learning_rate": 2.936229906132025e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009506868757307529, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041476869722828266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02316142413765192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9534412920475006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16620, + "weighted_flow_matching_loss": 0.025390291400253773, + "weighted_rkd_loss": 0.022083062771707775 + }, + { + "epoch": 4.838521966831539, + "flow_matching_loss": 0.026210296899080276, + "grad_norm": 0.28316131234169006, + "learning_rate": 2.9361106042908984e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009761219378560781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004502014094032347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02402445301413536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9533861219882965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16630, + "weighted_flow_matching_loss": 0.026210296899080276, + "weighted_rkd_loss": 0.02290455102920532 + }, + { + "epoch": 4.841431480942683, + "flow_matching_loss": 0.02612002268433571, + "grad_norm": 0.3158828914165497, + "learning_rate": 2.9359911933866486e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01040609972551465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048031135695055125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025615312904119492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9533309161663055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16640, + "weighted_flow_matching_loss": 0.02612002268433571, + "weighted_rkd_loss": 0.0244198739528656 + }, + { + "epoch": 4.844340995053826, + "flow_matching_loss": 0.02361602522432804, + "grad_norm": 0.27818581461906433, + "learning_rate": 2.935871673428344e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010567593248561024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004828331340104341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02596351783722639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9532756745815277, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16650, + "weighted_flow_matching_loss": 0.02361602522432804, + "weighted_rkd_loss": 0.024750364199280738 + }, + { + "epoch": 4.84725050916497, + "flow_matching_loss": 0.023880279250442983, + "grad_norm": 0.2332277148962021, + "learning_rate": 2.9357520444250627e-05, + "loss": 0.0461, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009526955103501678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004217572114430368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02327148225158453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9532204091548919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16660, + "weighted_flow_matching_loss": 0.023880279250442983, + "weighted_rkd_loss": 0.022182877548038958 + }, + { + "epoch": 4.850160023276112, + "flow_matching_loss": 0.024048615992069245, + "grad_norm": 0.26975691318511963, + "learning_rate": 2.935632306385888e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009665938606485725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004308843007311225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02364071998745203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9531651139259338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16670, + "weighted_flow_matching_loss": 0.024048615992069245, + "weighted_rkd_loss": 0.02253348082304001 + }, + { + "epoch": 4.853069537387256, + "flow_matching_loss": 0.024099050275981426, + "grad_norm": 0.23627004027366638, + "learning_rate": 2.9355124593199143e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00961293843574822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004465489368885756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023691366240382195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9531097829341888, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16680, + "weighted_flow_matching_loss": 0.024099050275981426, + "weighted_rkd_loss": 0.022580486349761487 + }, + { + "epoch": 4.8559790514984, + "flow_matching_loss": 0.02753213308751583, + "grad_norm": 0.3229916989803314, + "learning_rate": 2.9353925032362424e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010728741157799959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004937963513657451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026395445875823497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.953054416179657, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16690, + "weighted_flow_matching_loss": 0.02753213308751583, + "weighted_rkd_loss": 0.025156255066394805 + }, + { + "epoch": 4.8588885656095435, + "flow_matching_loss": 0.024531199410557748, + "grad_norm": 0.25747182965278625, + "learning_rate": 2.935272438143983e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008762604091316462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037388102151453494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02126401849091053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9529990255832672, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16700, + "weighted_flow_matching_loss": 0.024531199410557748, + "weighted_rkd_loss": 0.02026458252221346 + }, + { + "epoch": 4.861798079720686, + "flow_matching_loss": 0.02448675539344549, + "grad_norm": 0.27894142270088196, + "learning_rate": 2.935152264052254e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010140525083988905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004432844743132591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024713895097374917, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9529435932636261, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16710, + "weighted_flow_matching_loss": 0.02448675539344549, + "weighted_rkd_loss": 0.023550940863788127 + }, + { + "epoch": 4.86470759383183, + "flow_matching_loss": 0.02523156888782978, + "grad_norm": 0.2749464511871338, + "learning_rate": 2.9350319809701814e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009813091391697526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004302687500603497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02392887007445097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9528881549835205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16720, + "weighted_flow_matching_loss": 0.02523156888782978, + "weighted_rkd_loss": 0.022801529616117477 + }, + { + "epoch": 4.867617107942974, + "flow_matching_loss": 0.025617541559040547, + "grad_norm": 0.2432018220424652, + "learning_rate": 2.9349115889069005e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010047651967033744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004597284412011504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024692588485777378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9528326570987702, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16730, + "weighted_flow_matching_loss": 0.025617541559040547, + "weighted_rkd_loss": 0.023527883924543858 + }, + { + "epoch": 4.870526622054117, + "flow_matching_loss": 0.025851936265826225, + "grad_norm": 0.26611489057540894, + "learning_rate": 2.9347910878715542e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009373099310323596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00425645038485527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02300264947116375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9527771472930908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16740, + "weighted_flow_matching_loss": 0.025851936265826225, + "weighted_rkd_loss": 0.02191643100231886 + }, + { + "epoch": 4.87343613616526, + "flow_matching_loss": 0.024220577068626882, + "grad_norm": 0.251061350107193, + "learning_rate": 2.9346704778732926e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009268169291317463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004172460990957916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022708799690008163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9527215957641602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16750, + "weighted_flow_matching_loss": 0.024220577068626882, + "weighted_rkd_loss": 0.021635182201862335 + }, + { + "epoch": 4.876345650276404, + "flow_matching_loss": 0.024862840585410596, + "grad_norm": 0.24718426167964935, + "learning_rate": 2.934549758921277e-05, + "loss": 0.0465, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009194308193400502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004314387170597911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022703003697097303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9526660203933716, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16760, + "weighted_flow_matching_loss": 0.024862840585410596, + "weighted_rkd_loss": 0.021628371253609658 + }, + { + "epoch": 4.879255164387548, + "flow_matching_loss": 0.022955907881259917, + "grad_norm": 0.28578805923461914, + "learning_rate": 2.9344289310246745e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010158812534064055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004478167789056897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024795793183147907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9526104152202606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16770, + "weighted_flow_matching_loss": 0.022955907881259917, + "weighted_rkd_loss": 0.023620732873678208 + }, + { + "epoch": 4.8821646784986905, + "flow_matching_loss": 0.025409338995814323, + "grad_norm": 0.25233203172683716, + "learning_rate": 2.9343079941926607e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008881213376298547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00407707744743675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02183950413018465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9525547683238983, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16780, + "weighted_flow_matching_loss": 0.025409338995814323, + "weighted_rkd_loss": 0.02080334136262536 + }, + { + "epoch": 4.885074192609834, + "flow_matching_loss": 0.027274762839078905, + "grad_norm": 0.23366153240203857, + "learning_rate": 2.9341869484344206e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009806970926001668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004520945623517037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02413488756865263, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9524990975856781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16790, + "weighted_flow_matching_loss": 0.027274762839078905, + "weighted_rkd_loss": 0.022988471388816833 + }, + { + "epoch": 4.887983706720978, + "flow_matching_loss": 0.024404660053551196, + "grad_norm": 0.29562240839004517, + "learning_rate": 2.934065793759147e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01038238750770688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049052415182814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02567001646384597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.952443391084671, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16800, + "weighted_flow_matching_loss": 0.024404660053551196, + "weighted_rkd_loss": 0.024449204653501512 + }, + { + "epoch": 4.890893220832121, + "flow_matching_loss": 0.023175371065735816, + "grad_norm": 0.26535677909851074, + "learning_rate": 2.9339445301760396e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010135666001588106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004564324114471674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02483565602451563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.952387660741806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16810, + "weighted_flow_matching_loss": 0.023175371065735816, + "weighted_rkd_loss": 0.023653182201087475 + }, + { + "epoch": 4.8938027349432645, + "flow_matching_loss": 0.026658644527196886, + "grad_norm": 0.26058897376060486, + "learning_rate": 2.933823157694309e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010481512732803821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004627893469296396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025590918958187103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9523318946361542, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16820, + "weighted_flow_matching_loss": 0.026658644527196886, + "weighted_rkd_loss": 0.02437104918062687 + }, + { + "epoch": 4.896712249054408, + "flow_matching_loss": 0.023509199917316436, + "grad_norm": 0.2749086022377014, + "learning_rate": 2.9337016763231713e-05, + "loss": 0.0465, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009817523881793022, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004546573525294662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02418162114918232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.95227609872818, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16830, + "weighted_flow_matching_loss": 0.023509199917316436, + "weighted_rkd_loss": 0.023027582466602324 + }, + { + "epoch": 4.899621763165551, + "flow_matching_loss": 0.025609876215457916, + "grad_norm": 0.27585679292678833, + "learning_rate": 2.9335800860718528e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010533662326633931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004879476479254663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02594680115580559, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9522202730178833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16840, + "weighted_flow_matching_loss": 0.025609876215457916, + "weighted_rkd_loss": 0.024707093276083468 + }, + { + "epoch": 4.902531277276695, + "flow_matching_loss": 0.023754979483783245, + "grad_norm": 0.2633199393749237, + "learning_rate": 2.9334583869495887e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010209732875227927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004302360117435455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024721826054155828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9521644115447998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16850, + "weighted_flow_matching_loss": 0.023754979483783245, + "weighted_rkd_loss": 0.02353926859796047 + }, + { + "epoch": 4.905440791387838, + "flow_matching_loss": 0.026285064965486528, + "grad_norm": 0.24983444809913635, + "learning_rate": 2.9333365789656193e-05, + "loss": 0.0504, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010407405719161034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004461596766486764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025276408344507218, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9521085262298584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16860, + "weighted_flow_matching_loss": 0.026285064965486528, + "weighted_rkd_loss": 0.02406592294573784 + }, + { + "epoch": 4.908350305498981, + "flow_matching_loss": 0.02492987588047981, + "grad_norm": 0.21228544414043427, + "learning_rate": 2.9332146621291954e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009775897208601236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00432113625574857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02387293092906475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9520526111125946, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16870, + "weighted_flow_matching_loss": 0.02492987588047981, + "weighted_rkd_loss": 0.022728312574326993 + }, + { + "epoch": 4.911259819610125, + "flow_matching_loss": 0.023563569970428945, + "grad_norm": 0.289231538772583, + "learning_rate": 2.9330926364495775e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010494779516011476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004760093474760652, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02574965264648199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9519966602325439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16880, + "weighted_flow_matching_loss": 0.023563569970428945, + "weighted_rkd_loss": 0.02451356705278158 + }, + { + "epoch": 4.914169333721269, + "flow_matching_loss": 0.023730052821338176, + "grad_norm": 0.28447115421295166, + "learning_rate": 2.9329705019360305e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009704784490168095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00434757461771369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02375714387744665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9519406616687774, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16890, + "weighted_flow_matching_loss": 0.023730052821338176, + "weighted_rkd_loss": 0.022615390084683894 + }, + { + "epoch": 4.917078847832412, + "flow_matching_loss": 0.02491345666348934, + "grad_norm": 0.31408214569091797, + "learning_rate": 2.9328482585978314e-05, + "loss": 0.0482, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009992205351591111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004527028021402657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024511438980698585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9518846869468689, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16900, + "weighted_flow_matching_loss": 0.02491345666348934, + "weighted_rkd_loss": 0.02333205137401819 + }, + { + "epoch": 4.919988361943555, + "flow_matching_loss": 0.025432546995580195, + "grad_norm": 0.26150065660476685, + "learning_rate": 2.9327259064442627e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009732963377609849, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004588751192204654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024054678063839673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9518286108970642, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16910, + "weighted_flow_matching_loss": 0.025432546995580195, + "weighted_rkd_loss": 0.02289597149938345 + }, + { + "epoch": 4.922897876054699, + "flow_matching_loss": 0.025870626047253608, + "grad_norm": 0.2444661557674408, + "learning_rate": 2.9326034454846166e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009250815771520138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038903735345229506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02239200510084629, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9517725467681885, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16920, + "weighted_flow_matching_loss": 0.025870626047253608, + "weighted_rkd_loss": 0.021312106773257254 + }, + { + "epoch": 4.925807390165843, + "flow_matching_loss": 0.023387910239398478, + "grad_norm": 0.24661818146705627, + "learning_rate": 2.932480875728193e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01048080027103424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004566441453061998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025528041832149028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9517164468765259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16930, + "weighted_flow_matching_loss": 0.023387910239398478, + "weighted_rkd_loss": 0.02429545596241951 + }, + { + "epoch": 4.928716904276985, + "flow_matching_loss": 0.0257396399974823, + "grad_norm": 0.31257104873657227, + "learning_rate": 2.9323581971843004e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010559717938303948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004516541282646358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02563597708940506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9516602993011475, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16940, + "weighted_flow_matching_loss": 0.0257396399974823, + "weighted_rkd_loss": 0.02439671866595745 + }, + { + "epoch": 4.931626418388129, + "flow_matching_loss": 0.02553013451397419, + "grad_norm": 0.2804906964302063, + "learning_rate": 2.932235409862256e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010417762119323016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005034447996877134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025869972445070742, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9516041338443756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16950, + "weighted_flow_matching_loss": 0.02553013451397419, + "weighted_rkd_loss": 0.02461796663701534 + }, + { + "epoch": 4.934535932499273, + "flow_matching_loss": 0.023363616690039633, + "grad_norm": 0.26602837443351746, + "learning_rate": 2.9321125137713833e-05, + "loss": 0.0482, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010646544303745032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047831190284341575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02607620768249035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9515479326248169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16960, + "weighted_flow_matching_loss": 0.023363616690039633, + "weighted_rkd_loss": 0.02481277547776699 + }, + { + "epoch": 4.9374454466104165, + "flow_matching_loss": 0.02411932647228241, + "grad_norm": 0.29484954476356506, + "learning_rate": 2.931989508921017e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010774898109957576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005122876400128007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026672672480344772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9514917016029358, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16970, + "weighted_flow_matching_loss": 0.02411932647228241, + "weighted_rkd_loss": 0.0253788398578763 + }, + { + "epoch": 4.940354960721559, + "flow_matching_loss": 0.02523169033229351, + "grad_norm": 0.277042031288147, + "learning_rate": 2.9318663953204976e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010368355177342891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047392311040312055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025475941598415375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9514354467391968, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16980, + "weighted_flow_matching_loss": 0.02523169033229351, + "weighted_rkd_loss": 0.024238712154328823 + }, + { + "epoch": 4.943264474832703, + "flow_matching_loss": 0.026597934402525426, + "grad_norm": 0.23699097335338593, + "learning_rate": 2.931743172979175e-05, + "loss": 0.0482, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009274367336183786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004188571660779417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02273730617016554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9513791501522064, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16990, + "weighted_flow_matching_loss": 0.026597934402525426, + "weighted_rkd_loss": 0.021631849370896815 + }, + { + "epoch": 4.946173988943847, + "flow_matching_loss": 0.023800413496792315, + "grad_norm": 0.235436350107193, + "learning_rate": 2.931619841906407e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00933942743577063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004271442699246108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02295029778033495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9513228237628937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17000, + "weighted_flow_matching_loss": 0.023800413496792315, + "weighted_rkd_loss": 0.02183315549045801 + }, + { + "epoch": 4.94908350305499, + "flow_matching_loss": 0.025013337284326552, + "grad_norm": 0.2705077826976776, + "learning_rate": 2.9314964021115597e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009397485433146358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041465729009360075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022941543720662594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.951266485452652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17010, + "weighted_flow_matching_loss": 0.025013337284326552, + "weighted_rkd_loss": 0.02182349655777216 + }, + { + "epoch": 4.951993017166133, + "flow_matching_loss": 0.022944959811866283, + "grad_norm": 0.21147078275680542, + "learning_rate": 2.931372853604008e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008687452645972371, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00390431412961334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02127921935170889, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9512100934982299, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17020, + "weighted_flow_matching_loss": 0.022944959811866283, + "weighted_rkd_loss": 0.020240999571979047 + }, + { + "epoch": 4.954902531277277, + "flow_matching_loss": 0.02349622081965208, + "grad_norm": 0.3561236560344696, + "learning_rate": 2.9312491963931344e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009633862366899847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004367894399911166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02363561876118183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.95115367770195, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17030, + "weighted_flow_matching_loss": 0.02349622081965208, + "weighted_rkd_loss": 0.022481060586869717 + }, + { + "epoch": 4.95781204538842, + "flow_matching_loss": 0.02657253108918667, + "grad_norm": 0.27638551592826843, + "learning_rate": 2.9311254304883295e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009318228950724006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043012324254959825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022937690652906895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9510972321033477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17040, + "weighted_flow_matching_loss": 0.02657253108918667, + "weighted_rkd_loss": 0.021815978921949863 + }, + { + "epoch": 4.9607215594995635, + "flow_matching_loss": 0.02558692991733551, + "grad_norm": 0.28051042556762695, + "learning_rate": 2.931001555898993e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010176252666860819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004456362873315811, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024808868393301965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.951040756702423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17050, + "weighted_flow_matching_loss": 0.02558692991733551, + "weighted_rkd_loss": 0.023594242706894874 + }, + { + "epoch": 4.963631073610707, + "flow_matching_loss": 0.023884611018002034, + "grad_norm": 0.26343727111816406, + "learning_rate": 2.930877572634532e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010480067227035761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004770661285147071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025730795599520205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9509842455387115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17060, + "weighted_flow_matching_loss": 0.023884611018002034, + "weighted_rkd_loss": 0.024469623155891896 + }, + { + "epoch": 4.96654058772185, + "flow_matching_loss": 0.02603805661201477, + "grad_norm": 0.24715861678123474, + "learning_rate": 2.930753480704362e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00934370313771069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040227054618299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022710111737251282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9509277164936065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17070, + "weighted_flow_matching_loss": 0.02603805661201477, + "weighted_rkd_loss": 0.021595660503953694 + }, + { + "epoch": 4.969450101832994, + "flow_matching_loss": 0.022477381490170957, + "grad_norm": 0.25851741433143616, + "learning_rate": 2.930629280117908e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008431780990213156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003576855082064867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020440417155623437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9508711397647858, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17080, + "weighted_flow_matching_loss": 0.022477381490170957, + "weighted_rkd_loss": 0.019436190463602544 + }, + { + "epoch": 4.9723596159441374, + "flow_matching_loss": 0.02466166764497757, + "grad_norm": 0.3313184380531311, + "learning_rate": 2.930504970884601e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011298397555947304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005096220877021551, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027693016082048418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9508145153522491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17090, + "weighted_flow_matching_loss": 0.02466166764497757, + "weighted_rkd_loss": 0.026330887153744698 + }, + { + "epoch": 4.975269130055281, + "flow_matching_loss": 0.024218657426536082, + "grad_norm": 0.2623947262763977, + "learning_rate": 2.9303805530138822e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010534408874809742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004723411891609431, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02579222936183214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9507578909397125, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17100, + "weighted_flow_matching_loss": 0.024218657426536082, + "weighted_rkd_loss": 0.024522184394299984 + }, + { + "epoch": 4.978178644166424, + "flow_matching_loss": 0.024145964719355106, + "grad_norm": 0.26274481415748596, + "learning_rate": 2.9302560265152e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010080045089125633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004417601274326444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024577691592276095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9507012486457824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17110, + "weighted_flow_matching_loss": 0.024145964719355106, + "weighted_rkd_loss": 0.02336600683629513 + }, + { + "epoch": 4.981088158277568, + "flow_matching_loss": 0.023989454843103886, + "grad_norm": 0.28878819942474365, + "learning_rate": 2.9301313913980113e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009094429248943926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004113441752269864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022302300110459326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9506445527076721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17120, + "weighted_flow_matching_loss": 0.023989454843103886, + "weighted_rkd_loss": 0.02120158262550831 + }, + { + "epoch": 4.983997672388711, + "flow_matching_loss": 0.026738549396395684, + "grad_norm": 0.35266050696372986, + "learning_rate": 2.9300066476717815e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009824415296316147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004479482956230641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024128313176333904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9505878329277039, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17130, + "weighted_flow_matching_loss": 0.026738549396395684, + "weighted_rkd_loss": 0.02293609417974949 + }, + { + "epoch": 4.986907186499854, + "flow_matching_loss": 0.027651483565568923, + "grad_norm": 0.2528988718986511, + "learning_rate": 2.9298817953459843e-05, + "loss": 0.052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010391183895990252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004801051295362413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02558341920375824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9505310833454133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17140, + "weighted_flow_matching_loss": 0.027651483565568923, + "weighted_rkd_loss": 0.024317860789597034 + }, + { + "epoch": 4.989816700610998, + "flow_matching_loss": 0.026464690826833247, + "grad_norm": 0.25927212834358215, + "learning_rate": 2.929756834430101e-05, + "loss": 0.0485, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009399636229500175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004367107176221907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02316637970507145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9504742920398712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17150, + "weighted_flow_matching_loss": 0.026464690826833247, + "weighted_rkd_loss": 0.02201903034001589 + }, + { + "epoch": 4.992726214722142, + "flow_matching_loss": 0.02525605671107769, + "grad_norm": 0.2527254819869995, + "learning_rate": 2.9296317649336215e-05, + "loss": 0.0493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010276942001655697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047628681641072035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0253167524933815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9504174768924714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17160, + "weighted_flow_matching_loss": 0.02525605671107769, + "weighted_rkd_loss": 0.024061472713947298 + }, + { + "epoch": 4.995635728833285, + "flow_matching_loss": 0.02529959473758936, + "grad_norm": 0.22457373142242432, + "learning_rate": 2.9295065868660444e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009763666475191713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004421431454829872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02394876480102539, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9503606259822845, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17170, + "weighted_flow_matching_loss": 0.02529959473758936, + "weighted_rkd_loss": 0.022759997844696046 + }, + { + "epoch": 4.998545242944428, + "flow_matching_loss": 0.023929910361766817, + "grad_norm": 0.2567622661590576, + "learning_rate": 2.9293813002368758e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009568777028471231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004370532277971506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02350808624178171, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9503037512302399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17180, + "weighted_flow_matching_loss": 0.023929910361766817, + "weighted_rkd_loss": 0.022339843213558197 + }, + { + "epoch": 5.001454757055572, + "flow_matching_loss": 0.02380678942427039, + "grad_norm": 0.27327725291252136, + "learning_rate": 2.9292559050556306e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00943531678058207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004346248484216631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023216881975531577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9502468407154083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17190, + "weighted_flow_matching_loss": 0.02380678942427039, + "weighted_rkd_loss": 0.022061782702803612 + }, + { + "epoch": 5.004364271166716, + "flow_matching_loss": 0.024817070737481118, + "grad_norm": 0.2823299765586853, + "learning_rate": 2.929130401331832e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010170348174870013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004457142180763185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024797838553786276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9501899003982544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17200, + "weighted_flow_matching_loss": 0.024817070737481118, + "weighted_rkd_loss": 0.023562646098434926 + }, + { + "epoch": 5.007273785277858, + "flow_matching_loss": 0.02529163919389248, + "grad_norm": 0.3009320795536041, + "learning_rate": 2.9290047890750104e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00900286897085607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004063040693290531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022068778984248638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9501329302787781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17210, + "weighted_flow_matching_loss": 0.02529163919389248, + "weighted_rkd_loss": 0.020968245901167394 + }, + { + "epoch": 5.010183299389002, + "flow_matching_loss": 0.02618308924138546, + "grad_norm": 0.28017714619636536, + "learning_rate": 2.9288790682947062e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010323332156985998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004723676922731101, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025370341539382935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9500759303569793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17220, + "weighted_flow_matching_loss": 0.02618308924138546, + "weighted_rkd_loss": 0.024103754572570323 + }, + { + "epoch": 5.013092813500146, + "flow_matching_loss": 0.02472328320145607, + "grad_norm": 0.3513471186161041, + "learning_rate": 2.928753239000466e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009819709416478872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004598884261213243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024238303303718567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9500188887119293, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17230, + "weighted_flow_matching_loss": 0.02472328320145607, + "weighted_rkd_loss": 0.023026836663484575 + }, + { + "epoch": 5.016002327611289, + "flow_matching_loss": 0.02868523169308901, + "grad_norm": 0.2327997386455536, + "learning_rate": 2.9286273012018466e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009411143651232124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004162334674037993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02298462186008692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9499618291854859, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17240, + "weighted_flow_matching_loss": 0.02868523169308901, + "weighted_rkd_loss": 0.021834506280720235 + }, + { + "epoch": 5.018911841722432, + "flow_matching_loss": 0.025033091194927692, + "grad_norm": 0.2576275169849396, + "learning_rate": 2.928501254908412e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008601783635094762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039051078958436845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02110867518931627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9499047338962555, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17250, + "weighted_flow_matching_loss": 0.025033091194927692, + "weighted_rkd_loss": 0.020051226392388342 + }, + { + "epoch": 5.021821355833576, + "flow_matching_loss": 0.025311270356178285, + "grad_norm": 0.26642265915870667, + "learning_rate": 2.9283751001297338e-05, + "loss": 0.0464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009107775008305907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004025107598863542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022240657731890677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9498476088047028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17260, + "weighted_flow_matching_loss": 0.025311270356178285, + "weighted_rkd_loss": 0.021125233732163906 + }, + { + "epoch": 5.024730869944719, + "flow_matching_loss": 0.026542249880731106, + "grad_norm": 0.3106207251548767, + "learning_rate": 2.9282488368753937e-05, + "loss": 0.0493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00983094060793519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004314712202176452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02397659346461296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9497904539108276, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17270, + "weighted_flow_matching_loss": 0.026542249880731106, + "weighted_rkd_loss": 0.022772710770368576 + }, + { + "epoch": 5.0276403840558626, + "flow_matching_loss": 0.023325612396001817, + "grad_norm": 0.2318037897348404, + "learning_rate": 2.9281224651549797e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008895654603838921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004016719688661396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021808028779923917, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9497332572937012, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17280, + "weighted_flow_matching_loss": 0.023325612396001817, + "weighted_rkd_loss": 0.02071180772036314 + }, + { + "epoch": 5.030549898167006, + "flow_matching_loss": 0.025955786928534508, + "grad_norm": 0.2752703130245209, + "learning_rate": 2.927995984978089e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00967797227203846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041873733513057235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023543318174779414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9496760368347168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17290, + "weighted_flow_matching_loss": 0.025955786928534508, + "weighted_rkd_loss": 0.022358519583940507 + }, + { + "epoch": 5.03345941227815, + "flow_matching_loss": 0.02510519064962864, + "grad_norm": 0.24323798716068268, + "learning_rate": 2.9278693963543277e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010232660453766584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004535492765717209, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025000813603401183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9496188044548035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17300, + "weighted_flow_matching_loss": 0.02510519064962864, + "weighted_rkd_loss": 0.02374124210327864 + }, + { + "epoch": 5.036368926389293, + "flow_matching_loss": 0.023393883742392062, + "grad_norm": 0.24804633855819702, + "learning_rate": 2.9277426992933088e-05, + "loss": 0.0453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009347012639045716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004325857758522034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023019883222877978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9495615184307098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17310, + "weighted_flow_matching_loss": 0.023393883742392062, + "weighted_rkd_loss": 0.021858796197921037 + }, + { + "epoch": 5.0392784405004365, + "flow_matching_loss": 0.024087927676737307, + "grad_norm": 0.2547541856765747, + "learning_rate": 2.927615893804654e-05, + "loss": 0.0458, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009291094774380327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00423422462772578, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022816414572298528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9495042145252228, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17320, + "weighted_flow_matching_loss": 0.024087927676737307, + "weighted_rkd_loss": 0.0216642951592803 + }, + { + "epoch": 5.04218795461158, + "flow_matching_loss": 0.02334539070725441, + "grad_norm": 0.25027331709861755, + "learning_rate": 2.9274889798979938e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008757665241137146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038950454443693163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021410376019775867, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9494468688964843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17330, + "weighted_flow_matching_loss": 0.02334539070725441, + "weighted_rkd_loss": 0.020328016206622123 + }, + { + "epoch": 5.045097468722723, + "flow_matching_loss": 0.02668406944721937, + "grad_norm": 0.26237040758132935, + "learning_rate": 2.9273619575829663e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010665741004049777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005005288729444146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026336770690977573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.949389499425888, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17340, + "weighted_flow_matching_loss": 0.02668406944721937, + "weighted_rkd_loss": 0.025003860145807265 + }, + { + "epoch": 5.048006982833867, + "flow_matching_loss": 0.025879958271980287, + "grad_norm": 0.28371989727020264, + "learning_rate": 2.9272348268692172e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0087574967648834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004021068592555821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02153606228530407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9493320882320404, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17350, + "weighted_flow_matching_loss": 0.025879958271980287, + "weighted_rkd_loss": 0.020444873906672 + }, + { + "epoch": 5.05091649694501, + "flow_matching_loss": 0.02368533667176962, + "grad_norm": 0.24188627302646637, + "learning_rate": 2.9271075877664027e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009251564741134644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004116302798502147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02261943221092224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9492746591567993, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17360, + "weighted_flow_matching_loss": 0.02368533667176962, + "weighted_rkd_loss": 0.021472054161131383 + }, + { + "epoch": 5.053826011056153, + "flow_matching_loss": 0.026002819649875164, + "grad_norm": 0.2360302358865738, + "learning_rate": 2.9269802402841844e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010236843768507243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004631084157153964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025104771740734576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9492171943187714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17370, + "weighted_flow_matching_loss": 0.026002819649875164, + "weighted_rkd_loss": 0.02382986806333065 + }, + { + "epoch": 5.056735525167297, + "flow_matching_loss": 0.023142856359481812, + "grad_norm": 0.2696434259414673, + "learning_rate": 2.9268527844322343e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0094313845038414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004271534038707614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023134303279221058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.94915971159935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17380, + "weighted_flow_matching_loss": 0.023142856359481812, + "weighted_rkd_loss": 0.021958162263035774 + }, + { + "epoch": 5.059645039278441, + "flow_matching_loss": 0.025175968185067177, + "grad_norm": 0.2893894910812378, + "learning_rate": 2.9267252202202318e-05, + "loss": 0.05, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01051439791917801, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050842893775552515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026113084703683852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9491021811962128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17390, + "weighted_flow_matching_loss": 0.025175968185067177, + "weighted_rkd_loss": 0.02478403989225626 + }, + { + "epoch": 5.0625545533895835, + "flow_matching_loss": 0.02322189398109913, + "grad_norm": 0.26125243306159973, + "learning_rate": 2.926597547657864e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009829168813303113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004565762775018811, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024224100448191167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9490446209907532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17400, + "weighted_flow_matching_loss": 0.02322189398109913, + "weighted_rkd_loss": 0.022989782877266406 + }, + { + "epoch": 5.065464067500727, + "flow_matching_loss": 0.023683161474764347, + "grad_norm": 0.25647157430648804, + "learning_rate": 2.9264697667548275e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009478332148864866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004193952353671193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02315061669796705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9489870309829712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17410, + "weighted_flow_matching_loss": 0.023683161474764347, + "weighted_rkd_loss": 0.021969633642584085 + }, + { + "epoch": 5.068373581611871, + "flow_matching_loss": 0.0232124337926507, + "grad_norm": 0.3024626076221466, + "learning_rate": 2.926341877520826e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009262766921892762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040819673798978325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02260750122368336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9489294171333313, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17420, + "weighted_flow_matching_loss": 0.0232124337926507, + "weighted_rkd_loss": 0.021452923864126207 + }, + { + "epoch": 5.071283095723015, + "flow_matching_loss": 0.025972578674554825, + "grad_norm": 0.27658888697624207, + "learning_rate": 2.9262138799655717e-05, + "loss": 0.0493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009993528295308352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00457757324911654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0245646296069026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9488717615604401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17430, + "weighted_flow_matching_loss": 0.025972578674554825, + "weighted_rkd_loss": 0.023308682069182397 + }, + { + "epoch": 5.074192609834157, + "flow_matching_loss": 0.025260207056999207, + "grad_norm": 0.2672460675239563, + "learning_rate": 2.926085774098786e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009274145821109414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004003130435012281, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02255142219364643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9488140881061554, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17440, + "weighted_flow_matching_loss": 0.025260207056999207, + "weighted_rkd_loss": 0.02139710932970047 + }, + { + "epoch": 5.077102123945301, + "flow_matching_loss": 0.025535738840699197, + "grad_norm": 0.26151466369628906, + "learning_rate": 2.925957559930197e-05, + "loss": 0.0479, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009533996786922216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004551945743151009, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023619939386844636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9487563788890838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17450, + "weighted_flow_matching_loss": 0.025535738840699197, + "weighted_rkd_loss": 0.022409548051655293 + }, + { + "epoch": 5.080011638056445, + "flow_matching_loss": 0.02481464222073555, + "grad_norm": 0.31515634059906006, + "learning_rate": 2.9258292374695414e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009593229508027434, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004336465732194483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02352292463183403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9486986339092255, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17460, + "weighted_flow_matching_loss": 0.02481464222073555, + "weighted_rkd_loss": 0.022316141799092293 + }, + { + "epoch": 5.082921152167588, + "flow_matching_loss": 0.026359887234866618, + "grad_norm": 0.26476937532424927, + "learning_rate": 2.925700806726565e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010447804676368832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005083115398883819, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02597872484475374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9486408531665802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17470, + "weighted_flow_matching_loss": 0.026359887234866618, + "weighted_rkd_loss": 0.0246444845572114 + }, + { + "epoch": 5.085830666278731, + "flow_matching_loss": 0.024572912231087686, + "grad_norm": 0.2744428813457489, + "learning_rate": 2.9255722677110212e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00878507811576128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00393885113298893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021509007457643746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9485830366611481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17480, + "weighted_flow_matching_loss": 0.024572912231087686, + "weighted_rkd_loss": 0.020403087418526412 + }, + { + "epoch": 5.088740180389875, + "flow_matching_loss": 0.023874354735016822, + "grad_norm": 0.25873735547065735, + "learning_rate": 2.9254436204326716e-05, + "loss": 0.0458, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009523646626621484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004101596749387681, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023148890025913714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9485252201557159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17490, + "weighted_flow_matching_loss": 0.023874354735016822, + "weighted_rkd_loss": 0.02195731233805418 + }, + { + "epoch": 5.091649694501018, + "flow_matching_loss": 0.025767191685736178, + "grad_norm": 0.26525774598121643, + "learning_rate": 2.925314864901286e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0102905229665339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004633324127644301, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02521436959505081, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.948467367887497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17500, + "weighted_flow_matching_loss": 0.025767191685736178, + "weighted_rkd_loss": 0.023915050737559795 + }, + { + "epoch": 5.094559208612162, + "flow_matching_loss": 0.02369636483490467, + "grad_norm": 0.2457350343465805, + "learning_rate": 2.925186001126643e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00949184144847095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004379328852519393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02336301188915968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9484094679355621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17510, + "weighted_flow_matching_loss": 0.02369636483490467, + "weighted_rkd_loss": 0.022157743759453296 + }, + { + "epoch": 5.097468722723305, + "flow_matching_loss": 0.024067949689924716, + "grad_norm": 0.266458660364151, + "learning_rate": 2.9250570291185284e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010638187360018491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047354873735457655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02601186204701662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9483515441417694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17520, + "weighted_flow_matching_loss": 0.024067949689924716, + "weighted_rkd_loss": 0.024668395705521106 + }, + { + "epoch": 5.100378236834449, + "flow_matching_loss": 0.02509484514594078, + "grad_norm": 0.23887848854064941, + "learning_rate": 2.9249279488867373e-05, + "loss": 0.0461, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009076867252588272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004022659454494715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02217639423906803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9482935845851899, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17530, + "weighted_flow_matching_loss": 0.02509484514594078, + "weighted_rkd_loss": 0.0210297055542469 + }, + { + "epoch": 5.103287750945592, + "flow_matching_loss": 0.02303258180618286, + "grad_norm": 0.21085526049137115, + "learning_rate": 2.9247987604410717e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01013857745565474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00494855570141226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02522571086883545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9482356011867523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17540, + "weighted_flow_matching_loss": 0.02303258180618286, + "weighted_rkd_loss": 0.023919932544231415 + }, + { + "epoch": 5.1061972650567355, + "flow_matching_loss": 0.02400908339768648, + "grad_norm": 0.29629090428352356, + "learning_rate": 2.9246694637913434e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009720200952142478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004510258929803968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023950660973787306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9481775820255279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17550, + "weighted_flow_matching_loss": 0.02400908339768648, + "weighted_rkd_loss": 0.022709465958178042 + }, + { + "epoch": 5.109106779167879, + "flow_matching_loss": 0.0231013560667634, + "grad_norm": 0.3265940546989441, + "learning_rate": 2.9245400589473714e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010534676164388657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004658189043402672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02572754118591547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9481195390224457, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17560, + "weighted_flow_matching_loss": 0.0231013560667634, + "weighted_rkd_loss": 0.024392777122557163 + }, + { + "epoch": 5.112016293279022, + "flow_matching_loss": 0.025170831009745598, + "grad_norm": 0.27257537841796875, + "learning_rate": 2.924410545918983e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010331116709858179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00471909788902849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025381331145763398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9480614542961121, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17570, + "weighted_flow_matching_loss": 0.025170831009745598, + "weighted_rkd_loss": 0.024063038639724255 + }, + { + "epoch": 5.114925807390166, + "flow_matching_loss": 0.02446830812841654, + "grad_norm": 0.25451669096946716, + "learning_rate": 2.9242809247160137e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009297478385269642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040703345322981475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022665291279554366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.948003351688385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17580, + "weighted_flow_matching_loss": 0.02446830812841654, + "weighted_rkd_loss": 0.021486753411591054 + }, + { + "epoch": 5.1178353215013095, + "flow_matching_loss": 0.02498395647853613, + "grad_norm": 0.22763840854167938, + "learning_rate": 2.924151195348308e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009462750284001231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004450146714225412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023375647142529488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9479452013969422, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17590, + "weighted_flow_matching_loss": 0.02498395647853613, + "weighted_rkd_loss": 0.022158835269510745 + }, + { + "epoch": 5.120744835612452, + "flow_matching_loss": 0.023193465173244478, + "grad_norm": 0.31189391016960144, + "learning_rate": 2.9240213578257173e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009853214537724853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004525923077017069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024232352338731288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9478870332241058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17600, + "weighted_flow_matching_loss": 0.023193465173244478, + "weighted_rkd_loss": 0.02296954523772001 + }, + { + "epoch": 5.123654349723596, + "flow_matching_loss": 0.024503395706415177, + "grad_norm": 0.23549143970012665, + "learning_rate": 2.9238914121581023e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00912828352302313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004276857757940888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02253342531621456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9478288352489471, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17610, + "weighted_flow_matching_loss": 0.024503395706415177, + "weighted_rkd_loss": 0.021357841975986957 + }, + { + "epoch": 5.12656386383474, + "flow_matching_loss": 0.02395735327154398, + "grad_norm": 0.2534019947052002, + "learning_rate": 2.923761358355331e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009146570041775704, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003969135275110602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02226227559149265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9477705955505371, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17620, + "weighted_flow_matching_loss": 0.02395735327154398, + "weighted_rkd_loss": 0.021099529601633547 + }, + { + "epoch": 5.129473377945883, + "flow_matching_loss": 0.024181557074189185, + "grad_norm": 0.23389893770217896, + "learning_rate": 2.9236311964272814e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009215662442147731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004008627380244434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022439952194690704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9477123320102692, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17630, + "weighted_flow_matching_loss": 0.024181557074189185, + "weighted_rkd_loss": 0.021266622468829155 + }, + { + "epoch": 5.132382892057026, + "flow_matching_loss": 0.023579142056405544, + "grad_norm": 0.2780739665031433, + "learning_rate": 2.9235009263838368e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009085869789123536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004157922510057688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022329661808907986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9476540386676788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17640, + "weighted_flow_matching_loss": 0.023579142056405544, + "weighted_rkd_loss": 0.02116078622639179 + }, + { + "epoch": 5.13529240616817, + "flow_matching_loss": 0.02656095791608095, + "grad_norm": 0.2473905086517334, + "learning_rate": 2.923370548234891e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00893693077377975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003954533557407558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02182839494198561, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9475957155227661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17650, + "weighted_flow_matching_loss": 0.02656095791608095, + "weighted_rkd_loss": 0.020684497617185115 + }, + { + "epoch": 5.138201920279314, + "flow_matching_loss": 0.0232728973031044, + "grad_norm": 0.24328145384788513, + "learning_rate": 2.9232400619903463e-05, + "loss": 0.0464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009990262845531107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004418984754011035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02439951039850712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.947537362575531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17660, + "weighted_flow_matching_loss": 0.0232728973031044, + "weighted_rkd_loss": 0.0231194531545043 + }, + { + "epoch": 5.1411114343904565, + "flow_matching_loss": 0.025177383050322532, + "grad_norm": 0.2703301012516022, + "learning_rate": 2.923109467660111e-05, + "loss": 0.0458, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008847296191379427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040369789814576505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021731571480631828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9474789500236511, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17670, + "weighted_flow_matching_loss": 0.025177383050322532, + "weighted_rkd_loss": 0.020590180717408656 + }, + { + "epoch": 5.1440209485016, + "flow_matching_loss": 0.025644400157034397, + "grad_norm": 0.2302813082933426, + "learning_rate": 2.9229787652541037e-05, + "loss": 0.0465, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009009145153686404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003988845925778151, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02200713623315096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9474205374717712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17680, + "weighted_flow_matching_loss": 0.025644400157034397, + "weighted_rkd_loss": 0.020850011520087718 + }, + { + "epoch": 5.146930462612744, + "flow_matching_loss": 0.024262244813144207, + "grad_norm": 0.24077273905277252, + "learning_rate": 2.9228479547822495e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00869149323552847, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003916542837396264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021299529634416104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9473621129989624, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17690, + "weighted_flow_matching_loss": 0.024262244813144207, + "weighted_rkd_loss": 0.020178352110087873 + }, + { + "epoch": 5.149839976723887, + "flow_matching_loss": 0.02304593678563833, + "grad_norm": 0.2779816687107086, + "learning_rate": 2.9227170362544835e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009746616519987584, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004378395061939955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02387162819504738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9473036289215088, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17700, + "weighted_flow_matching_loss": 0.02304593678563833, + "weighted_rkd_loss": 0.022613673470914364 + }, + { + "epoch": 5.15274949083503, + "flow_matching_loss": 0.022558145970106126, + "grad_norm": 0.23540058732032776, + "learning_rate": 2.9225860096807475e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00933558288961649, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004208804853260517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022879970632493495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9472451269626617, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17710, + "weighted_flow_matching_loss": 0.022558145970106126, + "weighted_rkd_loss": 0.021672916039824484 + }, + { + "epoch": 5.155659004946174, + "flow_matching_loss": 0.023162574134767057, + "grad_norm": 0.21649810671806335, + "learning_rate": 2.922454875070992e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008382892701774836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033959723077714443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020161757618188857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9471865892410278, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17720, + "weighted_flow_matching_loss": 0.023162574134767057, + "weighted_rkd_loss": 0.01909695602953434 + }, + { + "epoch": 5.158568519057318, + "flow_matching_loss": 0.025680916383862495, + "grad_norm": 0.24186132848262787, + "learning_rate": 2.922323632435177e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009435825422406197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00437840458471328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023250055313110352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9471280157566071, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17730, + "weighted_flow_matching_loss": 0.025680916383862495, + "weighted_rkd_loss": 0.022020779363811017 + }, + { + "epoch": 5.161478033168461, + "flow_matching_loss": 0.022737484239041806, + "grad_norm": 0.21791797876358032, + "learning_rate": 2.9221922817832678e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008111042296513915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00361519786529243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019837282411754132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9470694184303283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17740, + "weighted_flow_matching_loss": 0.022737484239041806, + "weighted_rkd_loss": 0.01878730421885848 + }, + { + "epoch": 5.164387547279604, + "flow_matching_loss": 0.024983958527445792, + "grad_norm": 0.25338873267173767, + "learning_rate": 2.922060823125241e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008504365710541606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00401085689663887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02101958841085434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9470107793807984, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17750, + "weighted_flow_matching_loss": 0.024983958527445792, + "weighted_rkd_loss": 0.019905763771384956 + }, + { + "epoch": 5.167297061390748, + "flow_matching_loss": 0.024821629002690315, + "grad_norm": 0.27458298206329346, + "learning_rate": 2.9219292564710792e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010253827832639218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004619245929643512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02512690145522356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9469521164894104, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17760, + "weighted_flow_matching_loss": 0.024821629002690315, + "weighted_rkd_loss": 0.023793975263833998 + }, + { + "epoch": 5.170206575501891, + "flow_matching_loss": 0.024385521560907363, + "grad_norm": 0.26832315325737, + "learning_rate": 2.921797581830774e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010158480331301688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004440741683356464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024757702276110648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9468934237957001, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17770, + "weighted_flow_matching_loss": 0.024385521560907363, + "weighted_rkd_loss": 0.02344291191548109 + }, + { + "epoch": 5.173116089613035, + "flow_matching_loss": 0.024649146012961864, + "grad_norm": 0.32496896386146545, + "learning_rate": 2.9216657992143257e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009719580877572298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004451869498006999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02389103136956692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9468347072601319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17780, + "weighted_flow_matching_loss": 0.024649146012961864, + "weighted_rkd_loss": 0.022620862908661367 + }, + { + "epoch": 5.176025603724178, + "flow_matching_loss": 0.02439586687833071, + "grad_norm": 0.2414943426847458, + "learning_rate": 2.9215339086317426e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008631144044920802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038285626797005534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021090850792825223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9467759549617767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17790, + "weighted_flow_matching_loss": 0.02439586687833071, + "weighted_rkd_loss": 0.019968306832015514 + }, + { + "epoch": 5.178935117835321, + "flow_matching_loss": 0.025798094272613526, + "grad_norm": 0.27549272775650024, + "learning_rate": 2.92140191009304e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009052188461646437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039058557711541654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022010232880711557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9467171609401703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17800, + "weighted_flow_matching_loss": 0.025798094272613526, + "weighted_rkd_loss": 0.02083744890987873 + }, + { + "epoch": 5.181844631946465, + "flow_matching_loss": 0.024939189478754998, + "grad_norm": 0.27542224526405334, + "learning_rate": 2.921269803608243e-05, + "loss": 0.0464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009160175500437618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043211047071963545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022641455940902232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9466583490371704, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17810, + "weighted_flow_matching_loss": 0.024939189478754998, + "weighted_rkd_loss": 0.021433737128973007 + }, + { + "epoch": 5.1847541460576085, + "flow_matching_loss": 0.023843163810670375, + "grad_norm": 0.35927945375442505, + "learning_rate": 2.9211375891873843e-05, + "loss": 0.0461, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009564598416909575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004424028238281608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02355322539806366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9465995013713837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17820, + "weighted_flow_matching_loss": 0.023843163810670375, + "weighted_rkd_loss": 0.0222954117693007 + }, + { + "epoch": 5.187663660168752, + "flow_matching_loss": 0.022712686285376548, + "grad_norm": 0.27121448516845703, + "learning_rate": 2.9210052668405034e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009878829680383205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004546114942058921, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024303774163126946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.946540629863739, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17830, + "weighted_flow_matching_loss": 0.022712686285376548, + "weighted_rkd_loss": 0.023004516400396822 + }, + { + "epoch": 5.190573174279895, + "flow_matching_loss": 0.024081234633922578, + "grad_norm": 0.262495219707489, + "learning_rate": 2.920872836577651e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009035759419202805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003987710014916957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022059228830039503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.946481716632843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17840, + "weighted_flow_matching_loss": 0.024081234633922578, + "weighted_rkd_loss": 0.020878656394779684 + }, + { + "epoch": 5.193482688391039, + "flow_matching_loss": 0.023024290427565576, + "grad_norm": 0.30726391077041626, + "learning_rate": 2.9207402984088834e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008842449681833387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003963381238281727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02164828050881624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9464227855205536, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17850, + "weighted_flow_matching_loss": 0.023024290427565576, + "weighted_rkd_loss": 0.020488403923809528 + }, + { + "epoch": 5.1963922025021825, + "flow_matching_loss": 0.022571099549531938, + "grad_norm": 0.20305192470550537, + "learning_rate": 2.9206076523442664e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009497968107461929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004224601341411472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02322053723037243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9463638246059418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17860, + "weighted_flow_matching_loss": 0.022571099549531938, + "weighted_rkd_loss": 0.021975111216306686 + }, + { + "epoch": 5.199301716613325, + "flow_matching_loss": 0.02407842017710209, + "grad_norm": 0.2525763511657715, + "learning_rate": 2.9204748983938733e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008754183631390333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004064312041737139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02157267928123474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9463048279285431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17870, + "weighted_flow_matching_loss": 0.02407842017710209, + "weighted_rkd_loss": 0.020414322428405286 + }, + { + "epoch": 5.202211230724469, + "flow_matching_loss": 0.023378186114132404, + "grad_norm": 0.2495601326227188, + "learning_rate": 2.920342036567786e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009604895068332554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004328041360713542, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023537831753492354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9462457895278931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17880, + "weighted_flow_matching_loss": 0.023378186114132404, + "weighted_rkd_loss": 0.022272557020187378 + }, + { + "epoch": 5.205120744835613, + "flow_matching_loss": 0.022896575927734374, + "grad_norm": 0.26428744196891785, + "learning_rate": 2.920209066876094e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009674614761024714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044193619163706895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02376859150826931, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9461867272853851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17890, + "weighted_flow_matching_loss": 0.022896575927734374, + "weighted_rkd_loss": 0.022489545680582524 + }, + { + "epoch": 5.2080302589467555, + "flow_matching_loss": 0.023056911677122115, + "grad_norm": 0.25468581914901733, + "learning_rate": 2.9200759893288967e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008828956680372358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003894929401576519, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021552842669188978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9461276352405548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17900, + "weighted_flow_matching_loss": 0.023056911677122115, + "weighted_rkd_loss": 0.020391743630170822 + }, + { + "epoch": 5.210939773057899, + "flow_matching_loss": 0.02290007807314396, + "grad_norm": 0.1761925220489502, + "learning_rate": 2.9199428039362994e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008868108317255975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003758373949676752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021494590677320957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9460685133934021, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17910, + "weighted_flow_matching_loss": 0.02290007807314396, + "weighted_rkd_loss": 0.02033539591357112 + }, + { + "epoch": 5.213849287169043, + "flow_matching_loss": 0.024124172143638135, + "grad_norm": 0.2566893398761749, + "learning_rate": 2.919809510708417e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009236824931576848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004176010866649449, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022649660333991052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9460093677043915, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17920, + "weighted_flow_matching_loss": 0.024124172143638135, + "weighted_rkd_loss": 0.021426787972450255 + }, + { + "epoch": 5.216758801280186, + "flow_matching_loss": 0.02370349448174238, + "grad_norm": 0.2561681568622589, + "learning_rate": 2.919676109655372e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009248159732669592, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004161296505481004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022657615691423418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.945950186252594, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17930, + "weighted_flow_matching_loss": 0.02370349448174238, + "weighted_rkd_loss": 0.021432917471975087 + }, + { + "epoch": 5.2196683153913295, + "flow_matching_loss": 0.023879551328718662, + "grad_norm": 0.2752968370914459, + "learning_rate": 2.9195426007872954e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00938047505915165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004233838990330696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02299478892236948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9458909690380096, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17940, + "weighted_flow_matching_loss": 0.023879551328718662, + "weighted_rkd_loss": 0.021750551462173463 + }, + { + "epoch": 5.222577829502473, + "flow_matching_loss": 0.022959802858531476, + "grad_norm": 0.29756152629852295, + "learning_rate": 2.919408984114327e-05, + "loss": 0.0465, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010158621752634645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00455634156242013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02487358506768942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9458317279815673, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17950, + "weighted_flow_matching_loss": 0.022959802858531476, + "weighted_rkd_loss": 0.02352621927857399 + }, + { + "epoch": 5.225487343613617, + "flow_matching_loss": 0.025724085979163647, + "grad_norm": 0.2627018690109253, + "learning_rate": 2.919275259646613e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009972077701240778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004214034834876657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024158190190792083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9457724511623382, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17960, + "weighted_flow_matching_loss": 0.025724085979163647, + "weighted_rkd_loss": 0.02284813765436411 + }, + { + "epoch": 5.22839685772476, + "flow_matching_loss": 0.02408176548779011, + "grad_norm": 0.27459120750427246, + "learning_rate": 2.9191414273943098e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010035251546651126, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004433238808996975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024503741785883902, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9457131505012513, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17970, + "weighted_flow_matching_loss": 0.02408176548779011, + "weighted_rkd_loss": 0.023173475451767445 + }, + { + "epoch": 5.231306371835903, + "flow_matching_loss": 0.023973876610398293, + "grad_norm": 0.2793929874897003, + "learning_rate": 2.9190074873675802e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008604779746383429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003836148092523217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021045707631856203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9456538081169128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17980, + "weighted_flow_matching_loss": 0.023973876610398293, + "weighted_rkd_loss": 0.01990193100646138 + }, + { + "epoch": 5.234215885947047, + "flow_matching_loss": 0.024916450306773186, + "grad_norm": 0.28873345255851746, + "learning_rate": 2.918873439576597e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009923393605276942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004362720507197082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024209507554769517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9455944478511811, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17990, + "weighted_flow_matching_loss": 0.024916450306773186, + "weighted_rkd_loss": 0.02289239540696144 + }, + { + "epoch": 5.23712540005819, + "flow_matching_loss": 0.023888975195586682, + "grad_norm": 0.22050683200359344, + "learning_rate": 2.91873928403154e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008802529983222484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037325049517676233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021337565034627914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9455350518226624, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18000, + "weighted_flow_matching_loss": 0.023888975195586682, + "weighted_rkd_loss": 0.02017543949186802 + }, + { + "epoch": 5.240034914169334, + "flow_matching_loss": 0.025074292533099653, + "grad_norm": 0.24126824736595154, + "learning_rate": 2.9186050207425966e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011115206126123666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005167978163808584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027398390136659144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9454756259918213, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18010, + "weighted_flow_matching_loss": 0.025074292533099653, + "weighted_rkd_loss": 0.02590450532734394 + }, + { + "epoch": 5.242944428280477, + "flow_matching_loss": 0.02392652090638876, + "grad_norm": 0.22281472384929657, + "learning_rate": 2.918470649719964e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008615055866539479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003994119446724653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021224231086671354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9454161643981933, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18020, + "weighted_flow_matching_loss": 0.02392652090638876, + "weighted_rkd_loss": 0.02006575036793947 + }, + { + "epoch": 5.24585394239162, + "flow_matching_loss": 0.02449106313288212, + "grad_norm": 0.25422918796539307, + "learning_rate": 2.918336170973847e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008805675897747278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003975235903635621, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02158658755943179, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9453566789627075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18030, + "weighted_flow_matching_loss": 0.02449106313288212, + "weighted_rkd_loss": 0.02040698453783989 + }, + { + "epoch": 5.248763456502764, + "flow_matching_loss": 0.022595438174903393, + "grad_norm": 0.2586880326271057, + "learning_rate": 2.918201584514458e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009741557855159044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004255849099718034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023738964460790157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9452971577644348, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18040, + "weighted_flow_matching_loss": 0.022595438174903393, + "weighted_rkd_loss": 0.0224403427913785 + }, + { + "epoch": 5.251672970613908, + "flow_matching_loss": 0.022697314620018005, + "grad_norm": 0.23287932574748993, + "learning_rate": 2.9180668903520178e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008783880807459354, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004101174511015415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021668936125934124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9452376127243042, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18050, + "weighted_flow_matching_loss": 0.022697314620018005, + "weighted_rkd_loss": 0.020482301153242587 + }, + { + "epoch": 5.254582484725051, + "flow_matching_loss": 0.02557747606188059, + "grad_norm": 0.21965733170509338, + "learning_rate": 2.917932088496756e-05, + "loss": 0.0465, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009100543381646276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039784392109140755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022179525904357435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9451780319213867, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18060, + "weighted_flow_matching_loss": 0.02557747606188059, + "weighted_rkd_loss": 0.020963621325790882 + }, + { + "epoch": 5.257491998836194, + "flow_matching_loss": 0.024158182181417943, + "grad_norm": 0.2563323378562927, + "learning_rate": 2.9177971789589092e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008431365061551333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003992165555246174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020854895934462547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9451184272766113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18070, + "weighted_flow_matching_loss": 0.024158182181417943, + "weighted_rkd_loss": 0.019710351061075927 + }, + { + "epoch": 5.260401512947338, + "flow_matching_loss": 0.024205922335386276, + "grad_norm": 0.24592304229736328, + "learning_rate": 2.9176621617487237e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007802845584228635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035427465103566647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019148437678813933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9450587809085846, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18080, + "weighted_flow_matching_loss": 0.024205922335386276, + "weighted_rkd_loss": 0.01809643628075719 + }, + { + "epoch": 5.2633110270584815, + "flow_matching_loss": 0.024744217097759248, + "grad_norm": 0.3323797583580017, + "learning_rate": 2.917527036876452e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008905027760192752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003852508426643908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021662563644349576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9449991166591645, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18090, + "weighted_flow_matching_loss": 0.024744217097759248, + "weighted_rkd_loss": 0.02047113422304392 + }, + { + "epoch": 5.266220541169624, + "flow_matching_loss": 0.02186084296554327, + "grad_norm": 0.25463345646858215, + "learning_rate": 2.9173918043523573e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009910230757668614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004491908778436482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024312370084226133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9449394106864929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18100, + "weighted_flow_matching_loss": 0.02186084296554327, + "weighted_rkd_loss": 0.022973719611763956 + }, + { + "epoch": 5.269130055280768, + "flow_matching_loss": 0.025868583284318448, + "grad_norm": 0.24337153136730194, + "learning_rate": 2.9172564641867094e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009377016592770815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004210471617989242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022964504733681678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9448796808719635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18110, + "weighted_flow_matching_loss": 0.025868583284318448, + "weighted_rkd_loss": 0.021698706597089768 + }, + { + "epoch": 5.272039569391912, + "flow_matching_loss": 0.024005996994674206, + "grad_norm": 0.28941410779953003, + "learning_rate": 2.9171210163897857e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009124120511114598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003697274951264262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021945516206324102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9448199152946473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18120, + "weighted_flow_matching_loss": 0.024005996994674206, + "weighted_rkd_loss": 0.020734542235732078 + }, + { + "epoch": 5.274949083503055, + "flow_matching_loss": 0.023774625919759273, + "grad_norm": 0.2649575471878052, + "learning_rate": 2.9169854609718728e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00984020191244781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004436330287717283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02411673404276371, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.944760125875473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18130, + "weighted_flow_matching_loss": 0.023774625919759273, + "weighted_rkd_loss": 0.022784528322517872 + }, + { + "epoch": 5.277858597614198, + "flow_matching_loss": 0.02565637119114399, + "grad_norm": 0.21784166991710663, + "learning_rate": 2.9168497979432656e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009217284619808197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004404072323814035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0228386415168643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.944700300693512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18140, + "weighted_flow_matching_loss": 0.02565637119114399, + "weighted_rkd_loss": 0.02157566137611866 + }, + { + "epoch": 5.280768111725342, + "flow_matching_loss": 0.022639372758567332, + "grad_norm": 0.33543410897254944, + "learning_rate": 2.9167140273142667e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009615358989685774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00445246109738946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02368317898362875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9446404457092286, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18150, + "weighted_flow_matching_loss": 0.022639372758567332, + "weighted_rkd_loss": 0.022372041270136832 + }, + { + "epoch": 5.283677625836486, + "flow_matching_loss": 0.026478786394000052, + "grad_norm": 0.27251914143562317, + "learning_rate": 2.916578149095187e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008761892700567842, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040456305490806695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02156941592693329, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9445805549621582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18160, + "weighted_flow_matching_loss": 0.026478786394000052, + "weighted_rkd_loss": 0.020374013856053353 + }, + { + "epoch": 5.2865871399476285, + "flow_matching_loss": 0.02906488999724388, + "grad_norm": 0.23535104095935822, + "learning_rate": 2.916442163296345e-05, + "loss": 0.0504, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009261997090652585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040887143230065705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022612708434462548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9445206463336945, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18170, + "weighted_flow_matching_loss": 0.02906488999724388, + "weighted_rkd_loss": 0.02135818973183632 + }, + { + "epoch": 5.289496654058772, + "flow_matching_loss": 0.022971959598362447, + "grad_norm": 0.265487939119339, + "learning_rate": 2.9163060699280687e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00925957141444087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004092242778278887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02261138577014208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9444607019424438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18180, + "weighted_flow_matching_loss": 0.022971959598362447, + "weighted_rkd_loss": 0.0213555789552629 + }, + { + "epoch": 5.292406168169916, + "flow_matching_loss": 0.02331591844558716, + "grad_norm": 0.2268509864807129, + "learning_rate": 2.9161698690006937e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009427342051640152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004226984572596848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02308166865259409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9444007277488708, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18190, + "weighted_flow_matching_loss": 0.02331591844558716, + "weighted_rkd_loss": 0.021798383072018625 + }, + { + "epoch": 5.295315682281059, + "flow_matching_loss": 0.024650788679718972, + "grad_norm": 0.2597733438014984, + "learning_rate": 2.9160335605245626e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009061963204294443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004157285112887621, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022281211521476506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9443407237529755, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18200, + "weighted_flow_matching_loss": 0.024650788679718972, + "weighted_rkd_loss": 0.02104108165949583 + }, + { + "epoch": 5.2982251963922025, + "flow_matching_loss": 0.025978836603462695, + "grad_norm": 0.2240157425403595, + "learning_rate": 2.9158971445100276e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009186334302648902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004303578450344503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0226762467995286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9442806839942932, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18210, + "weighted_flow_matching_loss": 0.025978836603462695, + "weighted_rkd_loss": 0.021412788331508635 + }, + { + "epoch": 5.301134710503346, + "flow_matching_loss": 0.024994576908648015, + "grad_norm": 0.21979369223117828, + "learning_rate": 2.9157606209674485e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00885098623111844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038799671223387123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021581939421594143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9442206144332885, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18220, + "weighted_flow_matching_loss": 0.024994576908648015, + "weighted_rkd_loss": 0.02037813849747181 + }, + { + "epoch": 5.304044224614489, + "flow_matching_loss": 0.02338547511026263, + "grad_norm": 0.25873279571533203, + "learning_rate": 2.9156239899071935e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009986090939491987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004318086244165897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02429026775062084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.944160521030426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18230, + "weighted_flow_matching_loss": 0.02338547511026263, + "weighted_rkd_loss": 0.02293393649160862 + }, + { + "epoch": 5.306953738725633, + "flow_matching_loss": 0.024327893741428852, + "grad_norm": 0.22792983055114746, + "learning_rate": 2.9154872513396388e-05, + "loss": 0.0458, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009284248808398843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041915741050615905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02276007141917944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9441003918647766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18240, + "weighted_flow_matching_loss": 0.024327893741428852, + "weighted_rkd_loss": 0.02148782052099705 + }, + { + "epoch": 5.309863252836776, + "flow_matching_loss": 0.024080158583819865, + "grad_norm": 0.25392481684684753, + "learning_rate": 2.915350405275169e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008980432711541652, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004066254058852792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022027119621634485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9440402269363404, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18250, + "weighted_flow_matching_loss": 0.024080158583819865, + "weighted_rkd_loss": 0.020794440060853958 + }, + { + "epoch": 5.31277276694792, + "flow_matching_loss": 0.02510076817125082, + "grad_norm": 0.22465303540229797, + "learning_rate": 2.9152134517241758e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009951809188351035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004420909378677606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024324527941644192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9439800679683685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18260, + "weighted_flow_matching_loss": 0.02510076817125082, + "weighted_rkd_loss": 0.02296189982444048 + }, + { + "epoch": 5.315682281059063, + "flow_matching_loss": 0.02308919169008732, + "grad_norm": 0.4182365834712982, + "learning_rate": 2.9150763906970612e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009734124410897493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00443843468092382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02390668299049139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9439198136329651, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18270, + "weighted_flow_matching_loss": 0.02308919169008732, + "weighted_rkd_loss": 0.02256596852093935 + }, + { + "epoch": 5.318591795170207, + "flow_matching_loss": 0.024160216189920904, + "grad_norm": 0.2313491553068161, + "learning_rate": 2.9149392222042332e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008507654769346118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00366799533367157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020683304779231547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9438595771789551, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18280, + "weighted_flow_matching_loss": 0.024160216189920904, + "weighted_rkd_loss": 0.019522152189165354 + }, + { + "epoch": 5.32150130928135, + "flow_matching_loss": 0.025604036450386048, + "grad_norm": 0.3289383053779602, + "learning_rate": 2.9148019462561092e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00946664223447442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004425490624271333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023358774930238725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9437992990016937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18290, + "weighted_flow_matching_loss": 0.025604036450386048, + "weighted_rkd_loss": 0.022045956552028657 + }, + { + "epoch": 5.324410823392493, + "flow_matching_loss": 0.023520727828145026, + "grad_norm": 0.2721652686595917, + "learning_rate": 2.914664562863114e-05, + "loss": 0.0453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009424073435366154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042588143609464165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023106961138546468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9437389850616456, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18300, + "weighted_flow_matching_loss": 0.023520727828145026, + "weighted_rkd_loss": 0.021806934289634226 + }, + { + "epoch": 5.327320337503637, + "flow_matching_loss": 0.02261964324861765, + "grad_norm": 0.29943233728408813, + "learning_rate": 2.9145270720356813e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00914603448472917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004055566131137311, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022347635217010976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9436786472797394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18310, + "weighted_flow_matching_loss": 0.02261964324861765, + "weighted_rkd_loss": 0.021088984422385694 + }, + { + "epoch": 5.330229851614781, + "flow_matching_loss": 0.025097507424652575, + "grad_norm": 0.22836297750473022, + "learning_rate": 2.9143894737842524e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008959841076284647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004230096540413797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02214977890253067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9436182737350464, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18320, + "weighted_flow_matching_loss": 0.025097507424652575, + "weighted_rkd_loss": 0.020900943875312807 + }, + { + "epoch": 5.333139365725923, + "flow_matching_loss": 0.023554090596735477, + "grad_norm": 0.25339099764823914, + "learning_rate": 2.914251768119277e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010291993105784058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004376188106834889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024960174225270748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.943557870388031, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18330, + "weighted_flow_matching_loss": 0.023554090596735477, + "weighted_rkd_loss": 0.02355137886479497 + }, + { + "epoch": 5.336048879837067, + "flow_matching_loss": 0.024677231162786483, + "grad_norm": 0.23772136867046356, + "learning_rate": 2.9141139550512135e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008817423600703478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004011418507434428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02164626568555832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9434974372386933, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18340, + "weighted_flow_matching_loss": 0.024677231162786483, + "weighted_rkd_loss": 0.020423206966370344 + }, + { + "epoch": 5.338958393948211, + "flow_matching_loss": 0.0222161965444684, + "grad_norm": 0.2610177993774414, + "learning_rate": 2.9139760345905273e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008452240051701665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037367973709478974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020641277357935906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9434369802474976, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18350, + "weighted_flow_matching_loss": 0.0222161965444684, + "weighted_rkd_loss": 0.019473722949624062 + }, + { + "epoch": 5.3418679080593545, + "flow_matching_loss": 0.024098739586770536, + "grad_norm": 0.2310311496257782, + "learning_rate": 2.9138380067476925e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009402355598285794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004131266218610108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022935977764427663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9433764815330505, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18360, + "weighted_flow_matching_loss": 0.024098739586770536, + "weighted_rkd_loss": 0.021637261286377905 + }, + { + "epoch": 5.344777422170497, + "flow_matching_loss": 0.02139438856393099, + "grad_norm": 0.21966706216335297, + "learning_rate": 2.913699871533192e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008072602655738592, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00342590119689703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01957110669463873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.94331596493721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18370, + "weighted_flow_matching_loss": 0.02139438856393099, + "weighted_rkd_loss": 0.018461748398840426 + }, + { + "epoch": 5.347686936281641, + "flow_matching_loss": 0.024490374140441418, + "grad_norm": 0.2594243586063385, + "learning_rate": 2.9135616289575157e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00837417710572481, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037416874431073664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020490041561424734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9432554066181182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18380, + "weighted_flow_matching_loss": 0.024490374140441418, + "weighted_rkd_loss": 0.01932732928544283 + }, + { + "epoch": 5.350596450392785, + "flow_matching_loss": 0.023343441449105738, + "grad_norm": 0.2929975092411041, + "learning_rate": 2.9134232790311624e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009304606867954136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004316229000687599, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02292544301599264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9431948184967041, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18390, + "weighted_flow_matching_loss": 0.023343441449105738, + "weighted_rkd_loss": 0.02162313861772418 + }, + { + "epoch": 5.353505964503928, + "flow_matching_loss": 0.024712025374174117, + "grad_norm": 0.2553819715976715, + "learning_rate": 2.9132848217646388e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00867804461158812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037793059833347797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02113539520651102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9431342124938965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18400, + "weighted_flow_matching_loss": 0.024712025374174117, + "weighted_rkd_loss": 0.019933515414595603 + }, + { + "epoch": 5.356415478615071, + "flow_matching_loss": 0.024396041221916676, + "grad_norm": 0.237448051571846, + "learning_rate": 2.91314625716846e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009958831639960409, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004367311135865748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024284974299371244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9430735647678375, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18410, + "weighted_flow_matching_loss": 0.024396041221916676, + "weighted_rkd_loss": 0.022902514785528183 + }, + { + "epoch": 5.359324992726215, + "flow_matching_loss": 0.024670590460300446, + "grad_norm": 0.3091496229171753, + "learning_rate": 2.9130075852531484e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0096783893648535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004518297174945474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023875075951218606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9430128872394562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18420, + "weighted_flow_matching_loss": 0.024670590460300446, + "weighted_rkd_loss": 0.02251446321606636 + }, + { + "epoch": 5.362234506837358, + "flow_matching_loss": 0.023529704287648202, + "grad_norm": 0.21207626163959503, + "learning_rate": 2.9128688060292363e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008730010129511356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039653453510254625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02142536547034979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9429521858692169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18430, + "weighted_flow_matching_loss": 0.023529704287648202, + "weighted_rkd_loss": 0.02020312249660492 + }, + { + "epoch": 5.3651440209485015, + "flow_matching_loss": 0.023396276123821735, + "grad_norm": 0.2514120936393738, + "learning_rate": 2.912729919507262e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008654599124565721, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003856232529506087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021165430918335913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9428914487361908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18440, + "weighted_flow_matching_loss": 0.023396276123821735, + "weighted_rkd_loss": 0.019956659898161887 + }, + { + "epoch": 5.368053535059645, + "flow_matching_loss": 0.02288521695882082, + "grad_norm": 0.2300175130367279, + "learning_rate": 2.912590925697774e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009132717177271843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004007654543966055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02227308861911297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9428306818008423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18450, + "weighted_flow_matching_loss": 0.02288521695882082, + "weighted_rkd_loss": 0.020999769493937493 + }, + { + "epoch": 5.370963049170788, + "flow_matching_loss": 0.024340006150305273, + "grad_norm": 0.28852754831314087, + "learning_rate": 2.9124518246113274e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010061501106247307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044933858327567576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024616387858986854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9427698850631714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18460, + "weighted_flow_matching_loss": 0.024340006150305273, + "weighted_rkd_loss": 0.023207563161849975 + }, + { + "epoch": 5.373872563281932, + "flow_matching_loss": 0.021923836320638657, + "grad_norm": 0.23941828310489655, + "learning_rate": 2.9123126162584857e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009046109765768051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042482304852455854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022340450063347818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9427090585231781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18470, + "weighted_flow_matching_loss": 0.021923836320638657, + "weighted_rkd_loss": 0.02106054462492466 + }, + { + "epoch": 5.3767820773930755, + "flow_matching_loss": 0.02326583582907915, + "grad_norm": 0.2443697303533554, + "learning_rate": 2.9121733006498216e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008932919474318624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00399096286855638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021856801770627498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9426482021808624, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18480, + "weighted_flow_matching_loss": 0.02326583582907915, + "weighted_rkd_loss": 0.020603256952017545 + }, + { + "epoch": 5.379691591504219, + "flow_matching_loss": 0.02646271288394928, + "grad_norm": 0.2751830220222473, + "learning_rate": 2.912033877795915e-05, + "loss": 0.0464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008694738941267132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037987745134159924, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021188252419233323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9425873160362244, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18490, + "weighted_flow_matching_loss": 0.02646271288394928, + "weighted_rkd_loss": 0.01997177116572857 + }, + { + "epoch": 5.382601105615362, + "flow_matching_loss": 0.026255072839558125, + "grad_norm": 0.26403409242630005, + "learning_rate": 2.9118943477073538e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009420724771916866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042373720090836285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023078821413218976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9425264000892639, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18500, + "weighted_flow_matching_loss": 0.026255072839558125, + "weighted_rkd_loss": 0.02175240349024534 + }, + { + "epoch": 5.385510619726506, + "flow_matching_loss": 0.023446254804730414, + "grad_norm": 0.22031739354133606, + "learning_rate": 2.9117547103947343e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009627243690192699, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004613070096820593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023867557197809218, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9424654483795166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18510, + "weighted_flow_matching_loss": 0.023446254804730414, + "weighted_rkd_loss": 0.022494393028318883 + }, + { + "epoch": 5.388420133837649, + "flow_matching_loss": 0.023474783450365067, + "grad_norm": 0.24687470495700836, + "learning_rate": 2.9116149658686614e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008937679696828127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004013982694596052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02188934227451682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9424044728279114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18520, + "weighted_flow_matching_loss": 0.023474783450365067, + "weighted_rkd_loss": 0.02062859581783414 + }, + { + "epoch": 5.391329647948792, + "flow_matching_loss": 0.025336614437401296, + "grad_norm": 0.2569720149040222, + "learning_rate": 2.911475114139748e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009790846658870579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00447410368360579, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024055797047913073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9423434555530548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18530, + "weighted_flow_matching_loss": 0.025336614437401296, + "weighted_rkd_loss": 0.022668871842324734 + }, + { + "epoch": 5.394239162059936, + "flow_matching_loss": 0.02409585453569889, + "grad_norm": 0.2553357779979706, + "learning_rate": 2.9113351552186146e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008913748385384678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004101656703278422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021929153054952622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9422824263572693, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18540, + "weighted_flow_matching_loss": 0.02409585453569889, + "weighted_rkd_loss": 0.020663467608392237 + }, + { + "epoch": 5.39714867617108, + "flow_matching_loss": 0.023923554085195066, + "grad_norm": 0.21759656071662903, + "learning_rate": 2.91119508911589e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009308987949043512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00420604096725583, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022824016585946082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9422213435173035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18550, + "weighted_flow_matching_loss": 0.023923554085195066, + "weighted_rkd_loss": 0.021505281515419482 + }, + { + "epoch": 5.400058190282223, + "flow_matching_loss": 0.02399436477571726, + "grad_norm": 0.2403624951839447, + "learning_rate": 2.9110549158422116e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009030940849334002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003788350708782673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021850232034921646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9421602487564087, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18560, + "weighted_flow_matching_loss": 0.02399436477571726, + "weighted_rkd_loss": 0.02058640755712986 + }, + { + "epoch": 5.402967704393366, + "flow_matching_loss": 0.026132308505475522, + "grad_norm": 0.23278997838497162, + "learning_rate": 2.9109146354082243e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009133083326742054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041763505432754755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022442517057061195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9420991182327271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18570, + "weighted_flow_matching_loss": 0.026132308505475522, + "weighted_rkd_loss": 0.021143099293112755 + }, + { + "epoch": 5.40587721850451, + "flow_matching_loss": 0.02171245962381363, + "grad_norm": 0.24074812233448029, + "learning_rate": 2.9107742478245818e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0089166100602597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003845186228863895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02167840637266636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9420379579067231, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18580, + "weighted_flow_matching_loss": 0.02171245962381363, + "weighted_rkd_loss": 0.02042186800390482 + }, + { + "epoch": 5.408786732615654, + "flow_matching_loss": 0.023712733760476112, + "grad_norm": 0.3252512812614441, + "learning_rate": 2.9106337531019453e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008914419775828719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003991981828585267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021820821426808835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9419767677783966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18590, + "weighted_flow_matching_loss": 0.023712733760476112, + "weighted_rkd_loss": 0.020554713252931834 + }, + { + "epoch": 5.411696246726796, + "flow_matching_loss": 0.02442338764667511, + "grad_norm": 0.2648021876811981, + "learning_rate": 2.9104931512509846e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008908678311854601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003937673568725586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021755030378699303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9419155478477478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18600, + "weighted_flow_matching_loss": 0.02442338764667511, + "weighted_rkd_loss": 0.020491386763751507 + }, + { + "epoch": 5.41460576083794, + "flow_matching_loss": 0.02449138928204775, + "grad_norm": 0.2181805521249771, + "learning_rate": 2.910352442282378e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008591650100424886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003737364523112774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020920664817094804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9418542981147766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18610, + "weighted_flow_matching_loss": 0.02449138928204775, + "weighted_rkd_loss": 0.019704232830554246 + }, + { + "epoch": 5.417515274949084, + "flow_matching_loss": 0.026097685284912586, + "grad_norm": 0.224101260304451, + "learning_rate": 2.9102116262068105e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009181018359959126, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00385015660431236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02221219353377819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.941793018579483, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18620, + "weighted_flow_matching_loss": 0.026097685284912586, + "weighted_rkd_loss": 0.02091932259500027 + }, + { + "epoch": 5.420424789060227, + "flow_matching_loss": 0.023425758443772794, + "grad_norm": 0.2603244483470917, + "learning_rate": 2.9100707030349766e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009137210389599203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003989147394895554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022263568267226218, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9417317032814025, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18630, + "weighted_flow_matching_loss": 0.023425758443772794, + "weighted_rkd_loss": 0.02096630223095417 + }, + { + "epoch": 5.42333430317137, + "flow_matching_loss": 0.024421672523021697, + "grad_norm": 0.2549391984939575, + "learning_rate": 2.9099296727775784e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00894277635961771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040619269013404844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02194747989997268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9416703641414642, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18640, + "weighted_flow_matching_loss": 0.024421672523021697, + "weighted_rkd_loss": 0.020667336508631708 + }, + { + "epoch": 5.426243817282514, + "flow_matching_loss": 0.02267857491970062, + "grad_norm": 0.24548891186714172, + "learning_rate": 2.9097885354453263e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008908932702615857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004042498581111431, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021860364079475402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9416089951992035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18650, + "weighted_flow_matching_loss": 0.02267857491970062, + "weighted_rkd_loss": 0.020583920180797577 + }, + { + "epoch": 5.429153331393657, + "flow_matching_loss": 0.02206265777349472, + "grad_norm": 0.2899957299232483, + "learning_rate": 2.9096472910489388e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008831538306549192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003791551012545824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021454627532511948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9415476024150848, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18660, + "weighted_flow_matching_loss": 0.02206265777349472, + "weighted_rkd_loss": 0.0202005498111248 + }, + { + "epoch": 5.432062845504801, + "flow_matching_loss": 0.023492242954671384, + "grad_norm": 0.27686387300491333, + "learning_rate": 2.9095059395991426e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008445482375100255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00394631689414382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02083728164434433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9414861500263214, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18670, + "weighted_flow_matching_loss": 0.023492242954671384, + "weighted_rkd_loss": 0.019617998227477073 + }, + { + "epoch": 5.434972359615944, + "flow_matching_loss": 0.023950515687465666, + "grad_norm": 0.22682353854179382, + "learning_rate": 2.909364481106672e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009313058294355869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004136712546460331, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022762829437851906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9414246857166291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18680, + "weighted_flow_matching_loss": 0.023950515687465666, + "weighted_rkd_loss": 0.02142952010035515 + }, + { + "epoch": 5.437881873727088, + "flow_matching_loss": 0.02340331096202135, + "grad_norm": 0.24881528317928314, + "learning_rate": 2.9092229155822704e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00928782601840794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004535476444289088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023111128248274326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9413631916046142, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18690, + "weighted_flow_matching_loss": 0.02340331096202135, + "weighted_rkd_loss": 0.02175601478666067 + }, + { + "epoch": 5.440791387838231, + "flow_matching_loss": 0.02425830829888582, + "grad_norm": 0.2569337785243988, + "learning_rate": 2.9090812430366882e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008471901528537273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003950998629443347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020894801802933217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9413016736507416, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18700, + "weighted_flow_matching_loss": 0.02425830829888582, + "weighted_rkd_loss": 0.01966829998418689 + }, + { + "epoch": 5.4437009019493745, + "flow_matching_loss": 0.023530929163098336, + "grad_norm": 0.24575266242027283, + "learning_rate": 2.9089394634806847e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008609782997518778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038189235841855408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021038489416241646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9412401199340821, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18710, + "weighted_flow_matching_loss": 0.023530929163098336, + "weighted_rkd_loss": 0.019802263099700214 + }, + { + "epoch": 5.446610416060518, + "flow_matching_loss": 0.026136592403054236, + "grad_norm": 0.23416255414485931, + "learning_rate": 2.908797576925027e-05, + "loss": 0.0465, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008895265404134989, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038594468729570507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021649977564811705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9411785364151001, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18720, + "weighted_flow_matching_loss": 0.026136592403054236, + "weighted_rkd_loss": 0.02037650961428881 + }, + { + "epoch": 5.449519930171661, + "flow_matching_loss": 0.024518525041639806, + "grad_norm": 0.2585889995098114, + "learning_rate": 2.9086555833804905e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008782766293734312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038989952998235823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02146452758461237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9411169230937958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18730, + "weighted_flow_matching_loss": 0.024518525041639806, + "weighted_rkd_loss": 0.020200652815401555 + }, + { + "epoch": 5.452429444282805, + "flow_matching_loss": 0.02542017735540867, + "grad_norm": 0.2365483045578003, + "learning_rate": 2.908513482857859e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009609643835574389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004441182548180223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02366047017276287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9410552740097046, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18740, + "weighted_flow_matching_loss": 0.02542017735540867, + "weighted_rkd_loss": 0.022265856526792048 + }, + { + "epoch": 5.4553389583939484, + "flow_matching_loss": 0.023590815253555775, + "grad_norm": 0.25845566391944885, + "learning_rate": 2.908371275367924e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008768441807478666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003871421772055328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021408305130898954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.94099360704422, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18750, + "weighted_flow_matching_loss": 0.023590815253555775, + "weighted_rkd_loss": 0.020145076606422663 + }, + { + "epoch": 5.458248472505091, + "flow_matching_loss": 0.025067944079637527, + "grad_norm": 0.37577635049819946, + "learning_rate": 2.908228960921485e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010228656046092511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004607238038443029, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0250645499676466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.940931898355484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18760, + "weighted_flow_matching_loss": 0.025067944079637527, + "weighted_rkd_loss": 0.02358400970697403 + }, + { + "epoch": 5.461157986616235, + "flow_matching_loss": 0.025317589566111565, + "grad_norm": 0.3897009491920471, + "learning_rate": 2.9080865395293494e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009054797422140836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00413327703718096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02224287185817957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9408701598644257, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18770, + "weighted_flow_matching_loss": 0.025317589566111565, + "weighted_rkd_loss": 0.020927638188004492 + }, + { + "epoch": 5.464067500727379, + "flow_matching_loss": 0.024288037978112697, + "grad_norm": 0.2718219459056854, + "learning_rate": 2.907944011202334e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00912285135127604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004217026080004871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0224627286195755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9408083915710449, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18780, + "weighted_flow_matching_loss": 0.024288037978112697, + "weighted_rkd_loss": 0.021133145317435265 + }, + { + "epoch": 5.4669770148385215, + "flow_matching_loss": 0.024291751347482203, + "grad_norm": 0.23544608056545258, + "learning_rate": 2.9078013759512628e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009415486780926585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004250966478139162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02308194022625685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9407465994358063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18790, + "weighted_flow_matching_loss": 0.024291751347482203, + "weighted_rkd_loss": 0.021714274398982526 + }, + { + "epoch": 5.469886528949665, + "flow_matching_loss": 0.025099402666091918, + "grad_norm": 0.2957160472869873, + "learning_rate": 2.907658633786968e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008988334704190493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003969914931803942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021946584433317186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9406847715377807, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18800, + "weighted_flow_matching_loss": 0.025099402666091918, + "weighted_rkd_loss": 0.020644799061119556 + }, + { + "epoch": 5.472796043060809, + "flow_matching_loss": 0.02131858207285404, + "grad_norm": 0.24851734936237335, + "learning_rate": 2.907515784720289e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008781627891585231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003829811606556177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021393067575991152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9406229197978974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18810, + "weighted_flow_matching_loss": 0.02131858207285404, + "weighted_rkd_loss": 0.020122834760695697 + }, + { + "epoch": 5.475705557171953, + "flow_matching_loss": 0.024886402674019335, + "grad_norm": 0.2527907192707062, + "learning_rate": 2.9073728287620753e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00963876722380519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004363934975117445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023641469515860082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.940561032295227, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18820, + "weighted_flow_matching_loss": 0.024886402674019335, + "weighted_rkd_loss": 0.02223623637109995 + }, + { + "epoch": 5.478615071283095, + "flow_matching_loss": 0.023281477577984332, + "grad_norm": 0.26846644282341003, + "learning_rate": 2.9072297659231832e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008880327548831701, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004159687110222876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021920342370867728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9404991149902344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18830, + "weighted_flow_matching_loss": 0.023281477577984332, + "weighted_rkd_loss": 0.02061604559421539 + }, + { + "epoch": 5.481524585394239, + "flow_matching_loss": 0.026577320136129855, + "grad_norm": 0.2440614104270935, + "learning_rate": 2.9070865962144773e-05, + "loss": 0.0482, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009412526525557042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004162779776379466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02298783287405968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9404371678829193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18840, + "weighted_flow_matching_loss": 0.026577320136129855, + "weighted_rkd_loss": 0.021618621051311494 + }, + { + "epoch": 5.484434099505383, + "flow_matching_loss": 0.024587521329522133, + "grad_norm": 0.2986297905445099, + "learning_rate": 2.90694331964683e-05, + "loss": 0.0453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008957165433093906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004097678954713047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022012009844183923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9403752088546753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18850, + "weighted_flow_matching_loss": 0.024587521329522133, + "weighted_rkd_loss": 0.02069955999031663 + }, + { + "epoch": 5.487343613616526, + "flow_matching_loss": 0.024130858667194843, + "grad_norm": 0.30673128366470337, + "learning_rate": 2.906799936231123e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009893684228882194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004617065819911659, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024404434859752654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9403131783008576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18860, + "weighted_flow_matching_loss": 0.024130858667194843, + "weighted_rkd_loss": 0.022947834990918636 + }, + { + "epoch": 5.490253127727669, + "flow_matching_loss": 0.024023936688899995, + "grad_norm": 0.22472678124904633, + "learning_rate": 2.9066564459782445e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009410719387233257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004236360196955502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02305779904127121, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9402511477470398, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18870, + "weighted_flow_matching_loss": 0.024023936688899995, + "weighted_rkd_loss": 0.021680127084255218 + }, + { + "epoch": 5.493162641838813, + "flow_matching_loss": 0.023160367831587793, + "grad_norm": 0.25303077697753906, + "learning_rate": 2.9065128488990925e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00935438582673669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004127798974514008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022836570627987385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9401890754699707, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18880, + "weighted_flow_matching_loss": 0.023160367831587793, + "weighted_rkd_loss": 0.021470683440566064 + }, + { + "epoch": 5.496072155949957, + "flow_matching_loss": 0.02426058575510979, + "grad_norm": 0.2237338125705719, + "learning_rate": 2.9063691450045718e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008498470392078161, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004148152959533036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021145093813538553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9401269793510437, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18890, + "weighted_flow_matching_loss": 0.02426058575510979, + "weighted_rkd_loss": 0.019879091531038284 + }, + { + "epoch": 5.4989816700611, + "flow_matching_loss": 0.021726826392114162, + "grad_norm": 0.29413560032844543, + "learning_rate": 2.9062253343055956e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008831807039678096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003669512690976262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021333126723766326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9400648474693298, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18900, + "weighted_flow_matching_loss": 0.021726826392114162, + "weighted_rkd_loss": 0.020054500736296176 + }, + { + "epoch": 5.501891184172243, + "flow_matching_loss": 0.024709876254200937, + "grad_norm": 0.32673126459121704, + "learning_rate": 2.9060814168130852e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009396069403737783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041419472079724075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0229340860620141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.940002691745758, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18910, + "weighted_flow_matching_loss": 0.024709876254200937, + "weighted_rkd_loss": 0.021558073349297047 + }, + { + "epoch": 5.504800698283387, + "flow_matching_loss": 0.02352658472955227, + "grad_norm": 0.21694505214691162, + "learning_rate": 2.905937392537971e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008612099383026362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037837088806554676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02100790776312351, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9399405002593995, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18920, + "weighted_flow_matching_loss": 0.02352658472955227, + "weighted_rkd_loss": 0.019746200274676086 + }, + { + "epoch": 5.50771021239453, + "flow_matching_loss": 0.02663335856050253, + "grad_norm": 0.23547039926052094, + "learning_rate": 2.9057932614911902e-05, + "loss": 0.0479, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009276518365368247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004083246341906488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022636283189058304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9398782849311829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18930, + "weighted_flow_matching_loss": 0.02663335856050253, + "weighted_rkd_loss": 0.02127536665648222 + }, + { + "epoch": 5.5106197265056736, + "flow_matching_loss": 0.02440497577190399, + "grad_norm": 0.2628583610057831, + "learning_rate": 2.9056490236836885e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008674921933561563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003666793927550316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02101663760840893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9398160398006439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18940, + "weighted_flow_matching_loss": 0.02440497577190399, + "weighted_rkd_loss": 0.01975175617262721 + }, + { + "epoch": 5.513529240616817, + "flow_matching_loss": 0.024377612583339214, + "grad_norm": 0.32091888785362244, + "learning_rate": 2.9055046791264194e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009453157149255275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004102919297292828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023009233362972736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9397537648677826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18950, + "weighted_flow_matching_loss": 0.024377612583339214, + "weighted_rkd_loss": 0.021623038128018378 + }, + { + "epoch": 5.51643875472796, + "flow_matching_loss": 0.02418796718120575, + "grad_norm": 0.2566107213497162, + "learning_rate": 2.905360227830346e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009461613837629557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004301272449083626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023224500380456446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9396914482116699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18960, + "weighted_flow_matching_loss": 0.02418796718120575, + "weighted_rkd_loss": 0.021823879703879355 + }, + { + "epoch": 5.519348268839104, + "flow_matching_loss": 0.021212369948625565, + "grad_norm": 0.2763103246688843, + "learning_rate": 2.9052156698064384e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008763543842360378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004124303651042283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02165139131247997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9396291136741638, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18970, + "weighted_flow_matching_loss": 0.021212369948625565, + "weighted_rkd_loss": 0.020344278030097484 + }, + { + "epoch": 5.5222577829502475, + "flow_matching_loss": 0.02307151947170496, + "grad_norm": 0.2139436900615692, + "learning_rate": 2.9050710050656734e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008709182031452656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003893571742810309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021311935782432557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9395667493343354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18980, + "weighted_flow_matching_loss": 0.02307151947170496, + "weighted_rkd_loss": 0.020023988280445337 + }, + { + "epoch": 5.52516729706139, + "flow_matching_loss": 0.023127080406993628, + "grad_norm": 0.25223496556282043, + "learning_rate": 2.9049262336190388e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008654966577887535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040213484317064285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021331281773746012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.93950434923172, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18990, + "weighted_flow_matching_loss": 0.023127080406993628, + "weighted_rkd_loss": 0.020040806755423544 + }, + { + "epoch": 5.528076811172534, + "flow_matching_loss": 0.024210068210959435, + "grad_norm": 0.21319758892059326, + "learning_rate": 2.904781355477528e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008035478414967656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035403899615630507, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019611346814781428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9394419133663178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19000, + "weighted_flow_matching_loss": 0.024210068210959435, + "weighted_rkd_loss": 0.01842376673594117 + }, + { + "epoch": 5.530986325283678, + "flow_matching_loss": 0.024340588226914407, + "grad_norm": 0.20330394804477692, + "learning_rate": 2.904636370652144e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008461034949868917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00377853917889297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020700609125196934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9393794536590576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19010, + "weighted_flow_matching_loss": 0.024340588226914407, + "weighted_rkd_loss": 0.019445760920643807 + }, + { + "epoch": 5.533895839394821, + "flow_matching_loss": 0.021846460923552514, + "grad_norm": 0.21800251305103302, + "learning_rate": 2.904491279153898e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008787093870341777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003990367148071527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02156455460935831, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9393169641494751, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19020, + "weighted_flow_matching_loss": 0.021846460923552514, + "weighted_rkd_loss": 0.020255972072482108 + }, + { + "epoch": 5.536805353505964, + "flow_matching_loss": 0.02286473959684372, + "grad_norm": 0.22478990256786346, + "learning_rate": 2.9043460809938076e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009068924328312277, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00426084385253489, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022398692555725574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9392544448375701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19030, + "weighted_flow_matching_loss": 0.02286473959684372, + "weighted_rkd_loss": 0.021038095466792583 + }, + { + "epoch": 5.539714867617108, + "flow_matching_loss": 0.02334413807839155, + "grad_norm": 0.27963438630104065, + "learning_rate": 2.9042007761829005e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009142570476979018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040665312437340615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022351672500371934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9391919076442719, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19040, + "weighted_flow_matching_loss": 0.02334413807839155, + "weighted_rkd_loss": 0.020992543548345566 + }, + { + "epoch": 5.542624381728252, + "flow_matching_loss": 0.024203611724078655, + "grad_norm": 0.27469903230667114, + "learning_rate": 2.9040553647322113e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00905281831510365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040445031831040975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022150139696896075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9391293227672577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19050, + "weighted_flow_matching_loss": 0.024203611724078655, + "weighted_rkd_loss": 0.020801861956715583 + }, + { + "epoch": 5.5455338958393945, + "flow_matching_loss": 0.024082008749246597, + "grad_norm": 0.23089492321014404, + "learning_rate": 2.903909846652783e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008658529026433825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004021367849782109, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021338426135480405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9390667080879211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19060, + "weighted_flow_matching_loss": 0.024082008749246597, + "weighted_rkd_loss": 0.020038190484046935 + }, + { + "epoch": 5.548443409950538, + "flow_matching_loss": 0.02340853922069073, + "grad_norm": 0.24683509767055511, + "learning_rate": 2.9037642219556672e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008880602847784757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004116768622770906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021877974271774292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9390040695667267, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19070, + "weighted_flow_matching_loss": 0.02340853922069073, + "weighted_rkd_loss": 0.020543502923101187 + }, + { + "epoch": 5.551352924061682, + "flow_matching_loss": 0.023286159336566924, + "grad_norm": 0.2824164032936096, + "learning_rate": 2.9036184906519222e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008849435858428479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004005562444217503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021704434230923653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9389414012432098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19080, + "weighted_flow_matching_loss": 0.023286159336566924, + "weighted_rkd_loss": 0.02037921342998743 + }, + { + "epoch": 5.554262438172826, + "flow_matching_loss": 0.02108456827700138, + "grad_norm": 0.242328941822052, + "learning_rate": 2.9034726527526166e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008441143203526735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038209918420761824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02070327792316675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9388786971569061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19090, + "weighted_flow_matching_loss": 0.02108456827700138, + "weighted_rkd_loss": 0.019437878020107748 + }, + { + "epoch": 5.557171952283968, + "flow_matching_loss": 0.022546156123280526, + "grad_norm": 0.27619192004203796, + "learning_rate": 2.9033267082688246e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009722382621839642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004223954980261624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023668720200657845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9388159692287446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19100, + "weighted_flow_matching_loss": 0.022546156123280526, + "weighted_rkd_loss": 0.022220572642982005 + }, + { + "epoch": 5.560081466395112, + "flow_matching_loss": 0.02223956473171711, + "grad_norm": 0.3045024871826172, + "learning_rate": 2.9031806572116307e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009766433294862509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004359209653921425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02389207612723112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9387532055377961, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19110, + "weighted_flow_matching_loss": 0.02223956473171711, + "weighted_rkd_loss": 0.022428734973073006 + }, + { + "epoch": 5.562990980506255, + "flow_matching_loss": 0.02516499888151884, + "grad_norm": 0.21753259003162384, + "learning_rate": 2.9030344995921257e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008427428919821978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036804952193051576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020535352639853954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9386904180049896, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19120, + "weighted_flow_matching_loss": 0.02516499888151884, + "weighted_rkd_loss": 0.019276326335966588 + }, + { + "epoch": 5.565900494617399, + "flow_matching_loss": 0.022747710347175598, + "grad_norm": 0.29612666368484497, + "learning_rate": 2.90288823542141e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008873265562579035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00406520594842732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02181173712015152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9386275887489319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19130, + "weighted_flow_matching_loss": 0.022747710347175598, + "weighted_rkd_loss": 0.020473080687224865 + }, + { + "epoch": 5.568810008728542, + "flow_matching_loss": 0.02402761410921812, + "grad_norm": 0.3344840109348297, + "learning_rate": 2.902741864710591e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009099875995889305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004393912618979812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02259366475045681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9385647475719452, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19140, + "weighted_flow_matching_loss": 0.02402761410921812, + "weighted_rkd_loss": 0.02120565576478839 + }, + { + "epoch": 5.571719522839686, + "flow_matching_loss": 0.02254919894039631, + "grad_norm": 0.22222349047660828, + "learning_rate": 2.9025953874707842e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009432960161939263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004125815490260721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022991735953837633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9385018646717072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19150, + "weighted_flow_matching_loss": 0.02254919894039631, + "weighted_rkd_loss": 0.021577833406627177 + }, + { + "epoch": 5.574629036950829, + "flow_matching_loss": 0.021007004380226135, + "grad_norm": 0.25496503710746765, + "learning_rate": 2.9024488037131146e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009288633335381746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004295783257111907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022873050160706045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9384389579296112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19160, + "weighted_flow_matching_loss": 0.021007004380226135, + "weighted_rkd_loss": 0.02146495096385479 + }, + { + "epoch": 5.577538551061973, + "flow_matching_loss": 0.023636828176677227, + "grad_norm": 0.24553227424621582, + "learning_rate": 2.9023021134487135e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009060797048732638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041487478883937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022270341776311396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9383760094642639, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19170, + "weighted_flow_matching_loss": 0.023636828176677227, + "weighted_rkd_loss": 0.02089798413217068 + }, + { + "epoch": 5.580448065173116, + "flow_matching_loss": 0.021957249380648136, + "grad_norm": 0.250021755695343, + "learning_rate": 2.902155316688721e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008763923821970821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037186264991760252, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021246473863720895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9383130371570587, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19180, + "weighted_flow_matching_loss": 0.021957249380648136, + "weighted_rkd_loss": 0.019935854710638523 + }, + { + "epoch": 5.583357579284259, + "flow_matching_loss": 0.023411675356328487, + "grad_norm": 0.23812326788902283, + "learning_rate": 2.902008413444286e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008719163667410611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003981972788460552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02142030019313097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9382500410079956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19190, + "weighted_flow_matching_loss": 0.023411675356328487, + "weighted_rkd_loss": 0.02009760644286871 + }, + { + "epoch": 5.586267093395403, + "flow_matching_loss": 0.024362013675272465, + "grad_norm": 0.30824676156044006, + "learning_rate": 2.9018614037265647e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008107382385060192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00365018870215863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01986495330929756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9381870090961456, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19200, + "weighted_flow_matching_loss": 0.024362013675272465, + "weighted_rkd_loss": 0.018637003563344478 + }, + { + "epoch": 5.5891766075065465, + "flow_matching_loss": 0.023435207642614843, + "grad_norm": 0.2976977229118347, + "learning_rate": 2.9017142875467205e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00910562314093113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003956689941696822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022167936153709888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9381239473819732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19210, + "weighted_flow_matching_loss": 0.023435207642614843, + "weighted_rkd_loss": 0.020796218980103733 + }, + { + "epoch": 5.59208612161769, + "flow_matching_loss": 0.022869262285530566, + "grad_norm": 0.302004337310791, + "learning_rate": 2.9015670649159273e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010398274846374988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005293349036946893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02608989905565977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9380608558654785, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19220, + "weighted_flow_matching_loss": 0.022869262285530566, + "weighted_rkd_loss": 0.024473922699689864 + }, + { + "epoch": 5.594995635728833, + "flow_matching_loss": 0.02555858213454485, + "grad_norm": 0.25178372859954834, + "learning_rate": 2.901419735845365e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009354470064863563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042642598738893865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022973200120031834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9379977405071258, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19230, + "weighted_flow_matching_loss": 0.02555858213454485, + "weighted_rkd_loss": 0.021548830345273016 + }, + { + "epoch": 5.597905149839977, + "flow_matching_loss": 0.023144184798002242, + "grad_norm": 0.23544810712337494, + "learning_rate": 2.9012723003462224e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008828575117513537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038129273103550076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021470077522099017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9379345774650574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19240, + "weighted_flow_matching_loss": 0.023144184798002242, + "weighted_rkd_loss": 0.0201375387609005 + }, + { + "epoch": 5.6008146639511205, + "flow_matching_loss": 0.024055620282888414, + "grad_norm": 0.22664609551429749, + "learning_rate": 2.9011247584296957e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008263609278947115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038255910854786634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020352809689939022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9378713965415955, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19250, + "weighted_flow_matching_loss": 0.024055620282888414, + "weighted_rkd_loss": 0.019088329933583736 + }, + { + "epoch": 5.603724178062263, + "flow_matching_loss": 0.02434695102274418, + "grad_norm": 0.24103154242038727, + "learning_rate": 2.9009771101069908e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009825833048671484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004515915177762508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024167581275105477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9378081977367401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19260, + "weighted_flow_matching_loss": 0.02434695102274418, + "weighted_rkd_loss": 0.02266455367207527 + }, + { + "epoch": 5.606633692173407, + "flow_matching_loss": 0.02437534388154745, + "grad_norm": 0.23710250854492188, + "learning_rate": 2.9008293553893193e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008504406828433276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037205577129498122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020729371160268784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9377449512481689, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19270, + "weighted_flow_matching_loss": 0.02437534388154745, + "weighted_rkd_loss": 0.019438880030065775 + }, + { + "epoch": 5.609543206284551, + "flow_matching_loss": 0.02444952428340912, + "grad_norm": 0.2372005581855774, + "learning_rate": 2.900681494287903e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008497990434989333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038456990849226715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020841679722070693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9376816809177398, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19280, + "weighted_flow_matching_loss": 0.02444952428340912, + "weighted_rkd_loss": 0.019542830623686315 + }, + { + "epoch": 5.6124527203956935, + "flow_matching_loss": 0.021773321367800236, + "grad_norm": 0.25354641675949097, + "learning_rate": 2.9005335268139715e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009645227575674653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045486702118068935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023839125409722328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9376183807849884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19290, + "weighted_flow_matching_loss": 0.021773321367800236, + "weighted_rkd_loss": 0.022351980954408646 + }, + { + "epoch": 5.615362234506837, + "flow_matching_loss": 0.022634117119014262, + "grad_norm": 0.2815823554992676, + "learning_rate": 2.9003854529787615e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009367109462618827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004464373085647821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023198592104017736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9375550508499145, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19300, + "weighted_flow_matching_loss": 0.022634117119014262, + "weighted_rkd_loss": 0.021749952994287013 + }, + { + "epoch": 5.618271748617981, + "flow_matching_loss": 0.024095593579113485, + "grad_norm": 0.24831868708133698, + "learning_rate": 2.9002372727935174e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009505329141393303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004290520260110497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02330117840319872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9374916851520538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19310, + "weighted_flow_matching_loss": 0.024095593579113485, + "weighted_rkd_loss": 0.021844650618731976 + }, + { + "epoch": 5.621181262729124, + "flow_matching_loss": 0.023723260778933764, + "grad_norm": 0.2620788514614105, + "learning_rate": 2.9000889862694934e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009637643489986658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004426401574164629, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023701688461005687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9374282956123352, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19320, + "weighted_flow_matching_loss": 0.023723260778933764, + "weighted_rkd_loss": 0.02221863567829132 + }, + { + "epoch": 5.6240907768402675, + "flow_matching_loss": 0.022122927755117417, + "grad_norm": 0.3032990097999573, + "learning_rate": 2.8999405934179508e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008173947036266328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037650244077667592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020112918503582478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9373648703098297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19330, + "weighted_flow_matching_loss": 0.022122927755117417, + "weighted_rkd_loss": 0.01885313345119357 + }, + { + "epoch": 5.627000290951411, + "flow_matching_loss": 0.024058233201503753, + "grad_norm": 0.22501182556152344, + "learning_rate": 2.899792094250159e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008790749730542302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003875598032027483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021457097306847573, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9373014271259308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19340, + "weighted_flow_matching_loss": 0.024058233201503753, + "weighted_rkd_loss": 0.020111792720854282 + }, + { + "epoch": 5.629909805062555, + "flow_matching_loss": 0.022462382540106775, + "grad_norm": 0.2300107628107071, + "learning_rate": 2.899643488777395e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00842361617833376, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036106702871620657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020457902736961842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.937237948179245, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19350, + "weighted_flow_matching_loss": 0.022462382540106775, + "weighted_rkd_loss": 0.019173926021903754 + }, + { + "epoch": 5.632819319173698, + "flow_matching_loss": 0.022995303571224212, + "grad_norm": 0.22821776568889618, + "learning_rate": 2.899494777010945e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008752691512927412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038515201304107904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02135690292343497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9371744275093079, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19360, + "weighted_flow_matching_loss": 0.022995303571224212, + "weighted_rkd_loss": 0.020015135500580074 + }, + { + "epoch": 5.635728833284841, + "flow_matching_loss": 0.02135955188423395, + "grad_norm": 0.23961178958415985, + "learning_rate": 2.899345958962103e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009172544814646244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004012719937600195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02235780954360962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9371108949184418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19370, + "weighted_flow_matching_loss": 0.02135955188423395, + "weighted_rkd_loss": 0.02095177471637726 + }, + { + "epoch": 5.638638347395985, + "flow_matching_loss": 0.02370105590671301, + "grad_norm": 0.2346774935722351, + "learning_rate": 2.89919703464217e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009366063307970763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004244687804020941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022976813837885857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9370473206043244, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19380, + "weighted_flow_matching_loss": 0.02370105590671301, + "weighted_rkd_loss": 0.021530358865857125 + }, + { + "epoch": 5.641547861507128, + "flow_matching_loss": 0.023014633543789388, + "grad_norm": 0.24976450204849243, + "learning_rate": 2.8990480040624562e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008695319341495634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003912453120574355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021303091756999494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.936983722448349, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19390, + "weighted_flow_matching_loss": 0.023014633543789388, + "weighted_rkd_loss": 0.019960665516555308 + }, + { + "epoch": 5.644457375618272, + "flow_matching_loss": 0.021939928736537696, + "grad_norm": 0.1962006688117981, + "learning_rate": 2.8988988672342797e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00798470275476575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003433913108892739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019403318502008916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9369200885295867, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19400, + "weighted_flow_matching_loss": 0.021939928736537696, + "weighted_rkd_loss": 0.018179365526884796 + }, + { + "epoch": 5.647366889729415, + "flow_matching_loss": 0.023158374615013598, + "grad_norm": 0.23417796194553375, + "learning_rate": 2.898749624168966e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008145521907135845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003720728470943868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02001177240163088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9368564307689666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19410, + "weighted_flow_matching_loss": 0.023158374615013598, + "weighted_rkd_loss": 0.018748167622834443 + }, + { + "epoch": 5.650276403840559, + "flow_matching_loss": 0.022947581298649312, + "grad_norm": 0.23229017853736877, + "learning_rate": 2.898600274877849e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008404175471514463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003778199665248394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02058655060827732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9367927372455597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19420, + "weighted_flow_matching_loss": 0.022947581298649312, + "weighted_rkd_loss": 0.019285348430275916 + }, + { + "epoch": 5.653185917951702, + "flow_matching_loss": 0.02308585401624441, + "grad_norm": 0.24084101617336273, + "learning_rate": 2.898450819372271e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009271668456494808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004342102166265249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022885438986122608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9367290198802948, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19430, + "weighted_flow_matching_loss": 0.02308585401624441, + "weighted_rkd_loss": 0.02143746167421341 + }, + { + "epoch": 5.656095432062846, + "flow_matching_loss": 0.02368790227919817, + "grad_norm": 0.22830067574977875, + "learning_rate": 2.8983012576635825e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008596417028456926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004032242600806058, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021225076727569105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.936665266752243, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19440, + "weighted_flow_matching_loss": 0.02368790227919817, + "weighted_rkd_loss": 0.019880821462720633 + }, + { + "epoch": 5.659004946173989, + "flow_matching_loss": 0.02165221646428108, + "grad_norm": 0.24448202550411224, + "learning_rate": 2.8981515897631414e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009131712699308991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004197666724212468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022461092099547386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9366014897823334, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19450, + "weighted_flow_matching_loss": 0.02165221646428108, + "weighted_rkd_loss": 0.02103710062801838 + }, + { + "epoch": 5.661914460285132, + "flow_matching_loss": 0.021368843130767347, + "grad_norm": 0.2914360761642456, + "learning_rate": 2.8980018156823142e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008505642553791404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037818765034899116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02079316172748804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9365376770496369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19460, + "weighted_flow_matching_loss": 0.021368843130767347, + "weighted_rkd_loss": 0.019473561737686395 + }, + { + "epoch": 5.664823974396276, + "flow_matching_loss": 0.024398160353302955, + "grad_norm": 0.3014562129974365, + "learning_rate": 2.897851935432475e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009112056996673345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004024315741844475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02224842980504036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9364738404750824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19470, + "weighted_flow_matching_loss": 0.024398160353302955, + "weighted_rkd_loss": 0.020835085958242416 + }, + { + "epoch": 5.6677334885074195, + "flow_matching_loss": 0.021727373450994493, + "grad_norm": 0.22474589943885803, + "learning_rate": 2.8977019490250065e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008678060723468661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003940167231485247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021296288724988698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9364099681377411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19480, + "weighted_flow_matching_loss": 0.021727373450994493, + "weighted_rkd_loss": 0.019942105747759342 + }, + { + "epoch": 5.670643002618562, + "flow_matching_loss": 0.023846445605158806, + "grad_norm": 0.2972789704799652, + "learning_rate": 2.8975518564712986e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008627780806273222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003943948517553508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021199510619044303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9363460719585419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19490, + "weighted_flow_matching_loss": 0.023846445605158806, + "weighted_rkd_loss": 0.019850040879100563 + }, + { + "epoch": 5.673552516729706, + "flow_matching_loss": 0.022504043020308017, + "grad_norm": 0.22370152175426483, + "learning_rate": 2.8974016577827504e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007929900381714105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003576914663426578, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019436715357005595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9362821340560913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19500, + "weighted_flow_matching_loss": 0.022504043020308017, + "weighted_rkd_loss": 0.018198251724243164 + }, + { + "epoch": 5.67646203084085, + "flow_matching_loss": 0.02146202977746725, + "grad_norm": 0.2722848355770111, + "learning_rate": 2.8972513529707686e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009171130182221533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004320623818784952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022662884183228017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9362181842327117, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19510, + "weighted_flow_matching_loss": 0.02146202977746725, + "weighted_rkd_loss": 0.021217391453683378 + }, + { + "epoch": 5.679371544951993, + "flow_matching_loss": 0.02131421733647585, + "grad_norm": 0.28260374069213867, + "learning_rate": 2.8971009420467672e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008068476151674987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00356977516785264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019706727750599385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9361541867256165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19520, + "weighted_flow_matching_loss": 0.02131421733647585, + "weighted_rkd_loss": 0.018448537029325963 + }, + { + "epoch": 5.682281059063136, + "flow_matching_loss": 0.02254631109535694, + "grad_norm": 0.3337196707725525, + "learning_rate": 2.8969504250221697e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008541086362674833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038469151128083467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020929088070988654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9360901653766632, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19530, + "weighted_flow_matching_loss": 0.02254631109535694, + "weighted_rkd_loss": 0.0195914919488132 + }, + { + "epoch": 5.68519057317428, + "flow_matching_loss": 0.023682731576263903, + "grad_norm": 0.23791007697582245, + "learning_rate": 2.8967998019084065e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009087658533826471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004141081892885268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022316399216651916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9360261261463165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19540, + "weighted_flow_matching_loss": 0.023682731576263903, + "weighted_rkd_loss": 0.020888709649443626 + }, + { + "epoch": 5.688100087285424, + "flow_matching_loss": 0.023741587810218335, + "grad_norm": 0.2634146511554718, + "learning_rate": 2.8966490727169163e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008336228458210826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00384175768122077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020514214783906935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9359620392322541, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19550, + "weighted_flow_matching_loss": 0.023741587810218335, + "weighted_rkd_loss": 0.019200532138347624 + }, + { + "epoch": 5.6910096013965665, + "flow_matching_loss": 0.02367602400481701, + "grad_norm": 0.23660171031951904, + "learning_rate": 2.8964982374591456e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008959089685231448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004090829845517873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022009009122848512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9358979344367981, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19560, + "weighted_flow_matching_loss": 0.02367602400481701, + "weighted_rkd_loss": 0.02059816587716341 + }, + { + "epoch": 5.69391911550771, + "flow_matching_loss": 0.023208095878362655, + "grad_norm": 0.2695119380950928, + "learning_rate": 2.8963472961465504e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008773644268512727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041709663579240445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021718255057930947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9358337879180908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19570, + "weighted_flow_matching_loss": 0.023208095878362655, + "weighted_rkd_loss": 0.020324683003127574 + }, + { + "epoch": 5.696828629618854, + "flow_matching_loss": 0.023714998550713063, + "grad_norm": 0.2631915211677551, + "learning_rate": 2.896196248790593e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009849981311708689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004399014497175812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024098977260291577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9357696235179901, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19580, + "weighted_flow_matching_loss": 0.023714998550713063, + "weighted_rkd_loss": 0.02255112323909998 + }, + { + "epoch": 5.699738143729997, + "flow_matching_loss": 0.022531876154243947, + "grad_norm": 0.2526552081108093, + "learning_rate": 2.8960450954027447e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0086143902502954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003776312619447708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02100509312003851, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9357054233551025, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19590, + "weighted_flow_matching_loss": 0.022531876154243947, + "weighted_rkd_loss": 0.019654561392962934 + }, + { + "epoch": 5.7026476578411405, + "flow_matching_loss": 0.024173827655613423, + "grad_norm": 0.2856177091598511, + "learning_rate": 2.895893835994484e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009243191545829177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038591170217841864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02234550006687641, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9356411933898926, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19600, + "weighted_flow_matching_loss": 0.024173827655613423, + "weighted_rkd_loss": 0.020907410886138678 + }, + { + "epoch": 5.705557171952284, + "flow_matching_loss": 0.023295665346086025, + "grad_norm": 0.23509445786476135, + "learning_rate": 2.8957424705772992e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008680437644943594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003608421771787107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020969296991825103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9355769395828247, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19610, + "weighted_flow_matching_loss": 0.023295665346086025, + "weighted_rkd_loss": 0.019618392176926135 + }, + { + "epoch": 5.708466686063428, + "flow_matching_loss": 0.022668740712106228, + "grad_norm": 0.23296387493610382, + "learning_rate": 2.8955909991626848e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009173428872600198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003995823767036199, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02234268132597208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9355126440525054, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19620, + "weighted_flow_matching_loss": 0.022668740712106228, + "weighted_rkd_loss": 0.020901881158351898 + }, + { + "epoch": 5.711376200174571, + "flow_matching_loss": 0.02427540421485901, + "grad_norm": 0.23402898013591766, + "learning_rate": 2.8954394217621438e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008539315918460489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037978780223056675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020876510068774223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9354483306407928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19630, + "weighted_flow_matching_loss": 0.02427540421485901, + "weighted_rkd_loss": 0.01952890120446682 + }, + { + "epoch": 5.714285714285714, + "flow_matching_loss": 0.023648287728428842, + "grad_norm": 0.264037162065506, + "learning_rate": 2.8952877383871878e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008761329157277942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036465178010985254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021169176138937474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9353839755058289, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19640, + "weighted_flow_matching_loss": 0.023648287728428842, + "weighted_rkd_loss": 0.019801312871277333 + }, + { + "epoch": 5.717195228396857, + "flow_matching_loss": 0.024966716766357422, + "grad_norm": 0.22688907384872437, + "learning_rate": 2.8951359490493368e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009112111059948802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003992067533545196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02221628949046135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9353196024894714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19650, + "weighted_flow_matching_loss": 0.024966716766357422, + "weighted_rkd_loss": 0.020779332146048546 + }, + { + "epoch": 5.720104742508001, + "flow_matching_loss": 0.025078285299241544, + "grad_norm": 0.3375682830810547, + "learning_rate": 2.894984053760117e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010305675026029349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005125162866897881, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02573651298880577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9352551937103272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19660, + "weighted_flow_matching_loss": 0.025078285299241544, + "weighted_rkd_loss": 0.024070182628929616 + }, + { + "epoch": 5.723014256619145, + "flow_matching_loss": 0.023303444124758245, + "grad_norm": 0.21926642954349518, + "learning_rate": 2.8948320525310644e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008357161981984973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037752960342913866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020489619858562947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.935190761089325, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19670, + "weighted_flow_matching_loss": 0.023303444124758245, + "weighted_rkd_loss": 0.019161735381931066 + }, + { + "epoch": 5.725923770730288, + "flow_matching_loss": 0.024216876178979874, + "grad_norm": 0.26016122102737427, + "learning_rate": 2.8946799453737226e-05, + "loss": 0.0453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00915096844546497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004267774638719857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02256971150636673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9351262867450714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19680, + "weighted_flow_matching_loss": 0.024216876178979874, + "weighted_rkd_loss": 0.021105521637946368 + }, + { + "epoch": 5.728833284841431, + "flow_matching_loss": 0.02361010368913412, + "grad_norm": 0.23910297453403473, + "learning_rate": 2.894527732299643e-05, + "loss": 0.0455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009553732164204121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004263687296770513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02337115164846182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9350617945194244, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19690, + "weighted_flow_matching_loss": 0.02361010368913412, + "weighted_rkd_loss": 0.021853512059897183 + }, + { + "epoch": 5.731742798952575, + "flow_matching_loss": 0.023829376697540282, + "grad_norm": 0.2878577709197998, + "learning_rate": 2.8943754133203856e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008915065927430988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004024473787285388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02185460552573204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9349972605705261, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19700, + "weighted_flow_matching_loss": 0.023829376697540282, + "weighted_rkd_loss": 0.020434025768190624 + }, + { + "epoch": 5.734652313063719, + "flow_matching_loss": 0.021755552478134633, + "grad_norm": 0.2415662705898285, + "learning_rate": 2.894222988447517e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008220402617007494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036012139869853856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020042019337415694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9349326968193055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19710, + "weighted_flow_matching_loss": 0.021755552478134633, + "weighted_rkd_loss": 0.01873794188722968 + }, + { + "epoch": 5.737561827174861, + "flow_matching_loss": 0.021892431005835534, + "grad_norm": 0.24784179031848907, + "learning_rate": 2.894070457692614e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008719998225569726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037948181619867683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02123481472954154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9348681211471558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19720, + "weighted_flow_matching_loss": 0.021892431005835534, + "weighted_rkd_loss": 0.019851754326373337 + }, + { + "epoch": 5.740471341286005, + "flow_matching_loss": 0.024051347374916078, + "grad_norm": 0.3865951597690582, + "learning_rate": 2.8939178210672593e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007732213055714965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003613241342827678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019077667407691477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9348035037517548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19730, + "weighted_flow_matching_loss": 0.024051347374916078, + "weighted_rkd_loss": 0.01783386478200555 + }, + { + "epoch": 5.743380855397149, + "flow_matching_loss": 0.023320935294032098, + "grad_norm": 0.22484718263149261, + "learning_rate": 2.8937650785830455e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008932579448446632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003820480057038367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021685639023780824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9347388446331024, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19740, + "weighted_flow_matching_loss": 0.023320935294032098, + "weighted_rkd_loss": 0.020270435884594918 + }, + { + "epoch": 5.7462903695082925, + "flow_matching_loss": 0.02334144916385412, + "grad_norm": 0.22787196934223175, + "learning_rate": 2.893612230251572e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00884073693305254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004227327019907534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021908801048994064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9346741735935211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19750, + "weighted_flow_matching_loss": 0.02334144916385412, + "weighted_rkd_loss": 0.020477579534053804 + }, + { + "epoch": 5.749199883619435, + "flow_matching_loss": 0.023260318953543902, + "grad_norm": 0.24343785643577576, + "learning_rate": 2.8934592760844466e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009016383718699218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003932134225033223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021964901685714723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9346094727516174, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19760, + "weighted_flow_matching_loss": 0.023260318953543902, + "weighted_rkd_loss": 0.020528621040284634 + }, + { + "epoch": 5.752109397730579, + "flow_matching_loss": 0.022990349493920804, + "grad_norm": 0.24478188157081604, + "learning_rate": 2.8933062160932853e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009359529055655003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004312095255590976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023031152971088886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9345447421073914, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19770, + "weighted_flow_matching_loss": 0.022990349493920804, + "weighted_rkd_loss": 0.021523686684668066 + }, + { + "epoch": 5.755018911841723, + "flow_matching_loss": 0.023549068346619607, + "grad_norm": 0.25137513875961304, + "learning_rate": 2.8931530502897116e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008344222372397781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003805069555528462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020493514090776443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.934479969739914, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19780, + "weighted_flow_matching_loss": 0.023549068346619607, + "weighted_rkd_loss": 0.019150781258940697 + }, + { + "epoch": 5.757928425952866, + "flow_matching_loss": 0.022625031135976314, + "grad_norm": 0.22780855000019073, + "learning_rate": 2.892999778685358e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009167196368798614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004081712337210774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0224161047488451, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9344151794910431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19790, + "weighted_flow_matching_loss": 0.022625031135976314, + "weighted_rkd_loss": 0.020945946499705315 + }, + { + "epoch": 5.760837940064009, + "flow_matching_loss": 0.025462317280471325, + "grad_norm": 0.24157017469406128, + "learning_rate": 2.8928464012918642e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008267248701304197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00363240756560117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020166905038058756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9343503534793853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19800, + "weighted_flow_matching_loss": 0.025462317280471325, + "weighted_rkd_loss": 0.018842975609004498 + }, + { + "epoch": 5.763747454175153, + "flow_matching_loss": 0.02471289597451687, + "grad_norm": 0.2608735263347626, + "learning_rate": 2.8926929181208786e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009389127278700471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004212216404266655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022990470938384532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9342855036258697, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19810, + "weighted_flow_matching_loss": 0.02471289597451687, + "weighted_rkd_loss": 0.021479650773108004 + }, + { + "epoch": 5.766656968286296, + "flow_matching_loss": 0.024650216475129126, + "grad_norm": 0.2362309694290161, + "learning_rate": 2.8925393291840564e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008575297659263015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003895700955763459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02104629632085562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9342206120491028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19820, + "weighted_flow_matching_loss": 0.024650216475129126, + "weighted_rkd_loss": 0.019661908969283104 + }, + { + "epoch": 5.7695664823974395, + "flow_matching_loss": 0.023262246698141097, + "grad_norm": 0.24634085595607758, + "learning_rate": 2.892385634493062e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008718099072575569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036079355515539644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021044133603572844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9341556966304779, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19830, + "weighted_flow_matching_loss": 0.023262246698141097, + "weighted_rkd_loss": 0.019658520072698592 + }, + { + "epoch": 5.772475996508583, + "flow_matching_loss": 0.023944835364818572, + "grad_norm": 0.2428015172481537, + "learning_rate": 2.8922318340595677e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008894545864313841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038311612093821167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021620252914726735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9340907633304596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19840, + "weighted_flow_matching_loss": 0.023944835364818572, + "weighted_rkd_loss": 0.020195303857326506 + }, + { + "epoch": 5.775385510619726, + "flow_matching_loss": 0.022146704606711864, + "grad_norm": 0.21812766790390015, + "learning_rate": 2.8920779278952536e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008511229371652006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003659874643199146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020682333316653968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9340257942676544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19850, + "weighted_flow_matching_loss": 0.022146704606711864, + "weighted_rkd_loss": 0.01931780893355608 + }, + { + "epoch": 5.77829502473087, + "flow_matching_loss": 0.023167558014392853, + "grad_norm": 0.21817563474178314, + "learning_rate": 2.891923916011808e-05, + "loss": 0.0424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00847736168652773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036811148282140495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02063583806157112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9339607715606689, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19860, + "weighted_flow_matching_loss": 0.023167558014392853, + "weighted_rkd_loss": 0.019273106567561628 + }, + { + "epoch": 5.7812045388420135, + "flow_matching_loss": 0.023629009537398816, + "grad_norm": 0.24573945999145508, + "learning_rate": 2.8917697984209263e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008746364573016762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003906320291571319, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021399049367755652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.933895742893219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19870, + "weighted_flow_matching_loss": 0.023629009537398816, + "weighted_rkd_loss": 0.019984446559101345 + }, + { + "epoch": 5.784114052953157, + "flow_matching_loss": 0.023406386189162732, + "grad_norm": 0.25966158509254456, + "learning_rate": 2.891615575134314e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00789052709005773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033529773354530334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019134031422436238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9338306784629822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19880, + "weighted_flow_matching_loss": 0.023406386189162732, + "weighted_rkd_loss": 0.017867934703826905 + }, + { + "epoch": 5.7870235670643, + "flow_matching_loss": 0.02265129890292883, + "grad_norm": 0.23657649755477905, + "learning_rate": 2.891461246163682e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008364553330466151, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036925080232322214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020421614684164525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.933765584230423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19890, + "weighted_flow_matching_loss": 0.02265129890292883, + "weighted_rkd_loss": 0.01906901504844427 + }, + { + "epoch": 5.789933081175444, + "flow_matching_loss": 0.022515105456113814, + "grad_norm": 0.23098599910736084, + "learning_rate": 2.8913068115207514e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008026549685746431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032815562561154367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01933465600013733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9337004661560059, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19900, + "weighted_flow_matching_loss": 0.022515105456113814, + "weighted_rkd_loss": 0.018052779510617255 + }, + { + "epoch": 5.792842595286587, + "flow_matching_loss": 0.023618597909808157, + "grad_norm": 0.2761804461479187, + "learning_rate": 2.8911522712172508e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008672131039202214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003800159739330411, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02114442177116871, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9336353123188019, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19910, + "weighted_flow_matching_loss": 0.023618597909808157, + "weighted_rkd_loss": 0.019741176068782805 + }, + { + "epoch": 5.79575210939773, + "flow_matching_loss": 0.023763767816126345, + "grad_norm": 0.21297915279865265, + "learning_rate": 2.8909976252649156e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007881550444290042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003576750494539738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019339851569384336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9335701286792755, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19920, + "weighted_flow_matching_loss": 0.023763767816126345, + "weighted_rkd_loss": 0.01805513445287943 + }, + { + "epoch": 5.798661623508874, + "flow_matching_loss": 0.021937471255660056, + "grad_norm": 0.2560470998287201, + "learning_rate": 2.8908428736754905e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008991395449265837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038395059294998647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021822297014296055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9335049211978912, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19930, + "weighted_flow_matching_loss": 0.021937471255660056, + "weighted_rkd_loss": 0.02037120219320059 + }, + { + "epoch": 5.801571137620018, + "flow_matching_loss": 0.021825396921485662, + "grad_norm": 0.19634312391281128, + "learning_rate": 2.890688016460728e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007981455093249679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003643833938986063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01960674375295639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9334396719932556, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19940, + "weighted_flow_matching_loss": 0.021825396921485662, + "weighted_rkd_loss": 0.01830172389745712 + }, + { + "epoch": 5.804480651731161, + "flow_matching_loss": 0.022828700393438338, + "grad_norm": 0.2726397216320038, + "learning_rate": 2.890533053632388e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00817842073738575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036342251813039185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01999106667935848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9333744049072266, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19950, + "weighted_flow_matching_loss": 0.022828700393438338, + "weighted_rkd_loss": 0.01865911381319165 + }, + { + "epoch": 5.807390165842304, + "flow_matching_loss": 0.024314516969025136, + "grad_norm": 0.29573121666908264, + "learning_rate": 2.89037798520224e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009690018789842724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004672728921286761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024052766151726246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9333091020584107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19960, + "weighted_flow_matching_loss": 0.024314516969025136, + "weighted_rkd_loss": 0.0224487004801631 + }, + { + "epoch": 5.810299679953448, + "flow_matching_loss": 0.024087257124483584, + "grad_norm": 0.2511559724807739, + "learning_rate": 2.890222811182059e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008600732916966081, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003916047816164792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02111751362681389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9332437753677368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19970, + "weighted_flow_matching_loss": 0.024087257124483584, + "weighted_rkd_loss": 0.019707787595689297 + }, + { + "epoch": 5.813209194064592, + "flow_matching_loss": 0.02353779263794422, + "grad_norm": 0.25204065442085266, + "learning_rate": 2.8900675315836307e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008802350983023644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003743617096915841, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021348319016397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9331784129142762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19980, + "weighted_flow_matching_loss": 0.02353779263794422, + "weighted_rkd_loss": 0.019921784289181232 + }, + { + "epoch": 5.816118708175734, + "flow_matching_loss": 0.02483225502073765, + "grad_norm": 0.2143910527229309, + "learning_rate": 2.889912146418747e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007920106360688805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003660973836667836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019501186721026898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9331130266189576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19990, + "weighted_flow_matching_loss": 0.02483225502073765, + "weighted_rkd_loss": 0.01819680640473962 + }, + { + "epoch": 5.819028222286878, + "flow_matching_loss": 0.024393068440258504, + "grad_norm": 0.22974371910095215, + "learning_rate": 2.889756655699208e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008636699989438057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003855425608344376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021128825284540652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.933047604560852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20000, + "weighted_flow_matching_loss": 0.024393068440258504, + "weighted_rkd_loss": 0.019714199751615525 + }, + { + "epoch": 5.821937736398022, + "flow_matching_loss": 0.02547465022653341, + "grad_norm": 0.28458234667778015, + "learning_rate": 2.889601059436823e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008975445339456201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004018554487265647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02196944523602724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9329821586608886, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20010, + "weighted_flow_matching_loss": 0.02547465022653341, + "weighted_rkd_loss": 0.020497084036469458 + }, + { + "epoch": 5.824847250509165, + "flow_matching_loss": 0.02457576710730791, + "grad_norm": 0.2533929646015167, + "learning_rate": 2.8894453576434077e-05, + "loss": 0.0455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009188351733610035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00404644797090441, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022423151321709155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9329166769981384, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20020, + "weighted_flow_matching_loss": 0.02457576710730791, + "weighted_rkd_loss": 0.020918952114880086 + }, + { + "epoch": 5.827756764620308, + "flow_matching_loss": 0.021839451789855958, + "grad_norm": 0.26455384492874146, + "learning_rate": 2.8892895503307877e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008565803943201899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038792228791862727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0210108308121562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9328511655330658, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20030, + "weighted_flow_matching_loss": 0.021839451789855958, + "weighted_rkd_loss": 0.01959998719394207 + }, + { + "epoch": 5.830666278731452, + "flow_matching_loss": 0.023038704693317414, + "grad_norm": 0.2319878339767456, + "learning_rate": 2.8891336375107942e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009526249207556247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004528442933224142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023580940905958414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9327856302261353, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20040, + "weighted_flow_matching_loss": 0.023038704693317414, + "weighted_rkd_loss": 0.021995983179658653 + }, + { + "epoch": 5.833575792842595, + "flow_matching_loss": 0.023888245597481726, + "grad_norm": 0.24826198816299438, + "learning_rate": 2.888977619195269e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008202725974842905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003576453379355371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01998190525919199, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9327200651168823, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20050, + "weighted_flow_matching_loss": 0.023888245597481726, + "weighted_rkd_loss": 0.018637517094612123 + }, + { + "epoch": 5.836485306953739, + "flow_matching_loss": 0.022164162434637546, + "grad_norm": 0.26782792806625366, + "learning_rate": 2.8888214953960595e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008758666878566145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004081827122718096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021599160693585872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9326544582843781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20060, + "weighted_flow_matching_loss": 0.022164162434637546, + "weighted_rkd_loss": 0.02014455068856478 + }, + { + "epoch": 5.839394821064882, + "flow_matching_loss": 0.02227260321378708, + "grad_norm": 0.2904879152774811, + "learning_rate": 2.8886652661250227e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008534187637269496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037652828730642797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020833658054471016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9325888395309448, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20070, + "weighted_flow_matching_loss": 0.02227260321378708, + "weighted_rkd_loss": 0.019429171364754437 + }, + { + "epoch": 5.842304335176026, + "flow_matching_loss": 0.022863825038075446, + "grad_norm": 0.2958511412143707, + "learning_rate": 2.888508931394024e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00914178160019219, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004415184538811445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022698747366666792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9325231790542603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20080, + "weighted_flow_matching_loss": 0.022863825038075446, + "weighted_rkd_loss": 0.021167090348899364 + }, + { + "epoch": 5.845213849287169, + "flow_matching_loss": 0.02330967430025339, + "grad_norm": 0.22759532928466797, + "learning_rate": 2.888352491214935e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008465672982856632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037635758519172667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020694921538233758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9324574947357178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20090, + "weighted_flow_matching_loss": 0.02330967430025339, + "weighted_rkd_loss": 0.019297153130173682 + }, + { + "epoch": 5.8481233633983125, + "flow_matching_loss": 0.022509181499481203, + "grad_norm": 0.3702440559864044, + "learning_rate": 2.8881959455996365e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008330481266602874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037481361534446476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02040909882634878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9323917746543884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20100, + "weighted_flow_matching_loss": 0.022509181499481203, + "weighted_rkd_loss": 0.01902927625924349 + }, + { + "epoch": 5.851032877509456, + "flow_matching_loss": 0.02357027791440487, + "grad_norm": 0.2367781102657318, + "learning_rate": 2.8880392945600176e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009075334435328842, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004054430918768048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02220509946346283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9323260307312011, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20110, + "weighted_flow_matching_loss": 0.02357027791440487, + "weighted_rkd_loss": 0.020702404249459506 + }, + { + "epoch": 5.853942391620599, + "flow_matching_loss": 0.02480511460453272, + "grad_norm": 0.2639560103416443, + "learning_rate": 2.887882538107974e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008552977768704294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037024089368060233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020808364357799293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.932260251045227, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20120, + "weighted_flow_matching_loss": 0.02480511460453272, + "weighted_rkd_loss": 0.01939886696636677 + }, + { + "epoch": 5.856851905731743, + "flow_matching_loss": 0.02094244621694088, + "grad_norm": 0.24487780034542084, + "learning_rate": 2.8877256762554115e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007811090722680092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003431502962484956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019053684547543526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.932194447517395, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20130, + "weighted_flow_matching_loss": 0.02094244621694088, + "weighted_rkd_loss": 0.017761723883450032 + }, + { + "epoch": 5.8597614198428865, + "flow_matching_loss": 0.022823701612651348, + "grad_norm": 0.3126028776168823, + "learning_rate": 2.8875687090142415e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008471048204228282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003821197571232915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020763294026255607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9321286082267761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20140, + "weighted_flow_matching_loss": 0.022823701612651348, + "weighted_rkd_loss": 0.019354045018553735 + }, + { + "epoch": 5.86267093395403, + "flow_matching_loss": 0.024269686825573445, + "grad_norm": 0.2625453472137451, + "learning_rate": 2.8874116363963856e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009111611638218164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038529746467247604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02207619771361351, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9320627450942993, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20150, + "weighted_flow_matching_loss": 0.024269686825573445, + "weighted_rkd_loss": 0.02057638857513666 + }, + { + "epoch": 5.865580448065173, + "flow_matching_loss": 0.0224126810207963, + "grad_norm": 0.2298211306333542, + "learning_rate": 2.8872544584137717e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008046507136896253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003577689942903817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019670704100281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9319968581199646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20160, + "weighted_flow_matching_loss": 0.0224126810207963, + "weighted_rkd_loss": 0.018333065509796142 + }, + { + "epoch": 5.868489962176317, + "flow_matching_loss": 0.022901743091642855, + "grad_norm": 0.22360414266586304, + "learning_rate": 2.8870971750783373e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007554218824952841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003292660508304834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01840109843760729, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9319309294223785, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20170, + "weighted_flow_matching_loss": 0.022901743091642855, + "weighted_rkd_loss": 0.017148567270487546 + }, + { + "epoch": 5.8713994762874595, + "flow_matching_loss": 0.023254476301372052, + "grad_norm": 0.23798194527626038, + "learning_rate": 2.886939786402026e-05, + "loss": 0.0411, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007840601587668061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003436207864433527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01911741103976965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9318649768829346, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20180, + "weighted_flow_matching_loss": 0.023254476301372052, + "weighted_rkd_loss": 0.017814849503338337 + }, + { + "epoch": 5.874308990398603, + "flow_matching_loss": 0.022842469904571772, + "grad_norm": 0.21855251491069794, + "learning_rate": 2.886782292396791e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007663498120382428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003415483864955604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01874248031526804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9317989945411682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20190, + "weighted_flow_matching_loss": 0.022842469904571772, + "weighted_rkd_loss": 0.01746421717107296 + }, + { + "epoch": 5.877218504509747, + "flow_matching_loss": 0.022156627848744392, + "grad_norm": 0.25550445914268494, + "learning_rate": 2.886624693074593e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009250146942213178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004489645082503557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022989938966929913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9317329823970795, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20200, + "weighted_flow_matching_loss": 0.022156627848744392, + "weighted_rkd_loss": 0.0214204553514719 + }, + { + "epoch": 5.880128018620891, + "flow_matching_loss": 0.023452049307525157, + "grad_norm": 0.26915714144706726, + "learning_rate": 2.8864669884474003e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008078381046652794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003604932245798409, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01976169440895319, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9316669404506683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20210, + "weighted_flow_matching_loss": 0.023452049307525157, + "weighted_rkd_loss": 0.01841131681576371 + }, + { + "epoch": 5.8830375327320334, + "flow_matching_loss": 0.022620260901749133, + "grad_norm": 0.2768271267414093, + "learning_rate": 2.8863091785271903e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009410616243258119, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004028706043027341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022849938552826642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9316008687019348, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20220, + "weighted_flow_matching_loss": 0.022620260901749133, + "weighted_rkd_loss": 0.02128696646541357 + }, + { + "epoch": 5.885947046843177, + "flow_matching_loss": 0.021504119597375394, + "grad_norm": 0.21703475713729858, + "learning_rate": 2.886151263325946e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008203744050115347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003812756482511759, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020220244862139226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9315347671508789, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20230, + "weighted_flow_matching_loss": 0.021504119597375394, + "weighted_rkd_loss": 0.01883586831390858 + }, + { + "epoch": 5.888856560954321, + "flow_matching_loss": 0.02212196830660105, + "grad_norm": 0.2915310859680176, + "learning_rate": 2.885993242855662e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007870675064623356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003659109096042812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01940045915544033, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9314686357975006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20240, + "weighted_flow_matching_loss": 0.02212196830660105, + "weighted_rkd_loss": 0.01807091310620308 + }, + { + "epoch": 5.891766075065464, + "flow_matching_loss": 0.023861220851540566, + "grad_norm": 0.2882578670978546, + "learning_rate": 2.8858351171283375e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009256107732653618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004254772723652422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022766987979412078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9314024746417999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20250, + "weighted_flow_matching_loss": 0.023861220851540566, + "weighted_rkd_loss": 0.021205195412039756 + }, + { + "epoch": 5.894675589176607, + "flow_matching_loss": 0.022039920650422574, + "grad_norm": 0.27171939611434937, + "learning_rate": 2.885676886155982e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008233708003535867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003587969345971942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020055385399609806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9313362896442413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20260, + "weighted_flow_matching_loss": 0.022039920650422574, + "weighted_rkd_loss": 0.018678340874612332 + }, + { + "epoch": 5.897585103287751, + "flow_matching_loss": 0.02368914894759655, + "grad_norm": 0.23585374653339386, + "learning_rate": 2.885518549950611e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008149859262630343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036604347405955195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019960153475403787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9312700688838959, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20270, + "weighted_flow_matching_loss": 0.02368914894759655, + "weighted_rkd_loss": 0.018588317651301624 + }, + { + "epoch": 5.900494617398895, + "flow_matching_loss": 0.02540239319205284, + "grad_norm": 0.284652441740036, + "learning_rate": 2.88536010852425e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008817394962534308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004047166695818305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02168195676058531, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.931203818321228, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20280, + "weighted_flow_matching_loss": 0.02540239319205284, + "weighted_rkd_loss": 0.020190333388745783 + }, + { + "epoch": 5.903404131510038, + "flow_matching_loss": 0.0234245952218771, + "grad_norm": 0.28550636768341064, + "learning_rate": 2.8852015618889317e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009078242443501949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004022055980749428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022178540751338006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9311375379562378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20290, + "weighted_flow_matching_loss": 0.0234245952218771, + "weighted_rkd_loss": 0.020651250332593917 + }, + { + "epoch": 5.906313645621181, + "flow_matching_loss": 0.02100603561848402, + "grad_norm": 0.28104373812675476, + "learning_rate": 2.8850429100566958e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008077972987666725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003893165662884712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020049111917614937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9310712337493896, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20300, + "weighted_flow_matching_loss": 0.02100603561848402, + "weighted_rkd_loss": 0.018667155876755714 + }, + { + "epoch": 5.909223159732325, + "flow_matching_loss": 0.023466983065009116, + "grad_norm": 0.23705561459064484, + "learning_rate": 2.8848841530395917e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009036464989185334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004197999578900635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022270929627120496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9310048937797546, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20310, + "weighted_flow_matching_loss": 0.023466983065009116, + "weighted_rkd_loss": 0.02073439573869109 + }, + { + "epoch": 5.912132673843468, + "flow_matching_loss": 0.02349335178732872, + "grad_norm": 0.3110535144805908, + "learning_rate": 2.8847252908496757e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008047680044546724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00350704335141927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01960240360349417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9309385299682618, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20320, + "weighted_flow_matching_loss": 0.02349335178732872, + "weighted_rkd_loss": 0.018248603399842977 + }, + { + "epoch": 5.915042187954612, + "flow_matching_loss": 0.02359099667519331, + "grad_norm": 0.23905906081199646, + "learning_rate": 2.8845663234990124e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00855595087632537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003611020976677537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020722923055291175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9308721244335174, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20330, + "weighted_flow_matching_loss": 0.02359099667519331, + "weighted_rkd_loss": 0.019290348421782254 + }, + { + "epoch": 5.917951702065755, + "flow_matching_loss": 0.022867479547858238, + "grad_norm": 0.22881363332271576, + "learning_rate": 2.8844072509996736e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008195162564516068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038356322329491377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020225957594811918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9308057069778443, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20340, + "weighted_flow_matching_loss": 0.022867479547858238, + "weighted_rkd_loss": 0.018826441932469608 + }, + { + "epoch": 5.920861216176898, + "flow_matching_loss": 0.023897091671824456, + "grad_norm": 0.2239745706319809, + "learning_rate": 2.8842480733637407e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00871127974241972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003943587047979235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021366146672517062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9307392537593842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20350, + "weighted_flow_matching_loss": 0.023897091671824456, + "weighted_rkd_loss": 0.01988632595166564 + }, + { + "epoch": 5.923770730288042, + "flow_matching_loss": 0.022758979722857474, + "grad_norm": 0.20526711642742157, + "learning_rate": 2.8840887906033015e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00811668294481933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036084999796003104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019841866195201875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9306727588176728, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20360, + "weighted_flow_matching_loss": 0.022758979722857474, + "weighted_rkd_loss": 0.018466319981962444 + }, + { + "epoch": 5.9266802443991855, + "flow_matching_loss": 0.02226179149001837, + "grad_norm": 0.2494022250175476, + "learning_rate": 2.883929402730453e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008740616030991077, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038802817463874815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021361513435840605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9306062459945679, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20370, + "weighted_flow_matching_loss": 0.02226179149001837, + "weighted_rkd_loss": 0.019879139307886363 + }, + { + "epoch": 5.929589758510328, + "flow_matching_loss": 0.022494571283459663, + "grad_norm": 0.23075036704540253, + "learning_rate": 2.8837699097573e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008066752227023244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035429463721811773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019676450826227664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9305396974086761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20380, + "weighted_flow_matching_loss": 0.022494571283459663, + "weighted_rkd_loss": 0.018309699557721614 + }, + { + "epoch": 5.932499272621472, + "flow_matching_loss": 0.021915958262979984, + "grad_norm": 0.23250633478164673, + "learning_rate": 2.883610311695954e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008534823823720217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003784446232020855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020854093972593545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.930473130941391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20390, + "weighted_flow_matching_loss": 0.021915958262979984, + "weighted_rkd_loss": 0.01940420838072896 + }, + { + "epoch": 5.935408786732616, + "flow_matching_loss": 0.02253173980861902, + "grad_norm": 0.2736720144748688, + "learning_rate": 2.883450608558536e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008508448489010334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038602935383096337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020877190679311753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9304065227508544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20400, + "weighted_flow_matching_loss": 0.02253173980861902, + "weighted_rkd_loss": 0.01942424215376377 + }, + { + "epoch": 5.9383183008437594, + "flow_matching_loss": 0.02402003239840269, + "grad_norm": 0.28164082765579224, + "learning_rate": 2.883290800357174e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00857870066538453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036682696547359226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020825670659542085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9303398907184601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20410, + "weighted_flow_matching_loss": 0.02402003239840269, + "weighted_rkd_loss": 0.019374916329979897 + }, + { + "epoch": 5.941227814954902, + "flow_matching_loss": 0.02261544354259968, + "grad_norm": 0.2204427421092987, + "learning_rate": 2.883130887104005e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00792519454844296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003767439117655158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01961782807484269, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9302732229232789, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20420, + "weighted_flow_matching_loss": 0.02261544354259968, + "weighted_rkd_loss": 0.018249931186437605 + }, + { + "epoch": 5.944137329066046, + "flow_matching_loss": 0.022277214005589484, + "grad_norm": 0.2374480664730072, + "learning_rate": 2.8829708688111733e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008544987859204411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00379375999327749, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02088373564183712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9302065372467041, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20430, + "weighted_flow_matching_loss": 0.022277214005589484, + "weighted_rkd_loss": 0.01942616943269968 + }, + { + "epoch": 5.94704684317719, + "flow_matching_loss": 0.023891371488571168, + "grad_norm": 0.23736587166786194, + "learning_rate": 2.8828107454908303e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008030438888818025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037496810546144843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01981055848300457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9301398277282715, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20440, + "weighted_flow_matching_loss": 0.023891371488571168, + "weighted_rkd_loss": 0.018426595069468022 + }, + { + "epoch": 5.9499563572883325, + "flow_matching_loss": 0.02221934273838997, + "grad_norm": 0.2309359610080719, + "learning_rate": 2.8826505171551368e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00814984543249011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003557797148823738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019857487827539443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.930073082447052, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20450, + "weighted_flow_matching_loss": 0.02221934273838997, + "weighted_rkd_loss": 0.018468931689858435 + }, + { + "epoch": 5.952865871399476, + "flow_matching_loss": 0.023114792443811892, + "grad_norm": 0.2944169044494629, + "learning_rate": 2.8824901838162617e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00787675091996789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035314190201461313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019284920860081912, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9300062775611877, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20460, + "weighted_flow_matching_loss": 0.023114792443811892, + "weighted_rkd_loss": 0.01793509181588888 + }, + { + "epoch": 5.95577538551062, + "flow_matching_loss": 0.02275637350976467, + "grad_norm": 0.24544481933116913, + "learning_rate": 2.8823297454863805e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008489494305104017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003630763618275523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020609752181917428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.929939478635788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20470, + "weighted_flow_matching_loss": 0.02275637350976467, + "weighted_rkd_loss": 0.019165826961398125 + }, + { + "epoch": 5.958684899621764, + "flow_matching_loss": 0.021201794594526292, + "grad_norm": 0.26308199763298035, + "learning_rate": 2.882169202177678e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00800589071586728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00360654154792428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019618323259055614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9298726320266724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20480, + "weighted_flow_matching_loss": 0.021201794594526292, + "weighted_rkd_loss": 0.018242505937814714 + }, + { + "epoch": 5.961594413732906, + "flow_matching_loss": 0.022127947956323623, + "grad_norm": 0.2020115852355957, + "learning_rate": 2.8820085539023457e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008569008484482766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036677348427474497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020805751904845238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9298057675361633, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20490, + "weighted_flow_matching_loss": 0.022127947956323623, + "weighted_rkd_loss": 0.019345324020832777 + }, + { + "epoch": 5.96450392784405, + "flow_matching_loss": 0.021748473308980464, + "grad_norm": 0.3015550673007965, + "learning_rate": 2.8818478006725847e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008429344277828932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003771712118759751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020630400627851486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9297388672828675, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20500, + "weighted_flow_matching_loss": 0.021748473308980464, + "weighted_rkd_loss": 0.019180855620652436 + }, + { + "epoch": 5.967413441955194, + "flow_matching_loss": 0.022211700677871704, + "grad_norm": 0.21991387009620667, + "learning_rate": 2.8816869425006025e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007997972751036286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035145859699696304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019510531798005104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9296719372272492, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20510, + "weighted_flow_matching_loss": 0.022211700677871704, + "weighted_rkd_loss": 0.01813841098919511 + }, + { + "epoch": 5.970322956066337, + "flow_matching_loss": 0.0228552658110857, + "grad_norm": 0.20409902930259705, + "learning_rate": 2.881525979398615e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008594659063965083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003749689040705562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020939007587730884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9296049773693085, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20520, + "weighted_flow_matching_loss": 0.0228552658110857, + "weighted_rkd_loss": 0.019465019274502993 + }, + { + "epoch": 5.97323247017748, + "flow_matching_loss": 0.022134913690388204, + "grad_norm": 0.23969416320323944, + "learning_rate": 2.8813649113788467e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075503549538552765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030819976003840567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018182707019150257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9295379996299744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20530, + "weighted_flow_matching_loss": 0.022134913690388204, + "weighted_rkd_loss": 0.016901538148522376 + }, + { + "epoch": 5.976141984288624, + "flow_matching_loss": 0.023677728697657587, + "grad_norm": 0.21834327280521393, + "learning_rate": 2.8812037384535298e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008311892906203867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038294441532343627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020453230105340482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9294709801673889, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20540, + "weighted_flow_matching_loss": 0.023677728697657587, + "weighted_rkd_loss": 0.019010684825479983 + }, + { + "epoch": 5.979051498399767, + "flow_matching_loss": 0.021683630999177694, + "grad_norm": 0.27830496430397034, + "learning_rate": 2.8810424606349044e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008270466001704335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036503888200968503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020191320683807135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9294039309024811, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20550, + "weighted_flow_matching_loss": 0.021683630999177694, + "weighted_rkd_loss": 0.018765915371477603 + }, + { + "epoch": 5.981961012510911, + "flow_matching_loss": 0.02358080092817545, + "grad_norm": 0.2605171501636505, + "learning_rate": 2.8808810779352177e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008125011855736376, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003515878738835454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01976590249687433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9293368637561799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20560, + "weighted_flow_matching_loss": 0.02358080092817545, + "weighted_rkd_loss": 0.018369197193533183 + }, + { + "epoch": 5.984870526622054, + "flow_matching_loss": 0.023550682701170444, + "grad_norm": 0.28736409544944763, + "learning_rate": 2.8807195903667265e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008395864441990853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037788081681355834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020570536702871324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9292697608470917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20570, + "weighted_flow_matching_loss": 0.023550682701170444, + "weighted_rkd_loss": 0.01911559086292982 + }, + { + "epoch": 5.987780040733197, + "flow_matching_loss": 0.021749630197882654, + "grad_norm": 0.2603088617324829, + "learning_rate": 2.8805579979416943e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008698156708851457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036791505059227347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021075463946908714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9292026162147522, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20580, + "weighted_flow_matching_loss": 0.021749630197882654, + "weighted_rkd_loss": 0.01958336038514972 + }, + { + "epoch": 5.990689554844341, + "flow_matching_loss": 0.021576851978898047, + "grad_norm": 0.26981693506240845, + "learning_rate": 2.880396300672393e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008644172735512257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00379335090983659, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021081696823239328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9291354596614838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20590, + "weighted_flow_matching_loss": 0.021576851978898047, + "weighted_rkd_loss": 0.019587717950344086 + }, + { + "epoch": 5.9935990689554846, + "flow_matching_loss": 0.021967476792633535, + "grad_norm": 0.27902334928512573, + "learning_rate": 2.8802344985711025e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008819915680214762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004103468032553792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021743299439549446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9290682673454285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20600, + "weighted_flow_matching_loss": 0.021967476792633535, + "weighted_rkd_loss": 0.020200999919325113 + }, + { + "epoch": 5.996508583066628, + "flow_matching_loss": 0.021893160231411458, + "grad_norm": 0.19476650655269623, + "learning_rate": 2.8800725916501103e-05, + "loss": 0.0396, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007834900263696909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034108981024473907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01908069886267185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9290010452270507, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20610, + "weighted_flow_matching_loss": 0.021893160231411458, + "weighted_rkd_loss": 0.017726000305265187 + }, + { + "epoch": 5.999418097177771, + "flow_matching_loss": 0.0219405603595078, + "grad_norm": 0.33044928312301636, + "learning_rate": 2.8799105799217128e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008974012779071927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004187034093774855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0221350597217679, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9289337933063507, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20620, + "weighted_flow_matching_loss": 0.0219405603595078, + "weighted_rkd_loss": 0.02056200411170721 + }, + { + "epoch": 6.002327611288915, + "flow_matching_loss": 0.022448245249688626, + "grad_norm": 0.24057762324810028, + "learning_rate": 2.879748463398213e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007812688313424587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036316640675067902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01925704088062048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9288665115833282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20630, + "weighted_flow_matching_loss": 0.022448245249688626, + "weighted_rkd_loss": 0.01788719668984413 + }, + { + "epoch": 6.0052371254000585, + "flow_matching_loss": 0.022188085503876208, + "grad_norm": 0.3082675635814667, + "learning_rate": 2.8795862420919233e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008454330265522003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038943953113630414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020803055353462697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9287991881370544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20640, + "weighted_flow_matching_loss": 0.022188085503876208, + "weighted_rkd_loss": 0.019321827776730062 + }, + { + "epoch": 6.008146639511201, + "flow_matching_loss": 0.023479005508124828, + "grad_norm": 0.23302990198135376, + "learning_rate": 2.879423916015163e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008249599114060402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036750010680407287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020174199528992177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9287318885326385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20650, + "weighted_flow_matching_loss": 0.023479005508124828, + "weighted_rkd_loss": 0.01873641200363636 + }, + { + "epoch": 6.011056153622345, + "flow_matching_loss": 0.022392028383910655, + "grad_norm": 0.25950291752815247, + "learning_rate": 2.8792614851802594e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007315155677497387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003077531396411359, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017707843147218228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9286644756793976, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20660, + "weighted_flow_matching_loss": 0.022392028383910655, + "weighted_rkd_loss": 0.016444621421396733 + }, + { + "epoch": 6.013965667733489, + "flow_matching_loss": 0.024282185174524785, + "grad_norm": 0.21597221493721008, + "learning_rate": 2.8790989495995486e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008776075299829244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039026014506816863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02145475186407566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9285970985889435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20670, + "weighted_flow_matching_loss": 0.024282185174524785, + "weighted_rkd_loss": 0.019922806508839132 + }, + { + "epoch": 6.0168751818446315, + "flow_matching_loss": 0.020932783000171186, + "grad_norm": 0.35326242446899414, + "learning_rate": 2.878936309285374e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008162045711651444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003639430669136345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019963521882891654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9285296738147736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20680, + "weighted_flow_matching_loss": 0.020932783000171186, + "weighted_rkd_loss": 0.018536684568971395 + }, + { + "epoch": 6.019784695955775, + "flow_matching_loss": 0.024212090484797954, + "grad_norm": 0.29211893677711487, + "learning_rate": 2.878773564250086e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008461379818618298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003836543811485171, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020759303495287894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9284622073173523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20690, + "weighted_flow_matching_loss": 0.024212090484797954, + "weighted_rkd_loss": 0.019274199567735194 + }, + { + "epoch": 6.022694210066919, + "flow_matching_loss": 0.022100465930998327, + "grad_norm": 0.25989019870758057, + "learning_rate": 2.8786107145060458e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008535301918163896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039961309870705005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021066734939813613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9283947169780731, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20700, + "weighted_flow_matching_loss": 0.022100465930998327, + "weighted_rkd_loss": 0.019558257423341274 + }, + { + "epoch": 6.025603724178063, + "flow_matching_loss": 0.022930718027055264, + "grad_norm": 0.22757099568843842, + "learning_rate": 2.87844776006562e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008836792083457112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039671414997428656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021640725806355475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9283272087574005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20710, + "weighted_flow_matching_loss": 0.022930718027055264, + "weighted_rkd_loss": 0.020089678186923265 + }, + { + "epoch": 6.0285132382892055, + "flow_matching_loss": 0.021617138758301735, + "grad_norm": 0.24736040830612183, + "learning_rate": 2.8782847009411834e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007862882455810905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003419907693751156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019145672488957644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9282596707344055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20720, + "weighted_flow_matching_loss": 0.021617138758301735, + "weighted_rkd_loss": 0.01777215292677283 + }, + { + "epoch": 6.031422752400349, + "flow_matching_loss": 0.023304367996752264, + "grad_norm": 0.2191767394542694, + "learning_rate": 2.8781215371451194e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007954587694257498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034840188920497893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019393194280564786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9281920850276947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20730, + "weighted_flow_matching_loss": 0.023304367996752264, + "weighted_rkd_loss": 0.0180006368085742 + }, + { + "epoch": 6.034332266511493, + "flow_matching_loss": 0.022031837701797487, + "grad_norm": 0.24647235870361328, + "learning_rate": 2.87795826868982e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008512709755450487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038448352366685866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020870254561305045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9281244814395905, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20740, + "weighted_flow_matching_loss": 0.022031837701797487, + "weighted_rkd_loss": 0.019370221812278033 + }, + { + "epoch": 6.037241780622636, + "flow_matching_loss": 0.02294947039335966, + "grad_norm": 0.312610000371933, + "learning_rate": 2.8777948955876832e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008948396425694227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004016113700345159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02191290631890297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9280568540096283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20750, + "weighted_flow_matching_loss": 0.02294947039335966, + "weighted_rkd_loss": 0.02033639941364527 + }, + { + "epoch": 6.040151294733779, + "flow_matching_loss": 0.023342286236584187, + "grad_norm": 0.22311843931674957, + "learning_rate": 2.8776314178511176e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008176262443885206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003874207357876003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020226732548326253, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9279891848564148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20760, + "weighted_flow_matching_loss": 0.023342286236584187, + "weighted_rkd_loss": 0.01877020550891757 + }, + { + "epoch": 6.043060808844923, + "flow_matching_loss": 0.022250414453446866, + "grad_norm": 0.2637951076030731, + "learning_rate": 2.877467835492537e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008516761008650064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003738356428220868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02077187839895487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9279214978218079, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20770, + "weighted_flow_matching_loss": 0.022250414453446866, + "weighted_rkd_loss": 0.01927466094493866 + }, + { + "epoch": 6.045970322956066, + "flow_matching_loss": 0.02284017214551568, + "grad_norm": 0.2516140341758728, + "learning_rate": 2.8773041485243643e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008924437640234827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004172376380302012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02202125173062086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.927853775024414, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20780, + "weighted_flow_matching_loss": 0.02284017214551568, + "weighted_rkd_loss": 0.02043246254324913 + }, + { + "epoch": 6.04887983706721, + "flow_matching_loss": 0.021402724087238312, + "grad_norm": 0.19392551481723785, + "learning_rate": 2.8771403569590316e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007986391661688685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003277162089943886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01924994559958577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9277860224246979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20790, + "weighted_flow_matching_loss": 0.021402724087238312, + "weighted_rkd_loss": 0.017859831172972917 + }, + { + "epoch": 6.051789351178353, + "flow_matching_loss": 0.02274241764098406, + "grad_norm": 0.2554221749305725, + "learning_rate": 2.8769764608089766e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008232209272682667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003765389719046652, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020229808054864407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9277182459831238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20800, + "weighted_flow_matching_loss": 0.02274241764098406, + "weighted_rkd_loss": 0.01876754267141223 + }, + { + "epoch": 6.054698865289497, + "flow_matching_loss": 0.0222181236371398, + "grad_norm": 0.24677112698554993, + "learning_rate": 2.876812460086647e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007773278187960387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003678860259242356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019225416891276836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9276504337787628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20810, + "weighted_flow_matching_loss": 0.0222181236371398, + "weighted_rkd_loss": 0.017834479361772536 + }, + { + "epoch": 6.05760837940064, + "flow_matching_loss": 0.0229552099481225, + "grad_norm": 0.2906435430049896, + "learning_rate": 2.876648354804497e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008147903764620424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037613036576658486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02005711104720831, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9275825977325439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20820, + "weighted_flow_matching_loss": 0.0229552099481225, + "weighted_rkd_loss": 0.01860462613403797 + }, + { + "epoch": 6.060517893511784, + "flow_matching_loss": 0.022554895654320718, + "grad_norm": 0.2573074400424957, + "learning_rate": 2.87648414497499e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009457278344780207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041022698627784845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02301682662218809, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9275147259235382, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20830, + "weighted_flow_matching_loss": 0.022554895654320718, + "weighted_rkd_loss": 0.021348474733531474 + }, + { + "epoch": 6.063427407622927, + "flow_matching_loss": 0.022718033753335476, + "grad_norm": 0.2281186878681183, + "learning_rate": 2.8763198306105958e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008125637285411358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003756374213844538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020007648691534995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9274468421936035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20840, + "weighted_flow_matching_loss": 0.022718033753335476, + "weighted_rkd_loss": 0.018556003645062446 + }, + { + "epoch": 6.06633692173407, + "flow_matching_loss": 0.023312218487262726, + "grad_norm": 0.26275309920310974, + "learning_rate": 2.8761554117237932e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008439050102606416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037976976484060286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020675797946751118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9273788928985596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20850, + "weighted_flow_matching_loss": 0.023312218487262726, + "weighted_rkd_loss": 0.019174271170049904 + }, + { + "epoch": 6.069246435845214, + "flow_matching_loss": 0.02380710020661354, + "grad_norm": 0.2517724335193634, + "learning_rate": 2.875990888327069e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008767415909096599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040477513102814555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021582582779228688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9273109436035156, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20860, + "weighted_flow_matching_loss": 0.02380710020661354, + "weighted_rkd_loss": 0.020013726595789195 + }, + { + "epoch": 6.0721559499563575, + "flow_matching_loss": 0.022667812183499336, + "grad_norm": 0.2803571820259094, + "learning_rate": 2.875826260432918e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00828433227725327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036758589325472713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020244523603469133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9272429645061493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20870, + "weighted_flow_matching_loss": 0.022667812183499336, + "weighted_rkd_loss": 0.018771577160805464 + }, + { + "epoch": 6.0750654640675, + "flow_matching_loss": 0.020833294466137887, + "grad_norm": 0.24555255472660065, + "learning_rate": 2.875661528053842e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008650086540728807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038164077093824746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021116581000387668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9271749556064606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20880, + "weighted_flow_matching_loss": 0.020833294466137887, + "weighted_rkd_loss": 0.019578769523650407 + }, + { + "epoch": 6.077974978178644, + "flow_matching_loss": 0.024118846654891966, + "grad_norm": 0.2657126486301422, + "learning_rate": 2.8754966912023514e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008258650870993733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036516839172691107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020168985426425933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9271069049835206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20890, + "weighted_flow_matching_loss": 0.024118846654891966, + "weighted_rkd_loss": 0.01869878051802516 + }, + { + "epoch": 6.080884492289788, + "flow_matching_loss": 0.022793348506093027, + "grad_norm": 0.2683853209018707, + "learning_rate": 2.8753317498909648e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008429748006165028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003756892564706504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020616388786584137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9270388364791871, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20900, + "weighted_flow_matching_loss": 0.022793348506093027, + "weighted_rkd_loss": 0.019112162571400405 + }, + { + "epoch": 6.083794006400931, + "flow_matching_loss": 0.023786956816911696, + "grad_norm": 0.263160765171051, + "learning_rate": 2.875166704132208e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008405359741300345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003918816917575896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020729536842554807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9269707322120666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20910, + "weighted_flow_matching_loss": 0.023786956816911696, + "weighted_rkd_loss": 0.019215683080255987 + }, + { + "epoch": 6.086703520512074, + "flow_matching_loss": 0.024007930234074593, + "grad_norm": 0.25048840045928955, + "learning_rate": 2.8750015539386154e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008573440555483102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037408435251563787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020887724682688714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9269026041030883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20920, + "weighted_flow_matching_loss": 0.024007930234074593, + "weighted_rkd_loss": 0.019360896293073894 + }, + { + "epoch": 6.089613034623218, + "flow_matching_loss": 0.020894386805593967, + "grad_norm": 0.2445230334997177, + "learning_rate": 2.874836299322729e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007834503706544638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003554253536276519, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019223261065781115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9268344402313232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20930, + "weighted_flow_matching_loss": 0.020894386805593967, + "weighted_rkd_loss": 0.017816766723990442 + }, + { + "epoch": 6.092522548734362, + "flow_matching_loss": 0.024335913360118866, + "grad_norm": 0.23981943726539612, + "learning_rate": 2.8746709402970986e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009096289053559303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004080537287518382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02227311544120312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9267662525177002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20940, + "weighted_flow_matching_loss": 0.024335913360118866, + "weighted_rkd_loss": 0.020641980040818454 + }, + { + "epoch": 6.0954320628455045, + "flow_matching_loss": 0.02393374890089035, + "grad_norm": 0.24339865148067474, + "learning_rate": 2.8745054768742828e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008543208474293351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037834363291040065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020869853161275386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9266980290412903, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20950, + "weighted_flow_matching_loss": 0.02393374890089035, + "weighted_rkd_loss": 0.019340054132044315 + }, + { + "epoch": 6.098341576956648, + "flow_matching_loss": 0.02235928475856781, + "grad_norm": 0.20458608865737915, + "learning_rate": 2.8743399090668464e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008497280534356833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038954539224505423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020890015177428722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9266297817230225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20960, + "weighted_flow_matching_loss": 0.02235928475856781, + "weighted_rkd_loss": 0.019357331190258266 + }, + { + "epoch": 6.101251091067792, + "flow_matching_loss": 0.021348234638571738, + "grad_norm": 0.24276398122310638, + "learning_rate": 2.874174236887364e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008233985072001815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035765924490988254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020044562593102454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9265615046024323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20970, + "weighted_flow_matching_loss": 0.021348234638571738, + "weighted_rkd_loss": 0.018572529219090937 + }, + { + "epoch": 6.104160605178935, + "flow_matching_loss": 0.022726028971374036, + "grad_norm": 0.26333415508270264, + "learning_rate": 2.874008460348417e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008903412334620952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004046254209242761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02185307890176773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9264931976795197, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20980, + "weighted_flow_matching_loss": 0.022726028971374036, + "weighted_rkd_loss": 0.02024671994149685 + }, + { + "epoch": 6.1070701192900785, + "flow_matching_loss": 0.021999376825988293, + "grad_norm": 0.28626301884651184, + "learning_rate": 2.8738425794625952e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00828878334723413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036997525952756406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02027731891721487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9264248609542847, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20990, + "weighted_flow_matching_loss": 0.021999376825988293, + "weighted_rkd_loss": 0.018785409163683653 + }, + { + "epoch": 6.109979633401222, + "flow_matching_loss": 0.02466666083782911, + "grad_norm": 0.21452905237674713, + "learning_rate": 2.873676594242496e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008192129246890546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038354188902303575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02021967712789774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9263565003871918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21000, + "weighted_flow_matching_loss": 0.02466666083782911, + "weighted_rkd_loss": 0.01873066769912839 + }, + { + "epoch": 6.112889147512365, + "flow_matching_loss": 0.023501423187553883, + "grad_norm": 0.2189486026763916, + "learning_rate": 2.8735105047007255e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008508921042084694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036595589248463512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020677401218563318, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.926288104057312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21010, + "weighted_flow_matching_loss": 0.023501423187553883, + "weighted_rkd_loss": 0.019153217785060406 + }, + { + "epoch": 6.115798661623509, + "flow_matching_loss": 0.020527230016887187, + "grad_norm": 0.20221565663814545, + "learning_rate": 2.8733443108498957e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008678412716835737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003921324433758855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021278150007128717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9262196779251098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21020, + "weighted_flow_matching_loss": 0.020527230016887187, + "weighted_rkd_loss": 0.019708246644586324 + }, + { + "epoch": 6.118708175734652, + "flow_matching_loss": 0.02292131893336773, + "grad_norm": 0.26962539553642273, + "learning_rate": 2.8731780127026292e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008411674993112683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003790482971817255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020613832958042623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9261512219905853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21030, + "weighted_flow_matching_loss": 0.02292131893336773, + "weighted_rkd_loss": 0.01909153675660491 + }, + { + "epoch": 6.121617689845796, + "flow_matching_loss": 0.02166998777538538, + "grad_norm": 0.24051153659820557, + "learning_rate": 2.873011610271555e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009120946330949665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004432362574152649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022674255445599555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9260827481746674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21040, + "weighted_flow_matching_loss": 0.02166998777538538, + "weighted_rkd_loss": 0.020998256653547286 + }, + { + "epoch": 6.124527203956939, + "flow_matching_loss": 0.024362461268901826, + "grad_norm": 0.23140420019626617, + "learning_rate": 2.8728451035693095e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007968534762039781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036184508120641113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019555520359426738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9260142147541046, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21050, + "weighted_flow_matching_loss": 0.024362461268901826, + "weighted_rkd_loss": 0.018108714371919632 + }, + { + "epoch": 6.127436718068083, + "flow_matching_loss": 0.023223199136555194, + "grad_norm": 0.22502799332141876, + "learning_rate": 2.872678492608539e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007429976016283035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003372016968205571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018231968954205514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9259456694126129, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21060, + "weighted_flow_matching_loss": 0.023223199136555194, + "weighted_rkd_loss": 0.016881799418479206 + }, + { + "epoch": 6.130346232179226, + "flow_matching_loss": 0.02250107768923044, + "grad_norm": 0.24363857507705688, + "learning_rate": 2.872511777401895e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008097614534199238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034431441919878123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019638373423367737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9258771240711212, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21070, + "weighted_flow_matching_loss": 0.02250107768923044, + "weighted_rkd_loss": 0.018182739801704883 + }, + { + "epoch": 6.133255746290369, + "flow_matching_loss": 0.021826452016830443, + "grad_norm": 0.21505071222782135, + "learning_rate": 2.87234495796204e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009022253984585404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004113123309798539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02215763172134757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9258085250854492, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21080, + "weighted_flow_matching_loss": 0.021826452016830443, + "weighted_rkd_loss": 0.020513738226145505 + }, + { + "epoch": 6.136165260401513, + "flow_matching_loss": 0.02147336434572935, + "grad_norm": 0.2283715158700943, + "learning_rate": 2.8721780343016423e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007685398031026125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032369575696066023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018607753701508044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9257398903369903, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21090, + "weighted_flow_matching_loss": 0.02147336434572935, + "weighted_rkd_loss": 0.01722595626488328 + }, + { + "epoch": 6.139074774512657, + "flow_matching_loss": 0.023887988366186618, + "grad_norm": 0.20518459379673004, + "learning_rate": 2.8720110064333782e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009005829272791744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003991818078793586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02200347688049078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9256712436676026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21100, + "weighted_flow_matching_loss": 0.023887988366186618, + "weighted_rkd_loss": 0.020367980375885962 + }, + { + "epoch": 6.141984288623799, + "flow_matching_loss": 0.022804833203554153, + "grad_norm": 0.2671080529689789, + "learning_rate": 2.8718438743699325e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008135205507278443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036889128619804978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019959323666989805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9256025552749634, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21110, + "weighted_flow_matching_loss": 0.022804833203554153, + "weighted_rkd_loss": 0.018474385235458614 + }, + { + "epoch": 6.144893802734943, + "flow_matching_loss": 0.021937419660389423, + "grad_norm": 0.23142637312412262, + "learning_rate": 2.871676638123998e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008133811643347144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003489447315223515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01975707057863474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9255338311195374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21120, + "weighted_flow_matching_loss": 0.021937419660389423, + "weighted_rkd_loss": 0.01828580955043435 + }, + { + "epoch": 6.147803316846087, + "flow_matching_loss": 0.022920732386410236, + "grad_norm": 0.2142478972673416, + "learning_rate": 2.8715092977082754e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007777588022872805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034408242208883166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018996000476181506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9254650890827179, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21130, + "weighted_flow_matching_loss": 0.022920732386410236, + "weighted_rkd_loss": 0.01758011793717742 + }, + { + "epoch": 6.1507128309572305, + "flow_matching_loss": 0.021361346170306206, + "grad_norm": 0.23444102704524994, + "learning_rate": 2.8713418531354726e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007687756512314081, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003413586877286434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018789100088179113, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9253963172435761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21140, + "weighted_flow_matching_loss": 0.021361346170306206, + "weighted_rkd_loss": 0.01738737616688013 + }, + { + "epoch": 6.153622345068373, + "flow_matching_loss": 0.021817334368824957, + "grad_norm": 0.23182126879692078, + "learning_rate": 2.8711743044183057e-05, + "loss": 0.0396, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007903239410370588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034125103615224362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01921898890286684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9253275096416473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21150, + "weighted_flow_matching_loss": 0.021817334368824957, + "weighted_rkd_loss": 0.01778385266661644 + }, + { + "epoch": 6.156531859179517, + "flow_matching_loss": 0.024434698186814784, + "grad_norm": 0.2586787939071655, + "learning_rate": 2.8710066515695002e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072178429458290335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031280873110517858, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017563773039728402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9252586901187897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21160, + "weighted_flow_matching_loss": 0.024434698186814784, + "weighted_rkd_loss": 0.016251017712056635 + }, + { + "epoch": 6.159441373290661, + "flow_matching_loss": 0.02060577683150768, + "grad_norm": 0.19283217191696167, + "learning_rate": 2.870838894601787e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007876385375857354, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035822554025799038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019335025828331708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9251898169517517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21170, + "weighted_flow_matching_loss": 0.02060577683150768, + "weighted_rkd_loss": 0.01788855269551277 + }, + { + "epoch": 6.162350887401804, + "flow_matching_loss": 0.023623072169721128, + "grad_norm": 0.22131039202213287, + "learning_rate": 2.870671033527906e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008817564509809018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004278575209900737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02191370464861393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9251209259033203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21180, + "weighted_flow_matching_loss": 0.023623072169721128, + "weighted_rkd_loss": 0.020272849593311548 + }, + { + "epoch": 6.165260401512947, + "flow_matching_loss": 0.02248728387057781, + "grad_norm": 0.2615566551685333, + "learning_rate": 2.8705030683606067e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007724640797823668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034041175385937095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01885339953005314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9250520169734955, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21190, + "weighted_flow_matching_loss": 0.02248728387057781, + "weighted_rkd_loss": 0.017440362367779016 + }, + { + "epoch": 6.168169915624091, + "flow_matching_loss": 0.022723483853042126, + "grad_norm": 0.24373498558998108, + "learning_rate": 2.8703349991126432e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008148501953110098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036720269359648227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019969030842185022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9249830603599548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21200, + "weighted_flow_matching_loss": 0.022723483853042126, + "weighted_rkd_loss": 0.018471036665141584 + }, + { + "epoch": 6.171079429735234, + "flow_matching_loss": 0.023203765600919725, + "grad_norm": 0.27089637517929077, + "learning_rate": 2.87016682579678e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007933364855125547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035552470944821835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01942197708413005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9249140739440918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21210, + "weighted_flow_matching_loss": 0.023203765600919725, + "weighted_rkd_loss": 0.01796366088092327 + }, + { + "epoch": 6.1739889438463775, + "flow_matching_loss": 0.0223997313529253, + "grad_norm": 0.2503691613674164, + "learning_rate": 2.8699985484257883e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007324215397238732, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033716247882694004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018020055443048476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9248450756072998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21220, + "weighted_flow_matching_loss": 0.0223997313529253, + "weighted_rkd_loss": 0.016665758937597273 + }, + { + "epoch": 6.176898457957521, + "flow_matching_loss": 0.02220744490623474, + "grad_norm": 0.19805265963077545, + "learning_rate": 2.8698301670124484e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008330573095008731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035228947643190624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02018404081463814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.924776041507721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21230, + "weighted_flow_matching_loss": 0.02220744490623474, + "weighted_rkd_loss": 0.01866573914885521 + }, + { + "epoch": 6.179807972068665, + "flow_matching_loss": 0.023834651336073875, + "grad_norm": 0.2542620599269867, + "learning_rate": 2.869661681569548e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007827875344082713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00352153442800045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01917728502303362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9247069776058197, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21240, + "weighted_flow_matching_loss": 0.023834651336073875, + "weighted_rkd_loss": 0.01773337125778198 + }, + { + "epoch": 6.182717486179808, + "flow_matching_loss": 0.023452847823500635, + "grad_norm": 0.23425085842609406, + "learning_rate": 2.8694930921098813e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008095459826290608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003571364167146385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01976228430867195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9246378779411316, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21250, + "weighted_flow_matching_loss": 0.023452847823500635, + "weighted_rkd_loss": 0.018272949010133745 + }, + { + "epoch": 6.1856270002909515, + "flow_matching_loss": 0.02260776571929455, + "grad_norm": 0.31109535694122314, + "learning_rate": 2.8693243986462518e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008743934333324432, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040891102980822325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02157697919756174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.924568772315979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21260, + "weighted_flow_matching_loss": 0.02260776571929455, + "weighted_rkd_loss": 0.019949352368712425 + }, + { + "epoch": 6.188536514402095, + "flow_matching_loss": 0.022724047675728797, + "grad_norm": 0.2387009710073471, + "learning_rate": 2.8691556011914716e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007762839505448938, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003556164400652051, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019081843551248312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9244996070861816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21270, + "weighted_flow_matching_loss": 0.022724047675728797, + "weighted_rkd_loss": 0.01764115449041128 + }, + { + "epoch": 6.191446028513238, + "flow_matching_loss": 0.020654009468853474, + "grad_norm": 0.18326258659362793, + "learning_rate": 2.8689866997583585e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007522241678088903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003307825978845358, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01835230942815542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9244304239749909, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21280, + "weighted_flow_matching_loss": 0.020654009468853474, + "weighted_rkd_loss": 0.01696547046303749 + }, + { + "epoch": 6.194355542624382, + "flow_matching_loss": 0.02265725415199995, + "grad_norm": 0.23766042292118073, + "learning_rate": 2.8688176943597405e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00775248259305954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032870226306840776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018791988026350737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9243612110614776, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21290, + "weighted_flow_matching_loss": 0.02265725415199995, + "weighted_rkd_loss": 0.0173705842345953 + }, + { + "epoch": 6.197265056735525, + "flow_matching_loss": 0.023246585950255395, + "grad_norm": 0.22493509948253632, + "learning_rate": 2.8686485850084522e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008215088816359638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003861059690825641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02029123716056347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9242919743061065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21300, + "weighted_flow_matching_loss": 0.023246585950255395, + "weighted_rkd_loss": 0.018755022436380386 + }, + { + "epoch": 6.200174570846668, + "flow_matching_loss": 0.02443315964192152, + "grad_norm": 0.21016603708267212, + "learning_rate": 2.8684793717173358e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007844946440309286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00353197380900383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019221866503357887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9242227077484131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21310, + "weighted_flow_matching_loss": 0.02443315964192152, + "weighted_rkd_loss": 0.017765319813042878 + }, + { + "epoch": 6.203084084957812, + "flow_matching_loss": 0.022081106156110763, + "grad_norm": 0.22124934196472168, + "learning_rate": 2.8683100544992423e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008669599378481507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003995737107470631, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021334935910999775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9241534054279328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21320, + "weighted_flow_matching_loss": 0.022081106156110763, + "weighted_rkd_loss": 0.019716789200901984 + }, + { + "epoch": 6.205993599068956, + "flow_matching_loss": 0.02040244247764349, + "grad_norm": 0.25502997636795044, + "learning_rate": 2.8681406333670302e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007886220375075937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033672478515654803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019139688834547995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9240840792655944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21330, + "weighted_flow_matching_loss": 0.02040244247764349, + "weighted_rkd_loss": 0.01768665947020054 + }, + { + "epoch": 6.208903113180099, + "flow_matching_loss": 0.023006968572735785, + "grad_norm": 0.22132937610149384, + "learning_rate": 2.8679711083335664e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008474826766178011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037356138695031405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02068526726216078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9240147233009338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21340, + "weighted_flow_matching_loss": 0.023006968572735785, + "weighted_rkd_loss": 0.019113509077578784 + }, + { + "epoch": 6.211812627291242, + "flow_matching_loss": 0.021059119515120982, + "grad_norm": 0.2748109698295593, + "learning_rate": 2.8678014794117244e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00783038097433746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036776467924937604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01933840857818723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9239453315734864, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21350, + "weighted_flow_matching_loss": 0.021059119515120982, + "weighted_rkd_loss": 0.01786763742566109 + }, + { + "epoch": 6.214722141402386, + "flow_matching_loss": 0.02258762326091528, + "grad_norm": 0.22393806278705597, + "learning_rate": 2.8676317466143866e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007944197580218314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036404035286977886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019528798945248126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9238759279251099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21360, + "weighted_flow_matching_loss": 0.02258762326091528, + "weighted_rkd_loss": 0.01804218227043748 + }, + { + "epoch": 6.21763165551353, + "flow_matching_loss": 0.02146568298339844, + "grad_norm": 0.2247946411371231, + "learning_rate": 2.8674619099544437e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008251423388719559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003467587102204561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019970433972775935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9238064765930176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21370, + "weighted_flow_matching_loss": 0.02146568298339844, + "weighted_rkd_loss": 0.01844878736883402 + }, + { + "epoch": 6.220541169624672, + "flow_matching_loss": 0.02211761921644211, + "grad_norm": 0.23733852803707123, + "learning_rate": 2.8672919694447927e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009009760012850166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004339889227412641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02235940909013152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9237370073795319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21380, + "weighted_flow_matching_loss": 0.02211761921644211, + "weighted_rkd_loss": 0.020654298644512892 + }, + { + "epoch": 6.223450683735816, + "flow_matching_loss": 0.023544960469007493, + "grad_norm": 0.23886975646018982, + "learning_rate": 2.8671219250983404e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007996040303260088, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003563668858259916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019555749744176863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9236675024032592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21390, + "weighted_flow_matching_loss": 0.023544960469007493, + "weighted_rkd_loss": 0.018063006084412338 + }, + { + "epoch": 6.22636019784696, + "flow_matching_loss": 0.020648689195513725, + "grad_norm": 0.22044284641742706, + "learning_rate": 2.866951776928e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0082853680010885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035961697110906245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020166905783116817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9235979735851287, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21400, + "weighted_flow_matching_loss": 0.020648689195513725, + "weighted_rkd_loss": 0.01862609488889575 + }, + { + "epoch": 6.229269711958103, + "flow_matching_loss": 0.02346955016255379, + "grad_norm": 0.2580922842025757, + "learning_rate": 2.866781524946693e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008741052541881799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039670554921031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0214491605758667, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9235284090042114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21410, + "weighted_flow_matching_loss": 0.02346955016255379, + "weighted_rkd_loss": 0.01980895008891821 + }, + { + "epoch": 6.232179226069246, + "flow_matching_loss": 0.023363183066248894, + "grad_norm": 0.23345983028411865, + "learning_rate": 2.866611169167349e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008319711871445179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037121264962479473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020351550169289113, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9234588265419006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21420, + "weighted_flow_matching_loss": 0.023363183066248894, + "weighted_rkd_loss": 0.018793844059109686 + }, + { + "epoch": 6.23508874018039, + "flow_matching_loss": 0.02260481668636203, + "grad_norm": 0.2622460722923279, + "learning_rate": 2.866440709602906e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008445120695978404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003730692691169679, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02062093438580632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9233892023563385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21430, + "weighted_flow_matching_loss": 0.02260481668636203, + "weighted_rkd_loss": 0.019041137583553792 + }, + { + "epoch": 6.237998254291533, + "flow_matching_loss": 0.020667687989771366, + "grad_norm": 0.25768500566482544, + "learning_rate": 2.8662701462663082e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007777615264058113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033616469707340004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01891687735915184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9233195662498475, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21440, + "weighted_flow_matching_loss": 0.020667687989771366, + "weighted_rkd_loss": 0.017466321121901274 + }, + { + "epoch": 6.240907768402677, + "flow_matching_loss": 0.023290074802935122, + "grad_norm": 0.2561684250831604, + "learning_rate": 2.86609947917051e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008132260292768478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003988891676999629, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02025341223925352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9232498824596405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21450, + "weighted_flow_matching_loss": 0.023290074802935122, + "weighted_rkd_loss": 0.01869895588606596 + }, + { + "epoch": 6.24381728251382, + "flow_matching_loss": 0.019994533248245717, + "grad_norm": 0.23779185116291046, + "learning_rate": 2.865928708328472e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007522847084328532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036479346686974167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01869362872093916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9231801927089691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21460, + "weighted_flow_matching_loss": 0.019994533248245717, + "weighted_rkd_loss": 0.017257562931627037 + }, + { + "epoch": 6.246726796624964, + "flow_matching_loss": 0.02145880088210106, + "grad_norm": 0.18779344856739044, + "learning_rate": 2.8657578337531626e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007824625167995692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035437775310128926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01919302809983492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9231104552745819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21470, + "weighted_flow_matching_loss": 0.02145880088210106, + "weighted_rkd_loss": 0.01771728415042162 + }, + { + "epoch": 6.249636310736107, + "flow_matching_loss": 0.02377616409212351, + "grad_norm": 0.25203022360801697, + "learning_rate": 2.8655868554575588e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008418319653719664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038252994418144226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020661938562989236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9230406820774079, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21480, + "weighted_flow_matching_loss": 0.02377616409212351, + "weighted_rkd_loss": 0.019071777630597354 + }, + { + "epoch": 6.2525458248472505, + "flow_matching_loss": 0.02459085565060377, + "grad_norm": 0.256538063287735, + "learning_rate": 2.8654157734546453e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007849572552368045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003783260309137404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01948240539059043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9229708909988403, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21490, + "weighted_flow_matching_loss": 0.02459085565060377, + "weighted_rkd_loss": 0.01798162879422307 + }, + { + "epoch": 6.255455338958394, + "flow_matching_loss": 0.02292176689952612, + "grad_norm": 0.22392721474170685, + "learning_rate": 2.865244587757415e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008716224320232868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003967251908034086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021399700455367564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9229010760784149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21500, + "weighted_flow_matching_loss": 0.02292176689952612, + "weighted_rkd_loss": 0.019749813713133335 + }, + { + "epoch": 6.258364853069537, + "flow_matching_loss": 0.024390272423624994, + "grad_norm": 0.2464674413204193, + "learning_rate": 2.8650732983788684e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008227530820295214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035469422582536936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02000200403854251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9228312194347381, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21510, + "weighted_flow_matching_loss": 0.024390272423624994, + "weighted_rkd_loss": 0.018458435032516717 + }, + { + "epoch": 6.261274367180681, + "flow_matching_loss": 0.0215745959430933, + "grad_norm": 0.23198947310447693, + "learning_rate": 2.8649019053320133e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007494307076558471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003197028557769954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018185642547905445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9227613508701324, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21520, + "weighted_flow_matching_loss": 0.0215745959430933, + "weighted_rkd_loss": 0.016781020164489745 + }, + { + "epoch": 6.2641838812918245, + "flow_matching_loss": 0.021759853325784208, + "grad_norm": 0.2285957783460617, + "learning_rate": 2.8647304086298658e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008240258414298296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037468680646270515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02022738493978977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9226914405822754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21530, + "weighted_flow_matching_loss": 0.021759853325784208, + "weighted_rkd_loss": 0.018663634173572065 + }, + { + "epoch": 6.267093395402968, + "flow_matching_loss": 0.022810166142880915, + "grad_norm": 0.252444326877594, + "learning_rate": 2.8645588082854502e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008148574130609631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003676890511997044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019974038936197758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9226215004920959, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21540, + "weighted_flow_matching_loss": 0.022810166142880915, + "weighted_rkd_loss": 0.01842845045030117 + }, + { + "epoch": 6.270002909514111, + "flow_matching_loss": 0.022168591618537903, + "grad_norm": 0.29183995723724365, + "learning_rate": 2.864387104311799e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007986977882683276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034974073758348824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019471363350749015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9225515305995942, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21550, + "weighted_flow_matching_loss": 0.022168591618537903, + "weighted_rkd_loss": 0.01796329505741596 + }, + { + "epoch": 6.272912423625255, + "flow_matching_loss": 0.023428916931152344, + "grad_norm": 0.22061195969581604, + "learning_rate": 2.864215296721951e-05, + "loss": 0.0411, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007884084293618799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003398553328588605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019166721776127816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9224815368652344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21560, + "weighted_flow_matching_loss": 0.023428916931152344, + "weighted_rkd_loss": 0.01768096387386322 + }, + { + "epoch": 6.275821937736398, + "flow_matching_loss": 0.02279988192021847, + "grad_norm": 0.2528936564922333, + "learning_rate": 2.864043385528954e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007743126200512052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003446565824560821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01893281824886799, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9224115192890168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21570, + "weighted_flow_matching_loss": 0.02279988192021847, + "weighted_rkd_loss": 0.017463830299675465 + }, + { + "epoch": 6.278731451847541, + "flow_matching_loss": 0.023509651049971582, + "grad_norm": 0.2688572406768799, + "learning_rate": 2.8638713707458644e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008017481118440629, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003605834092013538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01964079635217786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9223414659500122, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21580, + "weighted_flow_matching_loss": 0.023509651049971582, + "weighted_rkd_loss": 0.018115503806620838 + }, + { + "epoch": 6.281640965958685, + "flow_matching_loss": 0.023037115484476088, + "grad_norm": 0.24049654603004456, + "learning_rate": 2.8636992523857442e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007897020224481821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036246546544134615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01941869528964162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9222713828086853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21590, + "weighted_flow_matching_loss": 0.023037115484476088, + "weighted_rkd_loss": 0.017909345962107182 + }, + { + "epoch": 6.284550480069829, + "flow_matching_loss": 0.022508988343179225, + "grad_norm": 0.22439326345920563, + "learning_rate": 2.863527030461666e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007404724042862654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032255297526717188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018034977838397026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9222012639045716, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21600, + "weighted_flow_matching_loss": 0.022508988343179225, + "weighted_rkd_loss": 0.016631858702749015 + }, + { + "epoch": 6.2874599941809715, + "flow_matching_loss": 0.02387847900390625, + "grad_norm": 0.2538411617279053, + "learning_rate": 2.863354704986708e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008043725322932006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036467877449467777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019734238274395466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9221311330795288, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21610, + "weighted_flow_matching_loss": 0.02387847900390625, + "weighted_rkd_loss": 0.018197538796812296 + }, + { + "epoch": 6.290369508292115, + "flow_matching_loss": 0.022120461240410803, + "grad_norm": 0.21979846060276031, + "learning_rate": 2.8631822759739578e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007478073984384537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003134677978232503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018090826086699964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9220609664916992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21620, + "weighted_flow_matching_loss": 0.022120461240410803, + "weighted_rkd_loss": 0.016680873837321995 + }, + { + "epoch": 6.293279022403259, + "flow_matching_loss": 0.023183738440275194, + "grad_norm": 0.23742768168449402, + "learning_rate": 2.8630097434365096e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007829674426466227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036487648263573646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019308113679289817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9219907701015473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21630, + "weighted_flow_matching_loss": 0.023183738440275194, + "weighted_rkd_loss": 0.017801903747022153 + }, + { + "epoch": 6.296188536514402, + "flow_matching_loss": 0.02248765919357538, + "grad_norm": 0.3383052945137024, + "learning_rate": 2.8628371073874665e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009007902583107352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003976269694976509, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021992075070738793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9219205439090729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21640, + "weighted_flow_matching_loss": 0.02248765919357538, + "weighted_rkd_loss": 0.020274939946830273 + }, + { + "epoch": 6.299098050625545, + "flow_matching_loss": 0.022094602324068548, + "grad_norm": 0.25180521607398987, + "learning_rate": 2.8626643678399393e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007405233522877097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003360485821031034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018170952890068294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9218502879142761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21650, + "weighted_flow_matching_loss": 0.022094602324068548, + "weighted_rkd_loss": 0.016750854533165695 + }, + { + "epoch": 6.302007564736689, + "flow_matching_loss": 0.023455921187996866, + "grad_norm": 0.23027220368385315, + "learning_rate": 2.862491524807046e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00803709914907813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003450711793266237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019524909928441046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.921779990196228, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21660, + "weighted_flow_matching_loss": 0.023455921187996866, + "weighted_rkd_loss": 0.01799766141921282 + }, + { + "epoch": 6.304917078847833, + "flow_matching_loss": 0.02115523535758257, + "grad_norm": 0.24581928551197052, + "learning_rate": 2.8623185783019133e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007722567254677415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003320542979054153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01876567741855979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9217097103595734, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21670, + "weighted_flow_matching_loss": 0.02115523535758257, + "weighted_rkd_loss": 0.01729648681357503 + }, + { + "epoch": 6.307826592958976, + "flow_matching_loss": 0.020615505427122115, + "grad_norm": 0.1895299106836319, + "learning_rate": 2.862145528337675e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007422599801793694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032713203923776745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018116520158946513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9216393411159516, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21680, + "weighted_flow_matching_loss": 0.020615505427122115, + "weighted_rkd_loss": 0.016696920804679393 + }, + { + "epoch": 6.310736107070119, + "flow_matching_loss": 0.022197546623647212, + "grad_norm": 0.19192449748516083, + "learning_rate": 2.861972374927473e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007879446214064955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033539455849677323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019112838059663774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9215689778327942, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21690, + "weighted_flow_matching_loss": 0.022197546623647212, + "weighted_rkd_loss": 0.01761384531855583 + }, + { + "epoch": 6.313645621181263, + "flow_matching_loss": 0.021129999682307242, + "grad_norm": 0.20885063707828522, + "learning_rate": 2.8617991180844575e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008525748737156391, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00372411347925663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020775610860437155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9214985787868499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21700, + "weighted_flow_matching_loss": 0.021129999682307242, + "weighted_rkd_loss": 0.01914469264447689 + }, + { + "epoch": 6.316555135292406, + "flow_matching_loss": 0.021658410131931306, + "grad_norm": 0.19886426627635956, + "learning_rate": 2.861625757821786e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007206846494227648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003231225605122745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017644918616861106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9214281499385834, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21710, + "weighted_flow_matching_loss": 0.021658410131931306, + "weighted_rkd_loss": 0.01625855388119817 + }, + { + "epoch": 6.31946464940355, + "flow_matching_loss": 0.0219474283978343, + "grad_norm": 0.2557066082954407, + "learning_rate": 2.861452294152624e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007780329464003444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034345845226198433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018995243497192858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9213576912879944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21720, + "weighted_flow_matching_loss": 0.0219474283978343, + "weighted_rkd_loss": 0.017501422110944985 + }, + { + "epoch": 6.322374163514693, + "flow_matching_loss": 0.02209309358149767, + "grad_norm": 0.1999579221010208, + "learning_rate": 2.8612787270901458e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075275029987096785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031869333470240235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018241939414292575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.921287202835083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21730, + "weighted_flow_matching_loss": 0.02209309358149767, + "weighted_rkd_loss": 0.016806098725646736 + }, + { + "epoch": 6.325283677625836, + "flow_matching_loss": 0.022442873939871787, + "grad_norm": 0.3047383725643158, + "learning_rate": 2.861105056647531e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007939124247059226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036889272974804043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019567175675183533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9212166965007782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21740, + "weighted_flow_matching_loss": 0.022442873939871787, + "weighted_rkd_loss": 0.01802559457719326 + }, + { + "epoch": 6.32819319173698, + "flow_matching_loss": 0.02283746860921383, + "grad_norm": 0.20517560839653015, + "learning_rate": 2.86093128283797e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007545499224215746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032310519367456437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018322050385177136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9211461484432221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21750, + "weighted_flow_matching_loss": 0.02283746860921383, + "weighted_rkd_loss": 0.01687732618302107 + }, + { + "epoch": 6.3311027058481235, + "flow_matching_loss": 0.021889721136540176, + "grad_norm": 0.20957806706428528, + "learning_rate": 2.8607574056746595e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008153313165530562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003668572660535574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0199751989915967, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.921075576543808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21760, + "weighted_flow_matching_loss": 0.021889721136540176, + "weighted_rkd_loss": 0.018398668617010117 + }, + { + "epoch": 6.334012219959266, + "flow_matching_loss": 0.025336302258074282, + "grad_norm": 0.2079809606075287, + "learning_rate": 2.8605834251708044e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0081641492433846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003587554395198822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019915853254497053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9210049748420716, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21770, + "weighted_flow_matching_loss": 0.025336302258074282, + "weighted_rkd_loss": 0.018342608585953712 + }, + { + "epoch": 6.33692173407041, + "flow_matching_loss": 0.021326636523008348, + "grad_norm": 0.2809374928474426, + "learning_rate": 2.8604093413396172e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00749111189506948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033250229666009546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01830724673345685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9209343492984772, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21780, + "weighted_flow_matching_loss": 0.021326636523008348, + "weighted_rkd_loss": 0.016859738901257516 + }, + { + "epoch": 6.339831248181554, + "flow_matching_loss": 0.024158889800310133, + "grad_norm": 0.2321476936340332, + "learning_rate": 2.8602351541943184e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008164919540286064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003686165879480541, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020016005262732506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9208636820316315, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21790, + "weighted_flow_matching_loss": 0.024158889800310133, + "weighted_rkd_loss": 0.018432029243558645 + }, + { + "epoch": 6.3427407622926975, + "flow_matching_loss": 0.021674088947474957, + "grad_norm": 0.2592616677284241, + "learning_rate": 2.8600608637481364e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008521490637212992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039008301217108967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020943811163306236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9207929968833923, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21800, + "weighted_flow_matching_loss": 0.021674088947474957, + "weighted_rkd_loss": 0.019284902978688477 + }, + { + "epoch": 6.34565027640384, + "flow_matching_loss": 0.0201641789637506, + "grad_norm": 0.2556638717651367, + "learning_rate": 2.859886470014307e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074565749615430835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003368278266862035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018281428050249814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9207222878932952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21810, + "weighted_flow_matching_loss": 0.0201641789637506, + "weighted_rkd_loss": 0.016832113172858952 + }, + { + "epoch": 6.348559790514984, + "flow_matching_loss": 0.02296834662556648, + "grad_norm": 0.24545526504516602, + "learning_rate": 2.8597119730060746e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008383220667019487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038021829444915057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02056862451136112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9206515312194824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21820, + "weighted_flow_matching_loss": 0.02296834662556648, + "weighted_rkd_loss": 0.018936529848724605 + }, + { + "epoch": 6.351469304626128, + "flow_matching_loss": 0.02190400827676058, + "grad_norm": 0.19257186353206635, + "learning_rate": 2.859537372736692e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008702303795143963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038068201392889023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02121142791584134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9205807626247406, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21830, + "weighted_flow_matching_loss": 0.02190400827676058, + "weighted_rkd_loss": 0.019526867568492888 + }, + { + "epoch": 6.3543788187372705, + "flow_matching_loss": 0.021740720234811307, + "grad_norm": 0.23962411284446716, + "learning_rate": 2.8593626692194174e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007784695504233241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035748762311413884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019144267402589323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9205099523067475, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21840, + "weighted_flow_matching_loss": 0.021740720234811307, + "weighted_rkd_loss": 0.01762245995923877 + }, + { + "epoch": 6.357288332848414, + "flow_matching_loss": 0.023127919621765615, + "grad_norm": 0.2556530833244324, + "learning_rate": 2.8591878624675197e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00821976657025516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003445110144093633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01988464305177331, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9204391241073608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21850, + "weighted_flow_matching_loss": 0.023127919621765615, + "weighted_rkd_loss": 0.01830266136676073 + }, + { + "epoch": 6.360197846959558, + "flow_matching_loss": 0.02151587065309286, + "grad_norm": 0.23158608376979828, + "learning_rate": 2.8590129524942735e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008293863292783499, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003815508144907653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02040323503315449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9203682661056518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21860, + "weighted_flow_matching_loss": 0.02151587065309286, + "weighted_rkd_loss": 0.0187785055488348 + }, + { + "epoch": 6.363107361070702, + "flow_matching_loss": 0.02358484510332346, + "grad_norm": 0.250933438539505, + "learning_rate": 2.8588379393129622e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070700184907764195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030964251374825833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017236462235450743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9202973544597626, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21870, + "weighted_flow_matching_loss": 0.02358484510332346, + "weighted_rkd_loss": 0.015862667467445135 + }, + { + "epoch": 6.3660168751818444, + "flow_matching_loss": 0.02296515740454197, + "grad_norm": 0.227028951048851, + "learning_rate": 2.8586628229368765e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008435610868036746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036030791932716967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02047430109232664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9202264249324799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21880, + "weighted_flow_matching_loss": 0.02296515740454197, + "weighted_rkd_loss": 0.0188410515896976 + }, + { + "epoch": 6.368926389292988, + "flow_matching_loss": 0.023991172015666962, + "grad_norm": 0.3078635632991791, + "learning_rate": 2.8584876033793165e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008491676580160857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003743583848699927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020726936869323254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9201554954051971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21890, + "weighted_flow_matching_loss": 0.023991172015666962, + "weighted_rkd_loss": 0.0190719504840672 + }, + { + "epoch": 6.371835903404132, + "flow_matching_loss": 0.02158594485372305, + "grad_norm": 0.21556760370731354, + "learning_rate": 2.8583122806535876e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008298354363068939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037266036262735723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020323312561959028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9200845301151276, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21900, + "weighted_flow_matching_loss": 0.02158594485372305, + "weighted_rkd_loss": 0.01869918154552579 + }, + { + "epoch": 6.374745417515275, + "flow_matching_loss": 0.022500307485461234, + "grad_norm": 0.28499636054039, + "learning_rate": 2.8581368547730057e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00829024608246982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003947449638508261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02052794164046645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9200135231018066, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21910, + "weighted_flow_matching_loss": 0.022500307485461234, + "weighted_rkd_loss": 0.018885932583361863 + }, + { + "epoch": 6.377654931626418, + "flow_matching_loss": 0.02398698776960373, + "grad_norm": 0.302505761384964, + "learning_rate": 2.8579613257508925e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008568332297727466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00406156515236944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021198230050504207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9199424862861634, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21920, + "weighted_flow_matching_loss": 0.02398698776960373, + "weighted_rkd_loss": 0.019501139223575593 + }, + { + "epoch": 6.380564445737562, + "flow_matching_loss": 0.02303113481029868, + "grad_norm": 0.24618877470493317, + "learning_rate": 2.857785693600578e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008601393084973097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037814255570992827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02098421137779951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9198714315891265, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21930, + "weighted_flow_matching_loss": 0.02303113481029868, + "weighted_rkd_loss": 0.019302742928266524 + }, + { + "epoch": 6.383473959848705, + "flow_matching_loss": 0.02166641838848591, + "grad_norm": 0.22621752321720123, + "learning_rate": 2.8576099583354006e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008626384800300002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003756123478524387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021008892729878426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9198003351688385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21940, + "weighted_flow_matching_loss": 0.02166641838848591, + "weighted_rkd_loss": 0.019324000366032122 + }, + { + "epoch": 6.386383473959849, + "flow_matching_loss": 0.021464958414435388, + "grad_norm": 0.34200519323349, + "learning_rate": 2.857434119968707e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007676602341234684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033347097458317878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018687914311885833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9197292149066925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21950, + "weighted_flow_matching_loss": 0.021464958414435388, + "weighted_rkd_loss": 0.017187790013849735 + }, + { + "epoch": 6.389292988070992, + "flow_matching_loss": 0.023805159889161587, + "grad_norm": 0.23163746297359467, + "learning_rate": 2.8572581785138498e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008148928405717015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037763341097161174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020074190758168697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9196580708026886, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21960, + "weighted_flow_matching_loss": 0.023805159889161587, + "weighted_rkd_loss": 0.018461412843316795 + }, + { + "epoch": 6.392202502182135, + "flow_matching_loss": 0.021884053573012353, + "grad_norm": 0.20729924738407135, + "learning_rate": 2.857082133984191e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008433480421081186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00355729591101408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020424256939440964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9195868849754334, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21970, + "weighted_flow_matching_loss": 0.021884053573012353, + "weighted_rkd_loss": 0.01878188140690327 + }, + { + "epoch": 6.395112016293279, + "flow_matching_loss": 0.021515484526753427, + "grad_norm": 0.22296656668186188, + "learning_rate": 2.8569059863931006e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007976515218615532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035423124907538297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01949534295126796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9195156753063202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21980, + "weighted_flow_matching_loss": 0.021515484526753427, + "weighted_rkd_loss": 0.017926277685910465 + }, + { + "epoch": 6.398021530404423, + "flow_matching_loss": 0.02020394280552864, + "grad_norm": 0.22601492702960968, + "learning_rate": 2.8567297357539548e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007318932050839067, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003140923613682389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017778787482529878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9194444417953491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21990, + "weighted_flow_matching_loss": 0.02020394280552864, + "weighted_rkd_loss": 0.016346623934805394 + }, + { + "epoch": 6.400931044515566, + "flow_matching_loss": 0.023105383664369584, + "grad_norm": 0.2794063985347748, + "learning_rate": 2.856553382080139e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008971172897145153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004007914313115179, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021950260177254675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9193731844425201, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22000, + "weighted_flow_matching_loss": 0.023105383664369584, + "weighted_rkd_loss": 0.020180487912148236 + }, + { + "epoch": 6.403840558626709, + "flow_matching_loss": 0.022649236582219602, + "grad_norm": 0.328759104013443, + "learning_rate": 2.856376925385047e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008773163612931966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004113826132379472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021660153567790986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9193018913269043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22010, + "weighted_flow_matching_loss": 0.022649236582219602, + "weighted_rkd_loss": 0.01991217117756605 + }, + { + "epoch": 6.406750072737853, + "flow_matching_loss": 0.023531066812574865, + "grad_norm": 0.25582456588745117, + "learning_rate": 2.8562003656820783e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008512364700436592, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039688936900347475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02099362276494503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9192305684089661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22020, + "weighted_flow_matching_loss": 0.023531066812574865, + "weighted_rkd_loss": 0.019297977164387702 + }, + { + "epoch": 6.4096595868489965, + "flow_matching_loss": 0.023395654186606406, + "grad_norm": 0.22105874121189117, + "learning_rate": 2.8560237029846425e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007985747558996082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035088918171823024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01948038712143898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.91915922164917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22030, + "weighted_flow_matching_loss": 0.023395654186606406, + "weighted_rkd_loss": 0.017905565910041332 + }, + { + "epoch": 6.412569100960139, + "flow_matching_loss": 0.02165318187326193, + "grad_norm": 0.24867919087409973, + "learning_rate": 2.8558469373061546e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008400786528363824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003576902812346816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02037847563624382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9190878391265869, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22040, + "weighted_flow_matching_loss": 0.02165318187326193, + "weighted_rkd_loss": 0.018729603663086892 + }, + { + "epoch": 6.415478615071283, + "flow_matching_loss": 0.023526525311172008, + "grad_norm": 0.27222397923469543, + "learning_rate": 2.85567006866004e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008186876215040684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037134930957108737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020087245386093855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.919016432762146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22050, + "weighted_flow_matching_loss": 0.023526525311172008, + "weighted_rkd_loss": 0.01846051514148712 + }, + { + "epoch": 6.418388129182427, + "flow_matching_loss": 0.023342701606452465, + "grad_norm": 0.20186111330986023, + "learning_rate": 2.8554930970597306e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00753692309372127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033388728043064475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01841271873563528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9189449965953826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22060, + "weighted_flow_matching_loss": 0.023342701606452465, + "weighted_rkd_loss": 0.01692027784883976 + }, + { + "epoch": 6.4212976432935704, + "flow_matching_loss": 0.019822884071618318, + "grad_norm": 0.24533186852931976, + "learning_rate": 2.8553160225186657e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008436898281797767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036894588032737375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020563255436718463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9188735365867615, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22070, + "weighted_flow_matching_loss": 0.019822884071618318, + "weighted_rkd_loss": 0.018895053863525392 + }, + { + "epoch": 6.424207157404713, + "flow_matching_loss": 0.022368496004492044, + "grad_norm": 0.1974335014820099, + "learning_rate": 2.8551388450502933e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008137603616341948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037083850475028157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01998359216377139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.91880202293396, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22080, + "weighted_flow_matching_loss": 0.022368496004492044, + "weighted_rkd_loss": 0.01836104514077306 + }, + { + "epoch": 6.427116671515857, + "flow_matching_loss": 0.025799683295190334, + "grad_norm": 0.25690147280693054, + "learning_rate": 2.854961564668069e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0078911735676229, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035909216618165373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01937326891347766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9187304973602295, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22090, + "weighted_flow_matching_loss": 0.025799683295190334, + "weighted_rkd_loss": 0.01779880179092288 + }, + { + "epoch": 6.430026185627001, + "flow_matching_loss": 0.02284196615219116, + "grad_norm": 0.26093462109565735, + "learning_rate": 2.8547841813854555e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008241859963163733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003693857043981552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020177576504647732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.918658971786499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22100, + "weighted_flow_matching_loss": 0.02284196615219116, + "weighted_rkd_loss": 0.01853628931567073 + }, + { + "epoch": 6.4329356997381435, + "flow_matching_loss": 0.022999350726604462, + "grad_norm": 0.2128739356994629, + "learning_rate": 2.8546066952159244e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007302814582362771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031303423456847668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017735971231013538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9185873806476593, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22110, + "weighted_flow_matching_loss": 0.022999350726604462, + "weighted_rkd_loss": 0.01629204899072647 + }, + { + "epoch": 6.435845213849287, + "flow_matching_loss": 0.02193032028153539, + "grad_norm": 0.2044125497341156, + "learning_rate": 2.854429106172954e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00832020929083228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003619402414187789, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02025982104241848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9185157716274261, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22120, + "weighted_flow_matching_loss": 0.02193032028153539, + "weighted_rkd_loss": 0.01860899468883872 + }, + { + "epoch": 6.438754727960431, + "flow_matching_loss": 0.02184986099600792, + "grad_norm": 0.22880403697490692, + "learning_rate": 2.8542514142700323e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007738352008163929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003247916046530008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018724620062857866, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.918444138765335, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22130, + "weighted_flow_matching_loss": 0.02184986099600792, + "weighted_rkd_loss": 0.01719751553609967 + }, + { + "epoch": 6.441664242071574, + "flow_matching_loss": 0.020154720358550548, + "grad_norm": 0.2588292062282562, + "learning_rate": 2.8540736195206523e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007569128694012761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031423795269802214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018280636891722678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9183724641799926, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22140, + "weighted_flow_matching_loss": 0.020154720358550548, + "weighted_rkd_loss": 0.01678844029083848 + }, + { + "epoch": 6.444573756182717, + "flow_matching_loss": 0.02314592879265547, + "grad_norm": 0.18893291056156158, + "learning_rate": 2.8538957219383168e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00846850099042058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003572787414304912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020509789511561394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9183007776737213, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22150, + "weighted_flow_matching_loss": 0.02314592879265547, + "weighted_rkd_loss": 0.01883417572826147 + }, + { + "epoch": 6.447483270293861, + "flow_matching_loss": 0.020484021026641132, + "grad_norm": 0.19733594357967377, + "learning_rate": 2.8537177215365363e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007894275384023786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003532273950986564, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019320824556052683, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.918229055404663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22160, + "weighted_flow_matching_loss": 0.020484021026641132, + "weighted_rkd_loss": 0.017740958277136086 + }, + { + "epoch": 6.450392784405004, + "flow_matching_loss": 0.021329620294272898, + "grad_norm": 0.2830113470554352, + "learning_rate": 2.853539618328829e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007837308570742607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003472656896337867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01914727399125695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9181573033332825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22170, + "weighted_flow_matching_loss": 0.021329620294272898, + "weighted_rkd_loss": 0.017580184154212475 + }, + { + "epoch": 6.453302298516148, + "flow_matching_loss": 0.022823082096874715, + "grad_norm": 0.24289570748806, + "learning_rate": 2.8533614123287203e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007177564268931747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003247104329057038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017602233029901982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.918085515499115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22180, + "weighted_flow_matching_loss": 0.022823082096874715, + "weighted_rkd_loss": 0.016160303447395562 + }, + { + "epoch": 6.456211812627291, + "flow_matching_loss": 0.02259146422147751, + "grad_norm": 0.2845386862754822, + "learning_rate": 2.8531831035497435e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008033404685556888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035593442153185606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019626153726130725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.918013721704483, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22190, + "weighted_flow_matching_loss": 0.02259146422147751, + "weighted_rkd_loss": 0.01801708871498704 + }, + { + "epoch": 6.459121326738435, + "flow_matching_loss": 0.022277965396642684, + "grad_norm": 0.21711644530296326, + "learning_rate": 2.8530046920054406e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008018800802528859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034443529788404704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019481954351067542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9179418742656708, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22200, + "weighted_flow_matching_loss": 0.022277965396642684, + "weighted_rkd_loss": 0.017883315868675708 + }, + { + "epoch": 6.462030840849578, + "flow_matching_loss": 0.020866508595645428, + "grad_norm": 0.20808139443397522, + "learning_rate": 2.85282617770936e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007922394247725607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003669805941171944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01951459441334009, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9178700029850007, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22210, + "weighted_flow_matching_loss": 0.020866508595645428, + "weighted_rkd_loss": 0.017911871336400507 + }, + { + "epoch": 6.464940354960722, + "flow_matching_loss": 0.021805425733327867, + "grad_norm": 0.2855854332447052, + "learning_rate": 2.8526475606750585e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00780373583547771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036419920390471814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019249463640153407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9177981078624725, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22220, + "weighted_flow_matching_loss": 0.021805425733327867, + "weighted_rkd_loss": 0.01766712898388505 + }, + { + "epoch": 6.467849869071865, + "flow_matching_loss": 0.021653434634208678, + "grad_norm": 0.17285466194152832, + "learning_rate": 2.8524688409161022e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007853180728852749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003650366864167154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019356728438287972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9177261888980865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22230, + "weighted_flow_matching_loss": 0.021653434634208678, + "weighted_rkd_loss": 0.017764228954911233 + }, + { + "epoch": 6.470759383183008, + "flow_matching_loss": 0.02213554233312607, + "grad_norm": 0.2199404537677765, + "learning_rate": 2.852290018446063e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007913721026852727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003513071220368147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019340513460338115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9176542341709137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22240, + "weighted_flow_matching_loss": 0.02213554233312607, + "weighted_rkd_loss": 0.017747884057462214 + }, + { + "epoch": 6.473668897294152, + "flow_matching_loss": 0.02308257408440113, + "grad_norm": 0.2572147846221924, + "learning_rate": 2.852111093278521e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007642928650602699, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003503280319273472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018789137806743385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9175822496414184, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22250, + "weighted_flow_matching_loss": 0.02308257408440113, + "weighted_rkd_loss": 0.017240604478865863 + }, + { + "epoch": 6.4765784114052956, + "flow_matching_loss": 0.022664212714880706, + "grad_norm": 0.26064321398735046, + "learning_rate": 2.8519320654270645e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008081477647647262, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003581763245165348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019744718633592128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9175102412700653, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22260, + "weighted_flow_matching_loss": 0.022664212714880706, + "weighted_rkd_loss": 0.018115957360714675 + }, + { + "epoch": 6.479487925516438, + "flow_matching_loss": 0.018941254634410144, + "grad_norm": 0.20114141702651978, + "learning_rate": 2.8517529349052898e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007679248182103038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034522446570917965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018810741137713194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9174382090568542, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22270, + "weighted_flow_matching_loss": 0.018941254634410144, + "weighted_rkd_loss": 0.01725772339850664 + }, + { + "epoch": 6.482397439627582, + "flow_matching_loss": 0.02201097197830677, + "grad_norm": 0.23438714444637299, + "learning_rate": 2.8515737017268007e-05, + "loss": 0.0396, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007952302554622292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032723759533837436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01917698122560978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9173661351203919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22280, + "weighted_flow_matching_loss": 0.02201097197830677, + "weighted_rkd_loss": 0.01759228892624378 + }, + { + "epoch": 6.485306953738726, + "flow_matching_loss": 0.02069000545889139, + "grad_norm": 0.21497495472431183, + "learning_rate": 2.8513943659052083e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072872302494943145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033183413790538905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017892802041023968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.917294055223465, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22290, + "weighted_flow_matching_loss": 0.02069000545889139, + "weighted_rkd_loss": 0.01641294788569212 + }, + { + "epoch": 6.488216467849869, + "flow_matching_loss": 0.02183196656405926, + "grad_norm": 0.2609002888202667, + "learning_rate": 2.8512149274541323e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007725191302597523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003424345050007105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01887472756206989, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9172219336032867, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22300, + "weighted_flow_matching_loss": 0.02183196656405926, + "weighted_rkd_loss": 0.017312338296324014 + }, + { + "epoch": 6.491125981961012, + "flow_matching_loss": 0.025743844732642174, + "grad_norm": 0.2262362688779831, + "learning_rate": 2.8510353863871995e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007553676748648286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034136970760300757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0185210507363081, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9171497583389282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22310, + "weighted_flow_matching_loss": 0.025743844732642174, + "weighted_rkd_loss": 0.01698657190427184 + }, + { + "epoch": 6.494035496072156, + "flow_matching_loss": 0.022718091122806074, + "grad_norm": 0.2596658766269684, + "learning_rate": 2.8508557427180456e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008212412893772125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034410969354212286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01986592272296548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9170775890350342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22320, + "weighted_flow_matching_loss": 0.022718091122806074, + "weighted_rkd_loss": 0.018218624778091906 + }, + { + "epoch": 6.4969450101833, + "flow_matching_loss": 0.023805245384573938, + "grad_norm": 0.2149401754140854, + "learning_rate": 2.8506759964603127e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008084063744172454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037415451603010296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019909672532230617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9170053660869598, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22330, + "weighted_flow_matching_loss": 0.023805245384573938, + "weighted_rkd_loss": 0.018257272709161044 + }, + { + "epoch": 6.4998545242944425, + "flow_matching_loss": 0.021687335334718227, + "grad_norm": 0.27564775943756104, + "learning_rate": 2.850496147627652e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007927416451275348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003377929888665676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01923276297748089, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9169331371784211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22340, + "weighted_flow_matching_loss": 0.021687335334718227, + "weighted_rkd_loss": 0.017635127902030943 + }, + { + "epoch": 6.502764038405586, + "flow_matching_loss": 0.020318549312651157, + "grad_norm": 0.19763708114624023, + "learning_rate": 2.8503161962337206e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007245051069185137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00306294874753803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017553051002323627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9168608665466309, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22350, + "weighted_flow_matching_loss": 0.020318549312651157, + "weighted_rkd_loss": 0.01609372505918145 + }, + { + "epoch": 6.50567355251673, + "flow_matching_loss": 0.02084148209542036, + "grad_norm": 0.23456446826457977, + "learning_rate": 2.850136142292186e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006818258017301559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032128846272826196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016849400755017997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9167885780334473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22360, + "weighted_flow_matching_loss": 0.02084148209542036, + "weighted_rkd_loss": 0.015447333548218012 + }, + { + "epoch": 6.508583066627873, + "flow_matching_loss": 0.02073051258921623, + "grad_norm": 0.25654885172843933, + "learning_rate": 2.849955985816721e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007827564934268593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036302538122981788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019285383727401496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9167162477970123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22370, + "weighted_flow_matching_loss": 0.02073051258921623, + "weighted_rkd_loss": 0.01767916465178132 + }, + { + "epoch": 6.5114925807390165, + "flow_matching_loss": 0.02149015348404646, + "grad_norm": 0.2545336186885834, + "learning_rate": 2.8497757268210085e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075075904373079535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003276947303675115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01829212820157409, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9166438996791839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22380, + "weighted_flow_matching_loss": 0.02149015348404646, + "weighted_rkd_loss": 0.016767369769513606 + }, + { + "epoch": 6.51440209485016, + "flow_matching_loss": 0.02246814761310816, + "grad_norm": 0.2548089623451233, + "learning_rate": 2.8495953653187364e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007013146765530109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030544799636118116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017080773320049047, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9165715157985688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22390, + "weighted_flow_matching_loss": 0.02246814761310816, + "weighted_rkd_loss": 0.015655734669417144 + }, + { + "epoch": 6.517311608961304, + "flow_matching_loss": 0.021505888737738132, + "grad_norm": 0.23506397008895874, + "learning_rate": 2.8494149013236038e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007341987639665604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033404660178348424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018024441506713628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9164991080760956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22400, + "weighted_flow_matching_loss": 0.021505888737738132, + "weighted_rkd_loss": 0.016519348695874216 + }, + { + "epoch": 6.520221123072447, + "flow_matching_loss": 0.021553476713597773, + "grad_norm": 0.20219236612319946, + "learning_rate": 2.8492343348493145e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007127714389935136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00297850682400167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01723393555730581, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9164266645908355, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22410, + "weighted_flow_matching_loss": 0.021553476713597773, + "weighted_rkd_loss": 0.015793650690466166 + }, + { + "epoch": 6.52313063718359, + "flow_matching_loss": 0.02226635944098234, + "grad_norm": 0.22086982429027557, + "learning_rate": 2.8490536659095818e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007443016208708286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033309820340946318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01821701433509588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9163542032241822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22420, + "weighted_flow_matching_loss": 0.02226635944098234, + "weighted_rkd_loss": 0.01669325092807412 + }, + { + "epoch": 6.526040151294734, + "flow_matching_loss": 0.02399229984730482, + "grad_norm": 0.21383067965507507, + "learning_rate": 2.8488728945181257e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007803938258439303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037133360980078577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01932121245190501, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9162817060947418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22430, + "weighted_flow_matching_loss": 0.02399229984730482, + "weighted_rkd_loss": 0.017703699879348278 + }, + { + "epoch": 6.528949665405877, + "flow_matching_loss": 0.022355486080050467, + "grad_norm": 0.2103848159313202, + "learning_rate": 2.8486920206886752e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008115872275084257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037918815156444906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020023625902831556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9162091791629792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22440, + "weighted_flow_matching_loss": 0.022355486080050467, + "weighted_rkd_loss": 0.018345825374126434 + }, + { + "epoch": 6.531859179517021, + "flow_matching_loss": 0.019830034393817187, + "grad_norm": 0.20728819072246552, + "learning_rate": 2.848511044434966e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00745204659178853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003251687274314463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018155780248343946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.916136622428894, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22450, + "weighted_flow_matching_loss": 0.019830034393817187, + "weighted_rkd_loss": 0.016633183788508178 + }, + { + "epoch": 6.534768693628164, + "flow_matching_loss": 0.01921029658988118, + "grad_norm": 0.21482224762439728, + "learning_rate": 2.848329965770743e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007715562311932445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003342576290015131, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018773700948804616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9160640418529511, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22460, + "weighted_flow_matching_loss": 0.01921029658988118, + "weighted_rkd_loss": 0.017197917401790618 + }, + { + "epoch": 6.537678207739307, + "flow_matching_loss": 0.02357873283326626, + "grad_norm": 0.2190832495689392, + "learning_rate": 2.848148784709757e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007710892194882035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033592839958146215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01878106836229563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9159914374351501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22470, + "weighted_flow_matching_loss": 0.02357873283326626, + "weighted_rkd_loss": 0.01720329523086548 + }, + { + "epoch": 6.540587721850451, + "flow_matching_loss": 0.023820359259843826, + "grad_norm": 0.2417314648628235, + "learning_rate": 2.8479675012657677e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007048207195475697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029550758423283695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01705149030312896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9159187972545624, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22480, + "weighted_flow_matching_loss": 0.023820359259843826, + "weighted_rkd_loss": 0.015617773123085498 + }, + { + "epoch": 6.543497235961595, + "flow_matching_loss": 0.02304745027795434, + "grad_norm": 0.21384169161319733, + "learning_rate": 2.8477861154525427e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007514471467584371, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034345082938671114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01846345141530037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9158461272716523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22490, + "weighted_flow_matching_loss": 0.02304745027795434, + "weighted_rkd_loss": 0.016909689735621215 + }, + { + "epoch": 6.546406750072737, + "flow_matching_loss": 0.02240695431828499, + "grad_norm": 0.24191196262836456, + "learning_rate": 2.847604627283857e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007771387696266174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034825366688892244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019025311805307865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9157734513282776, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22500, + "weighted_flow_matching_loss": 0.02240695431828499, + "weighted_rkd_loss": 0.017422887682914733 + }, + { + "epoch": 6.549316264183881, + "flow_matching_loss": 0.0213980526663363, + "grad_norm": 0.2413102686405182, + "learning_rate": 2.847423036773493e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007171650743111968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003228906192816794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017572207469493152, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9157007277011872, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22510, + "weighted_flow_matching_loss": 0.0213980526663363, + "weighted_rkd_loss": 0.01609083665534854 + }, + { + "epoch": 6.552225778295025, + "flow_matching_loss": 0.021999814733862876, + "grad_norm": 0.2495863139629364, + "learning_rate": 2.8472413439352416e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007412574160844088, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033974552527070044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018222603667527437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9156279563903809, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22520, + "weighted_flow_matching_loss": 0.021999814733862876, + "weighted_rkd_loss": 0.016685093101114035 + }, + { + "epoch": 6.5551352924061685, + "flow_matching_loss": 0.021980537474155425, + "grad_norm": 0.2587433457374573, + "learning_rate": 2.847059548782901e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007143653742969036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003101063962094486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01738837147131562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9155551791191101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22530, + "weighted_flow_matching_loss": 0.021980537474155425, + "weighted_rkd_loss": 0.015919973514974116 + }, + { + "epoch": 6.558044806517311, + "flow_matching_loss": 0.021388989686965943, + "grad_norm": 0.2359149158000946, + "learning_rate": 2.8468776513302774e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008322619460523128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003621882270090282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020267121121287347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.915482372045517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22540, + "weighted_flow_matching_loss": 0.021388989686965943, + "weighted_rkd_loss": 0.01855418672785163 + }, + { + "epoch": 6.560954320628455, + "flow_matching_loss": 0.021368441917002202, + "grad_norm": 0.21842549741268158, + "learning_rate": 2.8466956515911854e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00760780549608171, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034269213676452635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018642532266676427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.915409529209137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22550, + "weighted_flow_matching_loss": 0.021368441917002202, + "weighted_rkd_loss": 0.017065555974841117 + }, + { + "epoch": 6.563863834739599, + "flow_matching_loss": 0.018714205268770455, + "grad_norm": 0.25125423073768616, + "learning_rate": 2.8465135495794458e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007897739531472325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033638391410931944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01915931813418865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9153366684913635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22560, + "weighted_flow_matching_loss": 0.018714205268770455, + "weighted_rkd_loss": 0.017537211161106826 + }, + { + "epoch": 6.566773348850742, + "flow_matching_loss": 0.023218230344355106, + "grad_norm": 0.30249834060668945, + "learning_rate": 2.8463313453088882e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008403017604723572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003908362705260515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02071439791470766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9152637720108032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22570, + "weighted_flow_matching_loss": 0.023218230344355106, + "weighted_rkd_loss": 0.01895907772704959 + }, + { + "epoch": 6.569682862961885, + "flow_matching_loss": 0.02114482633769512, + "grad_norm": 0.22414863109588623, + "learning_rate": 2.8461490387933498e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007891260366886855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003470728173851967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01925324909389019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9151908457279205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22580, + "weighted_flow_matching_loss": 0.02114482633769512, + "weighted_rkd_loss": 0.0176204152405262 + }, + { + "epoch": 6.572592377073029, + "flow_matching_loss": 0.021023448556661606, + "grad_norm": 0.2581407129764557, + "learning_rate": 2.8459666300466762e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00667177252471447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002856071782298386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016199616622179745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9151178956031799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22590, + "weighted_flow_matching_loss": 0.021023448556661606, + "weighted_rkd_loss": 0.014824523217976093 + }, + { + "epoch": 6.575501891184173, + "flow_matching_loss": 0.023599282652139664, + "grad_norm": 0.21780550479888916, + "learning_rate": 2.845784119082719e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008225556323304772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003449639701284468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01990075223147869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9150449156761169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22600, + "weighted_flow_matching_loss": 0.023599282652139664, + "weighted_rkd_loss": 0.0182101015932858 + }, + { + "epoch": 6.5784114052953155, + "flow_matching_loss": 0.02307963613420725, + "grad_norm": 0.2415223866701126, + "learning_rate": 2.84560150591534e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008184484113007784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034591431729495527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01982811139896512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9149719178676605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22610, + "weighted_flow_matching_loss": 0.02307963613420725, + "weighted_rkd_loss": 0.018142147921025753 + }, + { + "epoch": 6.581320919406459, + "flow_matching_loss": 0.023092657513916492, + "grad_norm": 0.23166212439537048, + "learning_rate": 2.8454187905584063e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008282938925549387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003974596667103469, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020540474727749825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9148988664150238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22620, + "weighted_flow_matching_loss": 0.023092657513916492, + "weighted_rkd_loss": 0.018792498763650654 + }, + { + "epoch": 6.584230433517602, + "flow_matching_loss": 0.022385959699749946, + "grad_norm": 0.22731640934944153, + "learning_rate": 2.8452359730257946e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007730966014787555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034293411299586296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01889127278700471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9148258030414581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22630, + "weighted_flow_matching_loss": 0.022385959699749946, + "weighted_rkd_loss": 0.01728223990648985 + }, + { + "epoch": 6.587139947628746, + "flow_matching_loss": 0.02189868539571762, + "grad_norm": 0.21782837808132172, + "learning_rate": 2.845053053331389e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00752029693685472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033128622686490416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018353455979377032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9147527158260346, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22640, + "weighted_flow_matching_loss": 0.02189868539571762, + "weighted_rkd_loss": 0.016788871586322786 + }, + { + "epoch": 6.5900494617398895, + "flow_matching_loss": 0.019337622355669735, + "grad_norm": 0.20878642797470093, + "learning_rate": 2.8448700314890804e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007567785726860165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00324960770085454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018385179340839386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9146795928478241, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22650, + "weighted_flow_matching_loss": 0.019337622355669735, + "weighted_rkd_loss": 0.01681658187881112 + }, + { + "epoch": 6.592958975851033, + "flow_matching_loss": 0.02120960745960474, + "grad_norm": 0.23592878878116608, + "learning_rate": 2.8446869075127678e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007464184844866395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035283554578199984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018456725217401982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9146064341068267, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22660, + "weighted_flow_matching_loss": 0.02120960745960474, + "weighted_rkd_loss": 0.01688059829175472 + }, + { + "epoch": 6.595868489962176, + "flow_matching_loss": 0.019871667213737964, + "grad_norm": 0.21939389407634735, + "learning_rate": 2.844503681416359e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007191277854144573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034442493226379154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01782680517062545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.914533257484436, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22670, + "weighted_flow_matching_loss": 0.019871667213737964, + "weighted_rkd_loss": 0.016303225699812175 + }, + { + "epoch": 6.59877800407332, + "flow_matching_loss": 0.021809789724647997, + "grad_norm": 0.1709057241678238, + "learning_rate": 2.844320353213769e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007326192362233997, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003225308982655406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017877693753689528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9144600570201874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22680, + "weighted_flow_matching_loss": 0.021809789724647997, + "weighted_rkd_loss": 0.01634847018867731 + }, + { + "epoch": 6.601687518184463, + "flow_matching_loss": 0.020425502397119998, + "grad_norm": 0.24582087993621826, + "learning_rate": 2.844136922918919e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007569498289376498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033164416207000615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018455438129603863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9143868207931518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22690, + "weighted_flow_matching_loss": 0.020425502397119998, + "weighted_rkd_loss": 0.016875444538891317 + }, + { + "epoch": 6.604597032295606, + "flow_matching_loss": 0.02462596818804741, + "grad_norm": 0.2108890414237976, + "learning_rate": 2.8439533905457412e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007973317801952363, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003372331475839019, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01931896712630987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.914313554763794, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22700, + "weighted_flow_matching_loss": 0.02462596818804741, + "weighted_rkd_loss": 0.017663621716201304 + }, + { + "epoch": 6.60750654640675, + "flow_matching_loss": 0.02175913555547595, + "grad_norm": 0.20666919648647308, + "learning_rate": 2.8437697561081718e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007558500580489636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032731736544519665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018390174582600595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9142402708530426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22710, + "weighted_flow_matching_loss": 0.02175913555547595, + "weighted_rkd_loss": 0.016813060268759727 + }, + { + "epoch": 6.610416060517894, + "flow_matching_loss": 0.021267142705619337, + "grad_norm": 0.21916437149047852, + "learning_rate": 2.8435860196201583e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007707719132304192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003178408369421959, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018593846634030343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9141669571399689, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22720, + "weighted_flow_matching_loss": 0.021267142705619337, + "weighted_rkd_loss": 0.016997897811233998 + }, + { + "epoch": 6.613325574629037, + "flow_matching_loss": 0.023440034501254557, + "grad_norm": 0.21729017794132233, + "learning_rate": 2.8434021810956526e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007767262309789658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033620735164731743, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01889659808948636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9140935838222504, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22730, + "weighted_flow_matching_loss": 0.023440034501254557, + "weighted_rkd_loss": 0.017273250315338374 + }, + { + "epoch": 6.61623508874018, + "flow_matching_loss": 0.02116701677441597, + "grad_norm": 0.2412031888961792, + "learning_rate": 2.8432182405486177e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007915113354101777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003493432234972715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019323659036308526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9140202164649963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22740, + "weighted_flow_matching_loss": 0.02116701677441597, + "weighted_rkd_loss": 0.01766218915581703 + }, + { + "epoch": 6.619144602851324, + "flow_matching_loss": 0.022266166657209395, + "grad_norm": 0.29569560289382935, + "learning_rate": 2.8430341979930218e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00830600461922586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003766418294981122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02037842748686671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.913946807384491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22750, + "weighted_flow_matching_loss": 0.022266166657209395, + "weighted_rkd_loss": 0.018624802213162185 + }, + { + "epoch": 6.622054116962468, + "flow_matching_loss": 0.022158190421760083, + "grad_norm": 0.2371203750371933, + "learning_rate": 2.842850053442842e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007345173554494977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003238114877603948, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017928461823612452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9138733685016632, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22760, + "weighted_flow_matching_loss": 0.022158190421760083, + "weighted_rkd_loss": 0.016384334210306405 + }, + { + "epoch": 6.62496363107361, + "flow_matching_loss": 0.021425329241901638, + "grad_norm": 0.2364511340856552, + "learning_rate": 2.8426658069120625e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007755533466115594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003477913304232061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018988980259746314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.913799911737442, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22770, + "weighted_flow_matching_loss": 0.021425329241901638, + "weighted_rkd_loss": 0.01735210241749883 + }, + { + "epoch": 6.627873145184754, + "flow_matching_loss": 0.021091607213020325, + "grad_norm": 0.18170344829559326, + "learning_rate": 2.8424814584146752e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008043999830260874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003471414907835424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019559414684772493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.913726419210434, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22780, + "weighted_flow_matching_loss": 0.021091607213020325, + "weighted_rkd_loss": 0.017871990241110323 + }, + { + "epoch": 6.630782659295898, + "flow_matching_loss": 0.022297938540577888, + "grad_norm": 0.252345472574234, + "learning_rate": 2.8422970079646813e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007688679406419397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034775684005580844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018854927271604538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9136528968811035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22790, + "weighted_flow_matching_loss": 0.022297938540577888, + "weighted_rkd_loss": 0.01722684781998396 + }, + { + "epoch": 6.633692173407041, + "flow_matching_loss": 0.02204355299472809, + "grad_norm": 0.2770514190196991, + "learning_rate": 2.842112455576088e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007423087116330862, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003122021583840251, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01796819604933262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9135793507099151, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22800, + "weighted_flow_matching_loss": 0.02204355299472809, + "weighted_rkd_loss": 0.016415394283831118 + }, + { + "epoch": 6.636601687518184, + "flow_matching_loss": 0.021403672732412816, + "grad_norm": 0.2258807122707367, + "learning_rate": 2.8419278012629106e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008476096251979471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038645438849925994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02081673629581928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9135057747364044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22810, + "weighted_flow_matching_loss": 0.021403672732412816, + "weighted_rkd_loss": 0.0190162131562829 + }, + { + "epoch": 6.639511201629328, + "flow_matching_loss": 0.02135786972939968, + "grad_norm": 0.215902179479599, + "learning_rate": 2.8417430450391733e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007720486260950565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003390827355906367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018831799738109112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9134321808815002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22820, + "weighted_flow_matching_loss": 0.02135786972939968, + "weighted_rkd_loss": 0.017201575357466935 + }, + { + "epoch": 6.642420715740471, + "flow_matching_loss": 0.022882517985999586, + "grad_norm": 0.2642337381839752, + "learning_rate": 2.841558186918906e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008063837047666311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003483168734237552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019610843248665334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9133585333824158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22830, + "weighted_flow_matching_loss": 0.022882517985999586, + "weighted_rkd_loss": 0.017911753617227077 + }, + { + "epoch": 6.645330229851615, + "flow_matching_loss": 0.020418230444192886, + "grad_norm": 0.1951952427625656, + "learning_rate": 2.8413732269161476e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008030127361416816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003502974263392389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019563228730112313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9132848739624023, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22840, + "weighted_flow_matching_loss": 0.020418230444192886, + "weighted_rkd_loss": 0.017866838071495295 + }, + { + "epoch": 6.648239743962758, + "flow_matching_loss": 0.022420917823910713, + "grad_norm": 0.2606472969055176, + "learning_rate": 2.8411881650449458e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008345727529376745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038189529674127697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020510408282279968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.913211190700531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22850, + "weighted_flow_matching_loss": 0.022420917823910713, + "weighted_rkd_loss": 0.018730290699750185 + }, + { + "epoch": 6.651149258073902, + "flow_matching_loss": 0.02138913981616497, + "grad_norm": 0.2334287166595459, + "learning_rate": 2.8410030013193534e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077323027420789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034703986952081324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018935004342347385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9131374657154083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22860, + "weighted_flow_matching_loss": 0.02138913981616497, + "weighted_rkd_loss": 0.01729023242369294 + }, + { + "epoch": 6.654058772185045, + "flow_matching_loss": 0.020760687068104744, + "grad_norm": 0.2337893545627594, + "learning_rate": 2.840817735753433e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008333742851391435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036154377041384576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020282923243939877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9130637168884277, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22870, + "weighted_flow_matching_loss": 0.020760687068104744, + "weighted_rkd_loss": 0.018519589956849814 + }, + { + "epoch": 6.6569682862961885, + "flow_matching_loss": 0.022889424115419388, + "grad_norm": 0.2583245038986206, + "learning_rate": 2.8406323683612535e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007229923829436302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032753715990111233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017735219094902277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9129899442195892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22880, + "weighted_flow_matching_loss": 0.022889424115419388, + "weighted_rkd_loss": 0.016192058008164166 + }, + { + "epoch": 6.659877800407332, + "flow_matching_loss": 0.020116050727665424, + "grad_norm": 0.2442665696144104, + "learning_rate": 2.8404468991568936e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076632781885564326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034554158570244907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01878197230398655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9129161477088928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22890, + "weighted_flow_matching_loss": 0.020116050727665424, + "weighted_rkd_loss": 0.017146378755569458 + }, + { + "epoch": 6.662787314518475, + "flow_matching_loss": 0.02059882301837206, + "grad_norm": 0.19019295275211334, + "learning_rate": 2.840261328154438e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00696368319913745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002937627327628434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016864993795752525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9128423154354095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22900, + "weighted_flow_matching_loss": 0.02059882301837206, + "weighted_rkd_loss": 0.015395082905888558 + }, + { + "epoch": 6.665696828629619, + "flow_matching_loss": 0.021589356288313866, + "grad_norm": 0.24421566724777222, + "learning_rate": 2.8400756553679793e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00750379990786314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003448223741725087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01845582341775298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9127684593200683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22910, + "weighted_flow_matching_loss": 0.021589356288313866, + "weighted_rkd_loss": 0.01684589358046651 + }, + { + "epoch": 6.6686063427407625, + "flow_matching_loss": 0.021710217744112015, + "grad_norm": 0.2071070373058319, + "learning_rate": 2.8398898808116186e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007473173551261425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003402006742544472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018348353914916515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9126945734024048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22920, + "weighted_flow_matching_loss": 0.021710217744112015, + "weighted_rkd_loss": 0.016746440343558787 + }, + { + "epoch": 6.671515856851906, + "flow_matching_loss": 0.022578272223472595, + "grad_norm": 0.2508620023727417, + "learning_rate": 2.8397040044994636e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077688239980489016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034314480144530535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018969096057116986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9126206636428833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22930, + "weighted_flow_matching_loss": 0.022578272223472595, + "weighted_rkd_loss": 0.017311579920351504 + }, + { + "epoch": 6.674425370963049, + "flow_matching_loss": 0.021513116918504237, + "grad_norm": 0.23230069875717163, + "learning_rate": 2.8395180264456314e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075502013321965935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033006675541400908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01840107049793005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9125467002391815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22940, + "weighted_flow_matching_loss": 0.021513116918504237, + "weighted_rkd_loss": 0.016791837848722936 + }, + { + "epoch": 6.677334885074193, + "flow_matching_loss": 0.01989717874675989, + "grad_norm": 0.27621421217918396, + "learning_rate": 2.8393319466642445e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007501171063631773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033174962271004917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01831983858719468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9124727368354797, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22950, + "weighted_flow_matching_loss": 0.01989717874675989, + "weighted_rkd_loss": 0.016716347821056843 + }, + { + "epoch": 6.680244399185336, + "flow_matching_loss": 0.022251459397375585, + "grad_norm": 0.2058141827583313, + "learning_rate": 2.8391457651694352e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007010209001600742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003133256523869932, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017153674364089967, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9123987317085266, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22960, + "weighted_flow_matching_loss": 0.022251459397375585, + "weighted_rkd_loss": 0.015651020221412182 + }, + { + "epoch": 6.683153913296479, + "flow_matching_loss": 0.02100364016368985, + "grad_norm": 0.20496410131454468, + "learning_rate": 2.8389594819753425e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007745306193828583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00343768997117877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01892830217257142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.91232470870018, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22970, + "weighted_flow_matching_loss": 0.02100364016368985, + "weighted_rkd_loss": 0.017268741223961114 + }, + { + "epoch": 6.686063427407623, + "flow_matching_loss": 0.021712120808660983, + "grad_norm": 0.23546752333641052, + "learning_rate": 2.838773097096114e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007747221039608121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034938319586217403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018988274224102496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9122506439685821, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22980, + "weighted_flow_matching_loss": 0.021712120808660983, + "weighted_rkd_loss": 0.017322046961635352 + }, + { + "epoch": 6.688972941518767, + "flow_matching_loss": 0.022306391969323158, + "grad_norm": 0.2010953277349472, + "learning_rate": 2.838586610545903e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007326271198689937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034373659640550614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018089908454567195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9121765613555908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22990, + "weighted_flow_matching_loss": 0.022306391969323158, + "weighted_rkd_loss": 0.01650116601958871 + }, + { + "epoch": 6.6918824556299095, + "flow_matching_loss": 0.02263072896748781, + "grad_norm": 0.20344896614551544, + "learning_rate": 2.8384000223388736e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008510221494361758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038768861442804335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020897329039871692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9121024429798126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23000, + "weighted_flow_matching_loss": 0.02263072896748781, + "weighted_rkd_loss": 0.019060556031763553 + }, + { + "epoch": 6.694791969741053, + "flow_matching_loss": 0.019980586599558592, + "grad_norm": 0.21046507358551025, + "learning_rate": 2.8382133324891953e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008049298683181405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036207302939146755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019719327241182326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9120283007621766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23010, + "weighted_flow_matching_loss": 0.019980586599558592, + "weighted_rkd_loss": 0.01798464683815837 + }, + { + "epoch": 6.697701483852197, + "flow_matching_loss": 0.020437567681074142, + "grad_norm": 0.18291422724723816, + "learning_rate": 2.8380265410110454e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007638476137071848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034657247830182313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018742677103728057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9119541347026825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23020, + "weighted_flow_matching_loss": 0.020437567681074142, + "weighted_rkd_loss": 0.017092491779476405 + }, + { + "epoch": 6.70061099796334, + "flow_matching_loss": 0.021082289703190326, + "grad_norm": 0.1966305375099182, + "learning_rate": 2.83783964791861e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007821098808199168, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036553991958498955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019297596998512745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9118799269199371, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23030, + "weighted_flow_matching_loss": 0.021082289703190326, + "weighted_rkd_loss": 0.017597108148038387 + }, + { + "epoch": 6.703520512074483, + "flow_matching_loss": 0.020664289221167566, + "grad_norm": 0.2182522565126419, + "learning_rate": 2.8376526532260823e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007469795038923621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034029845148324965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018342574685811998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9118057131767273, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23040, + "weighted_flow_matching_loss": 0.020664289221167566, + "weighted_rkd_loss": 0.016724858805537222 + }, + { + "epoch": 6.706430026185627, + "flow_matching_loss": 0.021952751092612743, + "grad_norm": 0.21621394157409668, + "learning_rate": 2.837465556947663e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00768852666951716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032417636830359697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018618817254900933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9117314517498016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23050, + "weighted_flow_matching_loss": 0.021952751092612743, + "weighted_rkd_loss": 0.016975380573421718 + }, + { + "epoch": 6.709339540296771, + "flow_matching_loss": 0.02253379551693797, + "grad_norm": 0.20181602239608765, + "learning_rate": 2.837278359097562e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007566723367199301, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033409036928787827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018474350310862066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9116571724414826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23060, + "weighted_flow_matching_loss": 0.02253379551693797, + "weighted_rkd_loss": 0.01684228042140603 + }, + { + "epoch": 6.712249054407914, + "flow_matching_loss": 0.020521342009305953, + "grad_norm": 0.23564544320106506, + "learning_rate": 2.837091059689994e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007615649234503507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035657862899824977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01879708506166935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9115828573703766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23070, + "weighted_flow_matching_loss": 0.020521342009305953, + "weighted_rkd_loss": 0.017135077249258758 + }, + { + "epoch": 6.715158568519057, + "flow_matching_loss": 0.020474997349083424, + "grad_norm": 0.22374476492404938, + "learning_rate": 2.8369036587391847e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076147106941789385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035746406530961393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018804061971604825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9115085184574128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23080, + "weighted_flow_matching_loss": 0.020474997349083424, + "weighted_rkd_loss": 0.017140045296400785 + }, + { + "epoch": 6.718068082630201, + "flow_matching_loss": 0.021923987288028003, + "grad_norm": 0.21091575920581818, + "learning_rate": 2.836716156259365e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007459686091169715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003176728799007833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01809610091149807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9114341497421264, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23090, + "weighted_flow_matching_loss": 0.021923987288028003, + "weighted_rkd_loss": 0.01649338789284229 + }, + { + "epoch": 6.720977596741344, + "flow_matching_loss": 0.020274038426578045, + "grad_norm": 0.2186421900987625, + "learning_rate": 2.8365285522647745e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007289720280095935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032263729022815823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017805813811719418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9113597571849823, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23100, + "weighted_flow_matching_loss": 0.020274038426578045, + "weighted_rkd_loss": 0.016227502841502428 + }, + { + "epoch": 6.723887110852488, + "flow_matching_loss": 0.02168460190296173, + "grad_norm": 0.2386276125907898, + "learning_rate": 2.8363408467696613e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007916072150692344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037177793448790906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019549923483282328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9112853348255158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23110, + "weighted_flow_matching_loss": 0.02168460190296173, + "weighted_rkd_loss": 0.017815568298101426 + }, + { + "epoch": 6.726796624963631, + "flow_matching_loss": 0.021079255174845456, + "grad_norm": 0.17688585817813873, + "learning_rate": 2.83615303978828e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072312754113227126, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034432933665812016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01790584409609437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9112108647823334, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23120, + "weighted_flow_matching_loss": 0.021079255174845456, + "weighted_rkd_loss": 0.016316015645861624 + }, + { + "epoch": 6.729706139074775, + "flow_matching_loss": 0.020957848615944387, + "grad_norm": 0.20854993164539337, + "learning_rate": 2.8359651313348927e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00671548368409276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030222236877307297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01645319126546383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9111364185810089, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23130, + "weighted_flow_matching_loss": 0.020957848615944387, + "weighted_rkd_loss": 0.014991117082536221 + }, + { + "epoch": 6.732615653185918, + "flow_matching_loss": 0.02274949140846729, + "grad_norm": 0.23616239428520203, + "learning_rate": 2.8357771214237708e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008554672496393323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037460565567016602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020855401642620564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9110619008541108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23140, + "weighted_flow_matching_loss": 0.02274949140846729, + "weighted_rkd_loss": 0.019000530522316695 + }, + { + "epoch": 6.7355251672970615, + "flow_matching_loss": 0.019378129579126834, + "grad_norm": 0.18694961071014404, + "learning_rate": 2.8355890100691916e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007297923555597663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030797082232311367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0176755553111434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9109873473644257, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23150, + "weighted_flow_matching_loss": 0.019378129579126834, + "weighted_rkd_loss": 0.01610221406444907 + }, + { + "epoch": 6.738434681408204, + "flow_matching_loss": 0.021969511173665524, + "grad_norm": 0.203776016831398, + "learning_rate": 2.8354007972854405e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007715380005538464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003283629100769758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0187143892981112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9109127819538116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23160, + "weighted_flow_matching_loss": 0.021969511173665524, + "weighted_rkd_loss": 0.017047165799885987 + }, + { + "epoch": 6.741344195519348, + "flow_matching_loss": 0.02223795484751463, + "grad_norm": 0.2622496783733368, + "learning_rate": 2.8352124830868125e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008579142019152641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036711044376716018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020829388312995432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9108381927013397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23170, + "weighted_flow_matching_loss": 0.02223795484751463, + "weighted_rkd_loss": 0.018972185999155046 + }, + { + "epoch": 6.744253709630492, + "flow_matching_loss": 0.02166882362216711, + "grad_norm": 0.2641226351261139, + "learning_rate": 2.835024067487608e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008327675517648458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00364211262203753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020297463424503802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9107635736465454, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23180, + "weighted_flow_matching_loss": 0.02166882362216711, + "weighted_rkd_loss": 0.018486173078417777 + }, + { + "epoch": 6.7471632237416355, + "flow_matching_loss": 0.02088827546685934, + "grad_norm": 0.249041348695755, + "learning_rate": 2.834835550502136e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00789022222161293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035711822332814337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019351626560091973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9106889188289642, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23190, + "weighted_flow_matching_loss": 0.02088827546685934, + "weighted_rkd_loss": 0.017623356077820063 + }, + { + "epoch": 6.750072737852778, + "flow_matching_loss": 0.021684583462774753, + "grad_norm": 0.20095302164554596, + "learning_rate": 2.834646932144713e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007603385671973229, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00327135156840086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018478123005479574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9106142461299896, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23200, + "weighted_flow_matching_loss": 0.021684583462774753, + "weighted_rkd_loss": 0.01682647066190839 + }, + { + "epoch": 6.752982251963922, + "flow_matching_loss": 0.02581027839332819, + "grad_norm": 0.21708792448043823, + "learning_rate": 2.8344582124296634e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008134763361886144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036382277496159075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019907754193991423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9105395376682281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23210, + "weighted_flow_matching_loss": 0.02581027839332819, + "weighted_rkd_loss": 0.018126859050244092 + }, + { + "epoch": 6.755891766075066, + "flow_matching_loss": 0.019227315206080675, + "grad_norm": 0.21716155111789703, + "learning_rate": 2.8342693913713193e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007013200921937824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030782924965023994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01710469461977482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9104648053646087, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23220, + "weighted_flow_matching_loss": 0.019227315206080675, + "weighted_rkd_loss": 0.015573236625641584 + }, + { + "epoch": 6.7588012801862085, + "flow_matching_loss": 0.02026959666982293, + "grad_norm": 0.2154437154531479, + "learning_rate": 2.8340804689840203e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007832160592079163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033628295874223112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019027150794863702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.910390043258667, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23230, + "weighted_flow_matching_loss": 0.02026959666982293, + "weighted_rkd_loss": 0.017322123143821956 + }, + { + "epoch": 6.761710794297352, + "flow_matching_loss": 0.022504156827926634, + "grad_norm": 0.2274179756641388, + "learning_rate": 2.8338914452821136e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006920247804373503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003101125219836831, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016941620875149964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9103152453899384, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23240, + "weighted_flow_matching_loss": 0.022504156827926634, + "weighted_rkd_loss": 0.015422211308032275 + }, + { + "epoch": 6.764620308408496, + "flow_matching_loss": 0.020903610810637473, + "grad_norm": 0.21359211206436157, + "learning_rate": 2.8337023202799545e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007314313296228647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033145656809210776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01794319236651063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9102404296398163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23250, + "weighted_flow_matching_loss": 0.020903610810637473, + "weighted_rkd_loss": 0.016332626808434725 + }, + { + "epoch": 6.76752982251964, + "flow_matching_loss": 0.020005340687930584, + "grad_norm": 0.2232983112335205, + "learning_rate": 2.8335130939919066e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00763285243883729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034665154293179513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018732220120728015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9101655840873718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23260, + "weighted_flow_matching_loss": 0.020005340687930584, + "weighted_rkd_loss": 0.017049447260797025 + }, + { + "epoch": 6.7704393366307825, + "flow_matching_loss": 0.02132217213511467, + "grad_norm": 0.22005555033683777, + "learning_rate": 2.8333237664323386e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007780639920383692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036190805491060018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01918036015704274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9100907146930695, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23270, + "weighted_flow_matching_loss": 0.02132217213511467, + "weighted_rkd_loss": 0.017455904744565487 + }, + { + "epoch": 6.773348850741926, + "flow_matching_loss": 0.020974916219711304, + "grad_norm": 0.23380140960216522, + "learning_rate": 2.8331343376156305e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007898292550817133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038237783126533033, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019620363134890795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9100158095359803, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23280, + "weighted_flow_matching_loss": 0.020974916219711304, + "weighted_rkd_loss": 0.017854820378124713 + }, + { + "epoch": 6.77625836485307, + "flow_matching_loss": 0.02145763374865055, + "grad_norm": 0.1806805580854416, + "learning_rate": 2.8329448075561673e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008041926426813006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036214940482750535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019705347158014773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.909940880537033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23290, + "weighted_flow_matching_loss": 0.02145763374865055, + "weighted_rkd_loss": 0.0179306891746819 + }, + { + "epoch": 6.779167878964213, + "flow_matching_loss": 0.02226091418415308, + "grad_norm": 0.18804416060447693, + "learning_rate": 2.832755176268342e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008215813664719462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038128030952066182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020244430657476185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9098659217357635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23300, + "weighted_flow_matching_loss": 0.02226091418415308, + "weighted_rkd_loss": 0.018419739417731763 + }, + { + "epoch": 6.782077393075356, + "flow_matching_loss": 0.02295377813279629, + "grad_norm": 0.24224890768527985, + "learning_rate": 2.8325654437665573e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007720003509894014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035183243220672012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01895833145827055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9097909331321716, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23310, + "weighted_flow_matching_loss": 0.02295377813279629, + "weighted_rkd_loss": 0.017248126026242972 + }, + { + "epoch": 6.7849869071865, + "flow_matching_loss": 0.020848861522972585, + "grad_norm": 0.26846060156822205, + "learning_rate": 2.8323756100652208e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008621308533474804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003647753573022783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02089037075638771, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9097159206867218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23320, + "weighted_flow_matching_loss": 0.020848861522972585, + "weighted_rkd_loss": 0.019004266522824765 + }, + { + "epoch": 6.787896421297643, + "flow_matching_loss": 0.020447826012969017, + "grad_norm": 0.2060028463602066, + "learning_rate": 2.8321856751787498e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00688683083280921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003012331994250417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016785993799567224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9096408724784851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23330, + "weighted_flow_matching_loss": 0.020447826012969017, + "weighted_rkd_loss": 0.015269231610000133 + }, + { + "epoch": 6.790805935408787, + "flow_matching_loss": 0.020825378131121397, + "grad_norm": 0.28372979164123535, + "learning_rate": 2.831995639121568e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007035721326246858, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031036987085826693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017175141349434854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.909565806388855, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23340, + "weighted_flow_matching_loss": 0.020825378131121397, + "weighted_rkd_loss": 0.015621906239539384 + }, + { + "epoch": 6.79371544951993, + "flow_matching_loss": 0.0208162859082222, + "grad_norm": 0.1765723079442978, + "learning_rate": 2.8318055019081083e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006527968728914857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002726200711913407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015782138239592312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.909490704536438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23350, + "weighted_flow_matching_loss": 0.0208162859082222, + "weighted_rkd_loss": 0.014353710785508155 + }, + { + "epoch": 6.796624963631073, + "flow_matching_loss": 0.022363973036408424, + "grad_norm": 0.2511344254016876, + "learning_rate": 2.8316152635528102e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007120467210188508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029256788548082113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0171666132286191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9094155788421631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23360, + "weighted_flow_matching_loss": 0.022363973036408424, + "weighted_rkd_loss": 0.015611563809216022 + }, + { + "epoch": 6.799534477742217, + "flow_matching_loss": 0.02142266891896725, + "grad_norm": 0.23738166689872742, + "learning_rate": 2.83142492407012e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00803535752929747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036056597717106344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019676375016570092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9093404173851013, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23370, + "weighted_flow_matching_loss": 0.02142266891896725, + "weighted_rkd_loss": 0.017892593052238226 + }, + { + "epoch": 6.802443991853361, + "flow_matching_loss": 0.021929504908621313, + "grad_norm": 0.22037114202976227, + "learning_rate": 2.831234483474494e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007324971025809646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031219662167131903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01777190826833248, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9092652380466462, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23380, + "weighted_flow_matching_loss": 0.021929504908621313, + "weighted_rkd_loss": 0.01615934483706951 + }, + { + "epoch": 6.805353505964504, + "flow_matching_loss": 0.020408251881599428, + "grad_norm": 0.27371227741241455, + "learning_rate": 2.831043941780394e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007260003359988332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003127523954026401, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017647530883550644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9091900229454041, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23390, + "weighted_flow_matching_loss": 0.020408251881599428, + "weighted_rkd_loss": 0.01604494210332632 + }, + { + "epoch": 6.808263020075647, + "flow_matching_loss": 0.020652754139155148, + "grad_norm": 0.18939121067523956, + "learning_rate": 2.830853299002291e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007052574772387743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030528500210493803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017157999612390996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9091147780418396, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23400, + "weighted_flow_matching_loss": 0.020652754139155148, + "weighted_rkd_loss": 0.015598598308861256 + }, + { + "epoch": 6.811172534186791, + "flow_matching_loss": 0.02108522653579712, + "grad_norm": 0.2022375911474228, + "learning_rate": 2.8306625551546628e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008179566869512201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037947615375742315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02015389557927847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9090395152568818, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23410, + "weighted_flow_matching_loss": 0.02108522653579712, + "weighted_rkd_loss": 0.0183206788264215 + }, + { + "epoch": 6.8140820482979345, + "flow_matching_loss": 0.0211774755269289, + "grad_norm": 0.20431746542453766, + "learning_rate": 2.830471710251995e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007781770639121532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036385855404660107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019202126935124398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9089642226696014, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23420, + "weighted_flow_matching_loss": 0.0211774755269289, + "weighted_rkd_loss": 0.017453996930271387 + }, + { + "epoch": 6.816991562409077, + "flow_matching_loss": 0.024206540547311306, + "grad_norm": 0.2334764003753662, + "learning_rate": 2.830280764308781e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007333896262571216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003561922744847834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0182297152467072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9088889002799988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23430, + "weighted_flow_matching_loss": 0.024206540547311306, + "weighted_rkd_loss": 0.016568749956786633 + }, + { + "epoch": 6.819901076520221, + "flow_matching_loss": 0.021160761080682277, + "grad_norm": 0.21713542938232422, + "learning_rate": 2.8300897173395224e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007251634076237678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002972866781055927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017476134933531284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9088135421276092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23440, + "weighted_flow_matching_loss": 0.021160761080682277, + "weighted_rkd_loss": 0.01588255250826478 + }, + { + "epoch": 6.822810590631365, + "flow_matching_loss": 0.021511817723512648, + "grad_norm": 0.21472221612930298, + "learning_rate": 2.8298985693587274e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00685896840877831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002878708788193762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016596645768731832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9087381660938263, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23450, + "weighted_flow_matching_loss": 0.021511817723512648, + "weighted_rkd_loss": 0.015082021057605744 + }, + { + "epoch": 6.8257201047425085, + "flow_matching_loss": 0.02055668868124485, + "grad_norm": 0.21143940091133118, + "learning_rate": 2.8297073203809127e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006922312872484326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003082591271959245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01692721722647548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.908662760257721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23460, + "weighted_flow_matching_loss": 0.02055668868124485, + "weighted_rkd_loss": 0.015381124895066023 + }, + { + "epoch": 6.828629618853651, + "flow_matching_loss": 0.023089155927300452, + "grad_norm": 0.1799047887325287, + "learning_rate": 2.8295159704206022e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007125989068299532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003161510406062007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0174134885892272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9085873186588287, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23470, + "weighted_flow_matching_loss": 0.023089155927300452, + "weighted_rkd_loss": 0.015821719728410243 + }, + { + "epoch": 6.831539132964795, + "flow_matching_loss": 0.021144545823335647, + "grad_norm": 0.24326488375663757, + "learning_rate": 2.829324519492328e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007561168167740107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034132553962990643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018535591661930084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9085118532180786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23480, + "weighted_flow_matching_loss": 0.021144545823335647, + "weighted_rkd_loss": 0.016839771810919047 + }, + { + "epoch": 6.834448647075939, + "flow_matching_loss": 0.02001581098884344, + "grad_norm": 0.24855667352676392, + "learning_rate": 2.829132967610629e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007628987915813923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034284661756828426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018686441890895366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9084363579750061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23490, + "weighted_flow_matching_loss": 0.02001581098884344, + "weighted_rkd_loss": 0.01697541996836662 + }, + { + "epoch": 6.8373581611870815, + "flow_matching_loss": 0.020650014001876115, + "grad_norm": 0.22501443326473236, + "learning_rate": 2.8289413147900527e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007177054416388273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003251048410311341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0176051571033895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9083608448505401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23500, + "weighted_flow_matching_loss": 0.020650014001876115, + "weighted_rkd_loss": 0.015991799440234898 + }, + { + "epoch": 6.840267675298225, + "flow_matching_loss": 0.022761013358831406, + "grad_norm": 0.2288687378168106, + "learning_rate": 2.8287495610451542e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007171569112688303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030721368035301565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01741527523845434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9082852959632873, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23510, + "weighted_flow_matching_loss": 0.022761013358831406, + "weighted_rkd_loss": 0.015818051993846893 + }, + { + "epoch": 6.843177189409369, + "flow_matching_loss": 0.01955186706036329, + "grad_norm": 0.2356167733669281, + "learning_rate": 2.828557706390495e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008116693887859582, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003525402792729437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019758790358901025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9082097232341766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23520, + "weighted_flow_matching_loss": 0.01955186706036329, + "weighted_rkd_loss": 0.0179450960829854 + }, + { + "epoch": 6.846086703520512, + "flow_matching_loss": 0.02264052089303732, + "grad_norm": 0.20441046357154846, + "learning_rate": 2.8283657508406462e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008073136024177075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036052796756848695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01975155184045434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.908134114742279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23530, + "weighted_flow_matching_loss": 0.02264052089303732, + "weighted_rkd_loss": 0.017937019374221562 + }, + { + "epoch": 6.8489962176316554, + "flow_matching_loss": 0.020909207314252852, + "grad_norm": 0.24344438314437866, + "learning_rate": 2.8281736944101845e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007918962324038149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003578350064344704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019416274689137935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9080584883689881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23540, + "weighted_flow_matching_loss": 0.020909207314252852, + "weighted_rkd_loss": 0.01763111362233758 + }, + { + "epoch": 6.851905731742799, + "flow_matching_loss": 0.019886724278330804, + "grad_norm": 0.20159585773944855, + "learning_rate": 2.8279815371136967e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007431288762018084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003443903778679669, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018306481186300516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9079828262329102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23550, + "weighted_flow_matching_loss": 0.019886724278330804, + "weighted_rkd_loss": 0.016621982958167792 + }, + { + "epoch": 6.854815245853942, + "flow_matching_loss": 0.022343512624502182, + "grad_norm": 0.21749630570411682, + "learning_rate": 2.8277892789657745e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007235281262546777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032244920963421463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017695054691284896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9079071581363678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23560, + "weighted_flow_matching_loss": 0.022343512624502182, + "weighted_rkd_loss": 0.016065472550690175 + }, + { + "epoch": 6.857724759965086, + "flow_matching_loss": 0.020329652167856694, + "grad_norm": 0.21835607290267944, + "learning_rate": 2.8275969199810194e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007345583569258451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003072045906446874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017763213068246842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9078314125537872, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23570, + "weighted_flow_matching_loss": 0.020329652167856694, + "weighted_rkd_loss": 0.016126009356230497 + }, + { + "epoch": 6.860634274076229, + "flow_matching_loss": 0.019969388749450444, + "grad_norm": 0.23735246062278748, + "learning_rate": 2.8274044601740393e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007221005903556943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032459798501804473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017687991727143526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9077556788921356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23580, + "weighted_flow_matching_loss": 0.019969388749450444, + "weighted_rkd_loss": 0.01605636803433299 + }, + { + "epoch": 6.863543788187373, + "flow_matching_loss": 0.021266152150928974, + "grad_norm": 0.23948131501674652, + "learning_rate": 2.827211899559451e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008527377899736167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037425915244966744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02079734727740288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9076799094676972, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23590, + "weighted_flow_matching_loss": 0.021266152150928974, + "weighted_rkd_loss": 0.018877348583191633 + }, + { + "epoch": 6.866453302298516, + "flow_matching_loss": 0.021250553242862225, + "grad_norm": 0.2985173165798187, + "learning_rate": 2.8270192381518775e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007849649619311094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033365101320669056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019035809393972158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9076041042804718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23600, + "weighted_flow_matching_loss": 0.021250553242862225, + "weighted_rkd_loss": 0.017276953253895045 + }, + { + "epoch": 6.86936281640966, + "flow_matching_loss": 0.022283066622912885, + "grad_norm": 0.20508494973182678, + "learning_rate": 2.82682647596595e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008187714079394937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035938021959736944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019969230610877277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.907528281211853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23610, + "weighted_flow_matching_loss": 0.022283066622912885, + "weighted_rkd_loss": 0.018122663255780935 + }, + { + "epoch": 6.872272330520803, + "flow_matching_loss": 0.020217105560004712, + "grad_norm": 0.20240817964076996, + "learning_rate": 2.8266336130163085e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007143548969179392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003276969655416906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01756406743079424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9074524283409119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23620, + "weighted_flow_matching_loss": 0.020217105560004712, + "weighted_rkd_loss": 0.015938549116253854 + }, + { + "epoch": 6.875181844631946, + "flow_matching_loss": 0.021496175974607467, + "grad_norm": 0.2996926009654999, + "learning_rate": 2.8264406493175992e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008072675298899412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035782066406682134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01972355730831623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9073765456676484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23630, + "weighted_flow_matching_loss": 0.021496175974607467, + "weighted_rkd_loss": 0.017896640580147505 + }, + { + "epoch": 6.87809135874309, + "flow_matching_loss": 0.02022344749420881, + "grad_norm": 0.19968198239803314, + "learning_rate": 2.826247584884476e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006618696870282293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028748176991939545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016112211253494026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9073006272315979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23640, + "weighted_flow_matching_loss": 0.02022344749420881, + "weighted_rkd_loss": 0.014618605468422174 + }, + { + "epoch": 6.881000872854234, + "flow_matching_loss": 0.022548486664891242, + "grad_norm": 0.2357611358165741, + "learning_rate": 2.8260544197316012e-05, + "loss": 0.0396, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007779393717646599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00322381928563118, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018782606907188894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.907224690914154, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23650, + "weighted_flow_matching_loss": 0.022548486664891242, + "weighted_rkd_loss": 0.017040044255554675 + }, + { + "epoch": 6.883910386965377, + "flow_matching_loss": 0.021176850982010366, + "grad_norm": 0.2364053577184677, + "learning_rate": 2.8258611538736445e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007715317839756608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032896415097638963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01872027711942792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9071487188339233, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23660, + "weighted_flow_matching_loss": 0.021176850982010366, + "weighted_rkd_loss": 0.016982078924775122 + }, + { + "epoch": 6.88681990107652, + "flow_matching_loss": 0.020689934492111206, + "grad_norm": 0.2600575089454651, + "learning_rate": 2.8256677873252832e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008741327235475183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038073732750490306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02129002772271633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9070727229118347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23670, + "weighted_flow_matching_loss": 0.020689934492111206, + "weighted_rkd_loss": 0.019311549421399832 + }, + { + "epoch": 6.889729415187664, + "flow_matching_loss": 0.021768155880272387, + "grad_norm": 0.2504083216190338, + "learning_rate": 2.825474320101202e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007540990877896547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031815855065360664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018263566959649324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9069967091083526, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23680, + "weighted_flow_matching_loss": 0.021768155880272387, + "weighted_rkd_loss": 0.016564995422959326 + }, + { + "epoch": 6.892638929298807, + "flow_matching_loss": 0.021785771660506725, + "grad_norm": 0.24407532811164856, + "learning_rate": 2.825280752216094e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007598462933674455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003310574358329177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018507500272244216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9069206535816192, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23690, + "weighted_flow_matching_loss": 0.021785771660506725, + "weighted_rkd_loss": 0.016784786060452462 + }, + { + "epoch": 6.89554844340995, + "flow_matching_loss": 0.02137018907815218, + "grad_norm": 0.23732729256153107, + "learning_rate": 2.8250870836846582e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007860139757394791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034192368388175965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01913951653987169, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.906844574213028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23700, + "weighted_flow_matching_loss": 0.02137018907815218, + "weighted_rkd_loss": 0.017356562428176403 + }, + { + "epoch": 6.898457957521094, + "flow_matching_loss": 0.020250912010669708, + "grad_norm": 0.22833983600139618, + "learning_rate": 2.8248933145216037e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007976653426885605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036753969034180045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019628703501075505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9067684650421143, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23710, + "weighted_flow_matching_loss": 0.020250912010669708, + "weighted_rkd_loss": 0.017798664327710868 + }, + { + "epoch": 6.901367471632238, + "flow_matching_loss": 0.02186724469065666, + "grad_norm": 0.18414026498794556, + "learning_rate": 2.8246994447416453e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007424622867256403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034450638107955457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018294309545308352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9066923379898071, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23720, + "weighted_flow_matching_loss": 0.02186724469065666, + "weighted_rkd_loss": 0.016587320901453494 + }, + { + "epoch": 6.904276985743381, + "flow_matching_loss": 0.02089250534772873, + "grad_norm": 0.17474505305290222, + "learning_rate": 2.8245054743595065e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007707196101546287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003329273429699242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01874366533011198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9066161692142487, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23730, + "weighted_flow_matching_loss": 0.02089250534772873, + "weighted_rkd_loss": 0.016993334516882896 + }, + { + "epoch": 6.907186499854524, + "flow_matching_loss": 0.022852745186537505, + "grad_norm": 0.2773350179195404, + "learning_rate": 2.824311403389918e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00810010633431375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003615837171673775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01981604965403676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9065399825572967, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23740, + "weighted_flow_matching_loss": 0.022852745186537505, + "weighted_rkd_loss": 0.017964007332921028 + }, + { + "epoch": 6.910096013965668, + "flow_matching_loss": 0.022278380393981934, + "grad_norm": 0.20328208804130554, + "learning_rate": 2.8241172318476175e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007019389141350985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032983002252876758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017337078507989646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9064637660980225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23750, + "weighted_flow_matching_loss": 0.022278380393981934, + "weighted_rkd_loss": 0.015715439338237047 + }, + { + "epoch": 6.913005528076811, + "flow_matching_loss": 0.022947178967297078, + "grad_norm": 0.21905213594436646, + "learning_rate": 2.8239229597473522e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073740746360272166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003274132404476404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018022281769663095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9063875079154968, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23760, + "weighted_flow_matching_loss": 0.022947178967297078, + "weighted_rkd_loss": 0.016335221007466316 + }, + { + "epoch": 6.9159150421879545, + "flow_matching_loss": 0.021231544949114323, + "grad_norm": 0.2378583699464798, + "learning_rate": 2.823728587103875e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00796731342561543, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003590452531352639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019525079242885112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9063112556934356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23770, + "weighted_flow_matching_loss": 0.021231544949114323, + "weighted_rkd_loss": 0.01769578205421567 + }, + { + "epoch": 6.918824556299098, + "flow_matching_loss": 0.0221007589250803, + "grad_norm": 0.20483644306659698, + "learning_rate": 2.8235341139319474e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006998572964221239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030704484786838295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017067594267427922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9062349259853363, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23780, + "weighted_flow_matching_loss": 0.0221007589250803, + "weighted_rkd_loss": 0.015467230696231127 + }, + { + "epoch": 6.921734070410242, + "flow_matching_loss": 0.022847133688628674, + "grad_norm": 0.20638388395309448, + "learning_rate": 2.8233395402463388e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008137823780998588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003724908805452287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02000055620446801, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9061586141586304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23790, + "weighted_flow_matching_loss": 0.022847133688628674, + "weighted_rkd_loss": 0.018123627919703723 + }, + { + "epoch": 6.924643584521385, + "flow_matching_loss": 0.021078987792134286, + "grad_norm": 0.21943089365959167, + "learning_rate": 2.8231448660618255e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008014923008158804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003578990860842168, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019608836621046066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9060822546482086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23800, + "weighted_flow_matching_loss": 0.021078987792134286, + "weighted_rkd_loss": 0.01776722390204668 + }, + { + "epoch": 6.927553098632528, + "flow_matching_loss": 0.020864523760974408, + "grad_norm": 0.17141523957252502, + "learning_rate": 2.822950091393191e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00729639558121562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035481457831338047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018140936829149724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9060058712959289, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23810, + "weighted_flow_matching_loss": 0.020864523760974408, + "weighted_rkd_loss": 0.016435806266963483 + }, + { + "epoch": 6.930462612743672, + "flow_matching_loss": 0.020839406922459603, + "grad_norm": 0.20706933736801147, + "learning_rate": 2.822755216255228e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007132767513394356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031063089380040767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017371843848377467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9059294581413269, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23820, + "weighted_flow_matching_loss": 0.020839406922459603, + "weighted_rkd_loss": 0.015737677831202744 + }, + { + "epoch": 6.933372126854815, + "flow_matching_loss": 0.020321487821638583, + "grad_norm": 0.22468721866607666, + "learning_rate": 2.8225602406627357e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007424460304901004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031907120021060107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018039632588624954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9058530151844024, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23830, + "weighted_flow_matching_loss": 0.020321487821638583, + "weighted_rkd_loss": 0.016341221146285535 + }, + { + "epoch": 6.936281640965959, + "flow_matching_loss": 0.023393142223358154, + "grad_norm": 0.2570067346096039, + "learning_rate": 2.8223651646305212e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008081074338406324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00377409269567579, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019936241395771503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9057765483856202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23840, + "weighted_flow_matching_loss": 0.023393142223358154, + "weighted_rkd_loss": 0.018057733494788408 + }, + { + "epoch": 6.939191155077102, + "flow_matching_loss": 0.019755831640213727, + "grad_norm": 0.21569687128067017, + "learning_rate": 2.8221699881734e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007430092757567763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032641524914652107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018124338146299122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9057000517845154, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23850, + "weighted_flow_matching_loss": 0.019755831640213727, + "weighted_rkd_loss": 0.01641521919518709 + }, + { + "epoch": 6.942100669188245, + "flow_matching_loss": 0.02240971811115742, + "grad_norm": 0.225098118185997, + "learning_rate": 2.8219747113061926e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007823726069182158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036517794942483306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01929923165589571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9056235313415527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23860, + "weighted_flow_matching_loss": 0.02240971811115742, + "weighted_rkd_loss": 0.017477835435420273 + }, + { + "epoch": 6.945010183299389, + "flow_matching_loss": 0.021290972270071505, + "grad_norm": 0.20673753321170807, + "learning_rate": 2.821779334043731e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008509226934984326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003783907415345311, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02080236133188009, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9055469751358032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23870, + "weighted_flow_matching_loss": 0.021290972270071505, + "weighted_rkd_loss": 0.01883752616122365 + }, + { + "epoch": 6.947919697410533, + "flow_matching_loss": 0.021620746701955795, + "grad_norm": 0.18744425475597382, + "learning_rate": 2.8215838564008515e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074530802201479675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032125358935445546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018118696194142102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9054704010486603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23880, + "weighted_flow_matching_loss": 0.021620746701955795, + "weighted_rkd_loss": 0.016405915655195714 + }, + { + "epoch": 6.950829211521675, + "flow_matching_loss": 0.020439601596444847, + "grad_norm": 0.2613697350025177, + "learning_rate": 2.8213882783924004e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007783715473487973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036704144440591335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019237844925373792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9053937911987304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23890, + "weighted_flow_matching_loss": 0.020439601596444847, + "weighted_rkd_loss": 0.017417792789638044 + }, + { + "epoch": 6.953738725632819, + "flow_matching_loss": 0.02175428345799446, + "grad_norm": 0.21708457171916962, + "learning_rate": 2.8211926000332296e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007477682922035456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033116305246949198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01826699636876583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9053171515464783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23900, + "weighted_flow_matching_loss": 0.02175428345799446, + "weighted_rkd_loss": 0.016537453141063453 + }, + { + "epoch": 6.956648239743963, + "flow_matching_loss": 0.02166223172098398, + "grad_norm": 0.21634729206562042, + "learning_rate": 2.8209968213382e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007133415946736932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033080881694331765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017574920412153007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9052404880523681, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23910, + "weighted_flow_matching_loss": 0.02166223172098398, + "weighted_rkd_loss": 0.01590951830148697 + }, + { + "epoch": 6.9595577538551066, + "flow_matching_loss": 0.022733346745371818, + "grad_norm": 0.21886910498142242, + "learning_rate": 2.8208009423221794e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007017135759815574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003178265620954335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017212537117302418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9051638007164001, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23920, + "weighted_flow_matching_loss": 0.022733346745371818, + "weighted_rkd_loss": 0.015580190997570752 + }, + { + "epoch": 6.962467267966249, + "flow_matching_loss": 0.021004882454872132, + "grad_norm": 0.20128773152828217, + "learning_rate": 2.820604963000044e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076784272212535145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032430438324809073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018599898274987935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9050870835781097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23930, + "weighted_flow_matching_loss": 0.021004882454872132, + "weighted_rkd_loss": 0.016834517009556293 + }, + { + "epoch": 6.965376782077393, + "flow_matching_loss": 0.021717994846403597, + "grad_norm": 0.1926347315311432, + "learning_rate": 2.820408883386677e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006776414066553116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003207876766100526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016760704945772885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9050103306770325, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23940, + "weighted_flow_matching_loss": 0.021717994846403597, + "weighted_rkd_loss": 0.015168627444654702 + }, + { + "epoch": 6.968286296188537, + "flow_matching_loss": 0.021112794056534767, + "grad_norm": 0.20357894897460938, + "learning_rate": 2.820212703496969e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007084160996600986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003072287933900952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017240609601140024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9049335598945618, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23950, + "weighted_flow_matching_loss": 0.021112794056534767, + "weighted_rkd_loss": 0.015601599682122469 + }, + { + "epoch": 6.97119581029968, + "flow_matching_loss": 0.020202620700001717, + "grad_norm": 0.2123517245054245, + "learning_rate": 2.8200164233458187e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007742526289075613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034797357395291327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018964788224548102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9048567533493042, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23960, + "weighted_flow_matching_loss": 0.020202620700001717, + "weighted_rkd_loss": 0.01716040847823024 + }, + { + "epoch": 6.974105324410823, + "flow_matching_loss": 0.02041014488786459, + "grad_norm": 0.18802566826343536, + "learning_rate": 2.8198200429481324e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007052559731528163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031025208067148925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017207640409469604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9047799289226532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23970, + "weighted_flow_matching_loss": 0.02041014488786459, + "weighted_rkd_loss": 0.01556915808469057 + }, + { + "epoch": 6.977014838521967, + "flow_matching_loss": 0.01982197817414999, + "grad_norm": 0.21670471131801605, + "learning_rate": 2.8196235623188243e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074086700566113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032406818820163606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01805802220478654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9047030508518219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23980, + "weighted_flow_matching_loss": 0.01982197817414999, + "weighted_rkd_loss": 0.016337157506495715 + }, + { + "epoch": 6.979924352633111, + "flow_matching_loss": 0.02047972995787859, + "grad_norm": 0.21647146344184875, + "learning_rate": 2.819426981472815e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073010779451578856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003156299842521548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017758455779403447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9046261608600616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23990, + "weighted_flow_matching_loss": 0.02047972995787859, + "weighted_rkd_loss": 0.01606476353481412 + }, + { + "epoch": 6.9828338667442535, + "flow_matching_loss": 0.0218766326084733, + "grad_norm": 0.2488260120153427, + "learning_rate": 2.8192303004250343e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007456713076680899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033398014958947895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018253227416425943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9045492708683014, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24000, + "weighted_flow_matching_loss": 0.0218766326084733, + "weighted_rkd_loss": 0.01651093438267708 + }, + { + "epoch": 6.985743380855397, + "flow_matching_loss": 0.02089545503258705, + "grad_norm": 0.23883427679538727, + "learning_rate": 2.819033519190418e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076195431873202326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033640835899859667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01860317001119256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9044723331928253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24010, + "weighted_flow_matching_loss": 0.02089545503258705, + "weighted_rkd_loss": 0.016826043743640183 + }, + { + "epoch": 6.988652894966541, + "flow_matching_loss": 0.020276957470923663, + "grad_norm": 0.24604478478431702, + "learning_rate": 2.8188366377839114e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007417389191687107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032759432680904864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0181107216514647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9043953597545624, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24020, + "weighted_flow_matching_loss": 0.020276957470923663, + "weighted_rkd_loss": 0.016379223391413688 + }, + { + "epoch": 6.991562409077684, + "flow_matching_loss": 0.021173732355237007, + "grad_norm": 0.26009371876716614, + "learning_rate": 2.818639656220465e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007682719361037016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003638364328071475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01900380300357938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.904318368434906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24030, + "weighted_flow_matching_loss": 0.021173732355237007, + "weighted_rkd_loss": 0.01718544727191329 + }, + { + "epoch": 6.9944719231888275, + "flow_matching_loss": 0.020261042192578314, + "grad_norm": 0.21605898439884186, + "learning_rate": 2.8184425745150394e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007557329721748829, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034697035793215037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01858436306938529, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9042413413524628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24040, + "weighted_flow_matching_loss": 0.020261042192578314, + "weighted_rkd_loss": 0.016804762929677964 + }, + { + "epoch": 6.997381437299971, + "flow_matching_loss": 0.021223923191428185, + "grad_norm": 0.3186807334423065, + "learning_rate": 2.818245392682601e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007600543461740017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034464150900021195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018647501617670058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9041642904281616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24050, + "weighted_flow_matching_loss": 0.021223923191428185, + "weighted_rkd_loss": 0.016860404051840305 + }, + { + "epoch": 7.000290951411114, + "flow_matching_loss": 0.022335355915129185, + "grad_norm": 0.24028192460536957, + "learning_rate": 2.8180481107381247e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007565748505294323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003207124024629593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018338621035218237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9040872037410737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24060, + "weighted_flow_matching_loss": 0.022335355915129185, + "weighted_rkd_loss": 0.016579727828502654 + }, + { + "epoch": 7.003200465522258, + "flow_matching_loss": 0.019941187743097544, + "grad_norm": 0.20657794177532196, + "learning_rate": 2.8178507286965934e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007775604445487261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003497314709238708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019048523530364035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9040101051330567, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24070, + "weighted_flow_matching_loss": 0.019941187743097544, + "weighted_rkd_loss": 0.0172200508415699 + }, + { + "epoch": 7.006109979633401, + "flow_matching_loss": 0.021096646413207056, + "grad_norm": 0.22506262362003326, + "learning_rate": 2.8176532465729958e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006911374302580953, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028941557509824635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01671690456569195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9039329707622528, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24080, + "weighted_flow_matching_loss": 0.021096646413207056, + "weighted_rkd_loss": 0.015110931359231472 + }, + { + "epoch": 7.009019493744544, + "flow_matching_loss": 0.021804046630859376, + "grad_norm": 0.24712297320365906, + "learning_rate": 2.8174556643823298e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077208579052239655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003605911391787231, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019047626852989198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9038558006286621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24090, + "weighted_flow_matching_loss": 0.021804046630859376, + "weighted_rkd_loss": 0.017216315772384406 + }, + { + "epoch": 7.011929007855688, + "flow_matching_loss": 0.0219918517395854, + "grad_norm": 0.21817001700401306, + "learning_rate": 2.8172579821396005e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006439488334581256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002798599819652736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01567757623270154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.903778612613678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24100, + "weighted_flow_matching_loss": 0.0219918517395854, + "weighted_rkd_loss": 0.014169069565832615 + }, + { + "epoch": 7.014838521966832, + "flow_matching_loss": 0.020053163170814514, + "grad_norm": 0.20058678090572357, + "learning_rate": 2.8170601998598208e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007529448485001922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035219866782426833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0185808839276433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9037014007568359, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24110, + "weighted_flow_matching_loss": 0.020053163170814514, + "weighted_rkd_loss": 0.016791598312556742 + }, + { + "epoch": 7.017748036077975, + "flow_matching_loss": 0.022266332432627677, + "grad_norm": 0.2492697536945343, + "learning_rate": 2.8168623175580107e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008939574379473925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004018567549064755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021897716354578732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.903624153137207, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24120, + "weighted_flow_matching_loss": 0.022266332432627677, + "weighted_rkd_loss": 0.019787298887968062 + }, + { + "epoch": 7.020657550189118, + "flow_matching_loss": 0.02030355725437403, + "grad_norm": 0.2255534827709198, + "learning_rate": 2.8166643352491978e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007070788834244013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030464255949482323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017188003472983836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9035468816757202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24130, + "weighted_flow_matching_loss": 0.02030355725437403, + "weighted_rkd_loss": 0.015530158672481776 + }, + { + "epoch": 7.023567064300262, + "flow_matching_loss": 0.01946822740137577, + "grad_norm": 0.17736341059207916, + "learning_rate": 2.8164662529484186e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007274088310077786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003159274603240192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01770745124667883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.903469580411911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24140, + "weighted_flow_matching_loss": 0.01946822740137577, + "weighted_rkd_loss": 0.015998170617967843 + }, + { + "epoch": 7.026476578411406, + "flow_matching_loss": 0.02083340771496296, + "grad_norm": 0.24611727893352509, + "learning_rate": 2.816268070670715e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008011284191161394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036880332510918377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01971060149371624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9033922553062439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24150, + "weighted_flow_matching_loss": 0.02083340771496296, + "weighted_rkd_loss": 0.01780637223273516 + }, + { + "epoch": 7.029386092522548, + "flow_matching_loss": 0.020970048569142818, + "grad_norm": 0.21082815527915955, + "learning_rate": 2.8160697884311377e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00744643728248775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031451119109988214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018037986382842063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9033148944377899, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24160, + "weighted_flow_matching_loss": 0.020970048569142818, + "weighted_rkd_loss": 0.01629404295235872 + }, + { + "epoch": 7.032295606633692, + "flow_matching_loss": 0.019887408893555402, + "grad_norm": 0.2270015925168991, + "learning_rate": 2.8158714062447453e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007805866841226816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003417019289918244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01902875304222107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9032375037670135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24170, + "weighted_flow_matching_loss": 0.019887408893555402, + "weighted_rkd_loss": 0.017187502048909665 + }, + { + "epoch": 7.035205120744836, + "flow_matching_loss": 0.020734699629247188, + "grad_norm": 0.37177491188049316, + "learning_rate": 2.8156729241266043e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007470181165263057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033273595850914716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018267721589654683, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9031601011753082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24180, + "weighted_flow_matching_loss": 0.020734699629247188, + "weighted_rkd_loss": 0.016498646792024375 + }, + { + "epoch": 7.038114634855979, + "flow_matching_loss": 0.021211291290819646, + "grad_norm": 0.24237172305583954, + "learning_rate": 2.815474342091787e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007676313864067197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003508630907163024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01886125896126032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9030826687812805, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24190, + "weighted_flow_matching_loss": 0.021211291290819646, + "weighted_rkd_loss": 0.017033281736075878 + }, + { + "epoch": 7.041024148967122, + "flow_matching_loss": 0.02112657343968749, + "grad_norm": 0.2248975932598114, + "learning_rate": 2.8152756601553745e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006957152299582959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029075711267068983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01682187560945749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9030051827430725, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24200, + "weighted_flow_matching_loss": 0.02112657343968749, + "weighted_rkd_loss": 0.015190241020172834 + }, + { + "epoch": 7.043933663078266, + "flow_matching_loss": 0.021819095965474843, + "grad_norm": 0.21457938849925995, + "learning_rate": 2.815076878332456e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007521518040448427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032375360373407603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018280571978539228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9029276967048645, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24210, + "weighted_flow_matching_loss": 0.021819095965474843, + "weighted_rkd_loss": 0.016506065987050532 + }, + { + "epoch": 7.04684317718941, + "flow_matching_loss": 0.021062207967042924, + "grad_norm": 0.2372448891401291, + "learning_rate": 2.814877996638127e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007361625321209431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033704173285514115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018093668017536403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9028501808643341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24220, + "weighted_flow_matching_loss": 0.021062207967042924, + "weighted_rkd_loss": 0.016335832234472037 + }, + { + "epoch": 7.049752691300553, + "flow_matching_loss": 0.021997169777750968, + "grad_norm": 0.23442342877388, + "learning_rate": 2.814679015087492e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007646325929090381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003362064994871616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01865471703931689, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9027726292610169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24230, + "weighted_flow_matching_loss": 0.021997169777750968, + "weighted_rkd_loss": 0.016840949095785616 + }, + { + "epoch": 7.052662205411696, + "flow_matching_loss": 0.02013904396444559, + "grad_norm": 0.21775689721107483, + "learning_rate": 2.8144799336956624e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069470305927097796, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029855911619961263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0168796525336802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9026950478553772, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24240, + "weighted_flow_matching_loss": 0.02013904396444559, + "weighted_rkd_loss": 0.015237142611294986 + }, + { + "epoch": 7.05557171952284, + "flow_matching_loss": 0.02327273301780224, + "grad_norm": 0.20138698816299438, + "learning_rate": 2.814280752477756e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069221124518662695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030820264015346765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01692625107243657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9026174485683441, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24250, + "weighted_flow_matching_loss": 0.02327273301780224, + "weighted_rkd_loss": 0.015277974773198366 + }, + { + "epoch": 7.058481233633983, + "flow_matching_loss": 0.02113536177203059, + "grad_norm": 0.21351341903209686, + "learning_rate": 2.8140814714489e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007015045592561364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003173923888243735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017204015143215658, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9025398135185242, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24260, + "weighted_flow_matching_loss": 0.02113536177203059, + "weighted_rkd_loss": 0.015527294483035803 + }, + { + "epoch": 7.0613907477451265, + "flow_matching_loss": 0.022236561682075263, + "grad_norm": 0.1970042586326599, + "learning_rate": 2.8138820906242282e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007396298786625266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031326658790931106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01792526338249445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9024621546268463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24270, + "weighted_flow_matching_loss": 0.022236561682075263, + "weighted_rkd_loss": 0.016176899895071983 + }, + { + "epoch": 7.06430026185627, + "flow_matching_loss": 0.023173228465020658, + "grad_norm": 0.25174185633659363, + "learning_rate": 2.8136826100188828e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007951138261705636, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035029141465201972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019405190460383893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9023844718933105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24280, + "weighted_flow_matching_loss": 0.023173228465020658, + "weighted_rkd_loss": 0.01751090632751584 + }, + { + "epoch": 7.067209775967413, + "flow_matching_loss": 0.02052636332809925, + "grad_norm": 0.2214640974998474, + "learning_rate": 2.8134830296480125e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007044418202713132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002947922796010971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017036759294569492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.902306753396988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24290, + "weighted_flow_matching_loss": 0.02052636332809925, + "weighted_rkd_loss": 0.015372385550290347 + }, + { + "epoch": 7.070119290078557, + "flow_matching_loss": 0.02126759383827448, + "grad_norm": 0.24612988531589508, + "learning_rate": 2.8132833495267743e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007386726606637239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031054385704919695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01787889199331403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9022290110588074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24300, + "weighted_flow_matching_loss": 0.02126759383827448, + "weighted_rkd_loss": 0.016130826156586408 + }, + { + "epoch": 7.0730288041897005, + "flow_matching_loss": 0.019914021622389554, + "grad_norm": 0.18675631284713745, + "learning_rate": 2.8130835696703333e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070486496668308975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031604138668626545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017257713340222837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.902151244878769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24310, + "weighted_flow_matching_loss": 0.019914021622389554, + "weighted_rkd_loss": 0.015569084044545888 + }, + { + "epoch": 7.075938318300844, + "flow_matching_loss": 0.021260196156799794, + "grad_norm": 0.2614906430244446, + "learning_rate": 2.8128836900938604e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007971500884741545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003641396528109908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01958439825102687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9020734429359436, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24320, + "weighted_flow_matching_loss": 0.021260196156799794, + "weighted_rkd_loss": 0.017666522692888976 + }, + { + "epoch": 7.078847832411987, + "flow_matching_loss": 0.021937869116663934, + "grad_norm": 0.21004977822303772, + "learning_rate": 2.8126837108125353e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007285265484824777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00319376103579998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01776429172605276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9019956171512604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24330, + "weighted_flow_matching_loss": 0.021937869116663934, + "weighted_rkd_loss": 0.016023269109427928 + }, + { + "epoch": 7.081757346523131, + "flow_matching_loss": 0.019993736129254102, + "grad_norm": 0.229444682598114, + "learning_rate": 2.8124836318415454e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006676395470276475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027669075410813094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016119698248803615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9019177675247192, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24340, + "weighted_flow_matching_loss": 0.019993736129254102, + "weighted_rkd_loss": 0.014538655057549477 + }, + { + "epoch": 7.084666860634274, + "flow_matching_loss": 0.021567124873399734, + "grad_norm": 0.19052571058273315, + "learning_rate": 2.812283453196085e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008163135359063744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035562160657718776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019882486760616304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9018398880958557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24350, + "weighted_flow_matching_loss": 0.021567124873399734, + "weighted_rkd_loss": 0.017930898070335387 + }, + { + "epoch": 7.087576374745417, + "flow_matching_loss": 0.02137765921652317, + "grad_norm": 0.23877298831939697, + "learning_rate": 2.812083174891357e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007043123943731189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029958635801449417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017082111444324255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9017619788646698, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24360, + "weighted_flow_matching_loss": 0.02137765921652317, + "weighted_rkd_loss": 0.015403989981859923 + }, + { + "epoch": 7.090485888856561, + "flow_matching_loss": 0.023998607136309146, + "grad_norm": 0.2450440227985382, + "learning_rate": 2.8118827969425708e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008141820318996906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00358774452470243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01987138520926237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9016840398311615, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24370, + "weighted_flow_matching_loss": 0.023998607136309146, + "weighted_rkd_loss": 0.017917671520262957 + }, + { + "epoch": 7.093395402967705, + "flow_matching_loss": 0.020619630441069604, + "grad_norm": 0.22406597435474396, + "learning_rate": 2.8116823193649443e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007107240660116076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031195626826956866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017334043700248002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9016060829162598, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24380, + "weighted_flow_matching_loss": 0.020619630441069604, + "weighted_rkd_loss": 0.015628503635525703 + }, + { + "epoch": 7.0963049170788475, + "flow_matching_loss": 0.02143444661051035, + "grad_norm": 0.2298535853624344, + "learning_rate": 2.8114817421737014e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007486369414255023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003359864163212478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01833260301500559, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9015280902385712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24390, + "weighted_flow_matching_loss": 0.02143444661051035, + "weighted_rkd_loss": 0.016527348011732102 + }, + { + "epoch": 7.099214431189991, + "flow_matching_loss": 0.020641167368739843, + "grad_norm": 0.19979839026927948, + "learning_rate": 2.8112810653840755e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00649454197846353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002844711230136454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015833795256912708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9014500677585602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24400, + "weighted_flow_matching_loss": 0.020641167368739843, + "weighted_rkd_loss": 0.01427337247878313 + }, + { + "epoch": 7.102123945301135, + "flow_matching_loss": 0.020539582520723344, + "grad_norm": 0.2283061146736145, + "learning_rate": 2.8110802890113064e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069730819668620825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030900222482159734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017036186158657075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9013720393180847, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24410, + "weighted_flow_matching_loss": 0.020539582520723344, + "weighted_rkd_loss": 0.015355942025780677 + }, + { + "epoch": 7.105033459412278, + "flow_matching_loss": 0.0203342504799366, + "grad_norm": 0.23191151022911072, + "learning_rate": 2.8108794130706414e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007000806462019682, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003193910070694983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01719552306458354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9012939631938934, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24420, + "weighted_flow_matching_loss": 0.0203342504799366, + "weighted_rkd_loss": 0.015498228557407855 + }, + { + "epoch": 7.107942973523421, + "flow_matching_loss": 0.019534853659570216, + "grad_norm": 0.25427669286727905, + "learning_rate": 2.8106784375773363e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007042894279584289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00323519017547369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017320978734642268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9012158572673797, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24430, + "weighted_flow_matching_loss": 0.019534853659570216, + "weighted_rkd_loss": 0.015609917137771845 + }, + { + "epoch": 7.110852487634565, + "flow_matching_loss": 0.02089574784040451, + "grad_norm": 0.1863323599100113, + "learning_rate": 2.810477362546653e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00791914751753211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034511456033214927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019289440754801036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9011377215385437, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24440, + "weighted_flow_matching_loss": 0.02089574784040451, + "weighted_rkd_loss": 0.017382460832595825 + }, + { + "epoch": 7.113762001745709, + "flow_matching_loss": 0.020992217026650904, + "grad_norm": 0.23744289577007294, + "learning_rate": 2.8102761879938634e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074812720064073805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00328888816293329, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018251432105898857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9010595679283142, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24450, + "weighted_flow_matching_loss": 0.020992217026650904, + "weighted_rkd_loss": 0.01644565062597394 + }, + { + "epoch": 7.116671515856852, + "flow_matching_loss": 0.020043016783893107, + "grad_norm": 0.1771487295627594, + "learning_rate": 2.8100749139342437e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007285559270530939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031244661891832947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017695584613829852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9009813845157624, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24460, + "weighted_flow_matching_loss": 0.020043016783893107, + "weighted_rkd_loss": 0.015943441819399596 + }, + { + "epoch": 7.119581029967995, + "flow_matching_loss": 0.022981683351099493, + "grad_norm": 0.21673932671546936, + "learning_rate": 2.8098735403830803e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00838024471886456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003985657333396375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02074614688754082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9009031772613525, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24470, + "weighted_flow_matching_loss": 0.022981683351099493, + "weighted_rkd_loss": 0.018690280523151158 + }, + { + "epoch": 7.122490544079139, + "flow_matching_loss": 0.023309127427637577, + "grad_norm": 0.2217729240655899, + "learning_rate": 2.809672067355666e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007918924558907747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003384352894499898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019222201965749263, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9008249342441559, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24480, + "weighted_flow_matching_loss": 0.023309127427637577, + "weighted_rkd_loss": 0.01731580952182412 + }, + { + "epoch": 7.125400058190282, + "flow_matching_loss": 0.01995259216055274, + "grad_norm": 0.20637056231498718, + "learning_rate": 2.809470494867301e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007769469823688269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003434578003361821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01897351760417223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9007466614246369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24490, + "weighted_flow_matching_loss": 0.01995259216055274, + "weighted_rkd_loss": 0.01709034200757742 + }, + { + "epoch": 7.128309572301426, + "flow_matching_loss": 0.021188033744692802, + "grad_norm": 0.20219579339027405, + "learning_rate": 2.8092688229332937e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008345962688326835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038410435197874905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020532968733459712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9006683826446533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24500, + "weighted_flow_matching_loss": 0.021188033744692802, + "weighted_rkd_loss": 0.018493413738906384 + }, + { + "epoch": 7.131219086412569, + "flow_matching_loss": 0.021249824762344362, + "grad_norm": 0.2451942265033722, + "learning_rate": 2.80906705156896e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007812266424298287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034087386215105654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019033271446824075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9005900621414185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24510, + "weighted_flow_matching_loss": 0.021249824762344362, + "weighted_rkd_loss": 0.017141154780983926 + }, + { + "epoch": 7.134128600523712, + "flow_matching_loss": 0.020178070291876793, + "grad_norm": 0.1940200924873352, + "learning_rate": 2.8088651807896227e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006643261015415192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028872159076854587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016173737961798905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9005117118358612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24520, + "weighted_flow_matching_loss": 0.020178070291876793, + "weighted_rkd_loss": 0.014564657770097255 + }, + { + "epoch": 7.137038114634856, + "flow_matching_loss": 0.02144098673015833, + "grad_norm": 0.3025099039077759, + "learning_rate": 2.8086632106106124e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007384503912180662, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003264491260051727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01803349917754531, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9004333317279816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24530, + "weighted_flow_matching_loss": 0.02144098673015833, + "weighted_rkd_loss": 0.01623795460909605 + }, + { + "epoch": 7.1399476287459995, + "flow_matching_loss": 0.020637661684304476, + "grad_norm": 0.210938960313797, + "learning_rate": 2.808461141047268e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008218614151701331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00352752897888422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019964757189154626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9003549337387085, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24540, + "weighted_flow_matching_loss": 0.020637661684304476, + "weighted_rkd_loss": 0.017975401505827902 + }, + { + "epoch": 7.142857142857143, + "flow_matching_loss": 0.022014399245381355, + "grad_norm": 0.2084210216999054, + "learning_rate": 2.8082589721149348e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065782320685684684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031193404225632547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016275804582983257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9002765059471131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24550, + "weighted_flow_matching_loss": 0.022014399245381355, + "weighted_rkd_loss": 0.014652738440781831 + }, + { + "epoch": 7.145766656968286, + "flow_matching_loss": 0.020902454666793347, + "grad_norm": 0.20631271600723267, + "learning_rate": 2.8080567038289667e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00704432581551373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034060770412907004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017494728788733482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9001980483531952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24560, + "weighted_flow_matching_loss": 0.020902454666793347, + "weighted_rkd_loss": 0.015748747065663336 + }, + { + "epoch": 7.14867617107943, + "flow_matching_loss": 0.021490955166518687, + "grad_norm": 0.2138139307498932, + "learning_rate": 2.8078543362047242e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006876041321083904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029860347509384154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01673811748623848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.900119560956955, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24570, + "weighted_flow_matching_loss": 0.021490955166518687, + "weighted_rkd_loss": 0.015066340938210488 + }, + { + "epoch": 7.1515856851905735, + "flow_matching_loss": 0.020421948470175266, + "grad_norm": 0.22862032055854797, + "learning_rate": 2.807651869257576e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007427451992407441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034537630854174495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01830866700038314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9000410437583923, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24580, + "weighted_flow_matching_loss": 0.020421948470175266, + "weighted_rkd_loss": 0.016478582471609115 + }, + { + "epoch": 7.154495199301716, + "flow_matching_loss": 0.02184693291783333, + "grad_norm": 0.3422026038169861, + "learning_rate": 2.807449303002898e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007348703127354384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003374644345603883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018072050623595716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8999625086784363, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24590, + "weighted_flow_matching_loss": 0.02184693291783333, + "weighted_rkd_loss": 0.016264136880636215 + }, + { + "epoch": 7.15740471341286, + "flow_matching_loss": 0.020578276179730893, + "grad_norm": 0.24966967105865479, + "learning_rate": 2.8072466374560737e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007421679841354489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031962607288733126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01803962020203471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8998839437961579, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24600, + "weighted_flow_matching_loss": 0.020578276179730893, + "weighted_rkd_loss": 0.01623358326032758 + }, + { + "epoch": 7.160314227524004, + "flow_matching_loss": 0.02187656592577696, + "grad_norm": 0.22103610634803772, + "learning_rate": 2.807043872632495e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074758404865860936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034049572655931114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01835663812234998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8998053431510925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24610, + "weighted_flow_matching_loss": 0.02187656592577696, + "weighted_rkd_loss": 0.016517387609928845 + }, + { + "epoch": 7.1632237416351465, + "flow_matching_loss": 0.020956468023359775, + "grad_norm": 0.2730531096458435, + "learning_rate": 2.806841008547559e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00677483850158751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030122545547783377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016561931651085614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8997267246246338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24620, + "weighted_flow_matching_loss": 0.020956468023359775, + "weighted_rkd_loss": 0.014901210553944111 + }, + { + "epoch": 7.16613325574629, + "flow_matching_loss": 0.019173897337168455, + "grad_norm": 0.2291383147239685, + "learning_rate": 2.8066380452166738e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00724813905544579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032985099125653504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017794787790626286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8996480703353882, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24630, + "weighted_flow_matching_loss": 0.019173897337168455, + "weighted_rkd_loss": 0.016009051445871593 + }, + { + "epoch": 7.169042769857434, + "flow_matching_loss": 0.02223674561828375, + "grad_norm": 0.23616397380828857, + "learning_rate": 2.8064349826552513e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00751840490847826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003418734553270042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018455544579774143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8995693922042847, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24640, + "weighted_flow_matching_loss": 0.02223674561828375, + "weighted_rkd_loss": 0.016602076403796674 + }, + { + "epoch": 7.171952283968578, + "flow_matching_loss": 0.020044574327766896, + "grad_norm": 0.25245732069015503, + "learning_rate": 2.806231820878714e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007583655137568712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034726811340078712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018639991339296103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8994906961917877, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24650, + "weighted_flow_matching_loss": 0.020044574327766896, + "weighted_rkd_loss": 0.016766433976590632 + }, + { + "epoch": 7.1748617980797205, + "flow_matching_loss": 0.02028704099357128, + "grad_norm": 0.19237713515758514, + "learning_rate": 2.8060285599024904e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007556729577481746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034058034187182786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01851926278322935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8994119584560394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24660, + "weighted_flow_matching_loss": 0.02028704099357128, + "weighted_rkd_loss": 0.016656418796628714 + }, + { + "epoch": 7.177771312190864, + "flow_matching_loss": 0.02049462851136923, + "grad_norm": 0.20332391560077667, + "learning_rate": 2.8058251997420165e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007182748382911086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030889651272445915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017454461939632893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8993332028388977, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24670, + "weighted_flow_matching_loss": 0.02049462851136923, + "weighted_rkd_loss": 0.01569736786186695 + }, + { + "epoch": 7.180680826302008, + "flow_matching_loss": 0.022045475989580156, + "grad_norm": 0.2299106866121292, + "learning_rate": 2.805621740412736e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007178952638059855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003199637192301452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017557542771100998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8992544114589691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24680, + "weighted_flow_matching_loss": 0.022045475989580156, + "weighted_rkd_loss": 0.01578864473849535 + }, + { + "epoch": 7.183590340413151, + "flow_matching_loss": 0.02062083184719086, + "grad_norm": 0.18944846093654633, + "learning_rate": 2.8054181819301012e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006654107477515936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029260497190989553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016234264709055424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8991755962371826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24690, + "weighted_flow_matching_loss": 0.02062083184719086, + "weighted_rkd_loss": 0.01459745755419135 + }, + { + "epoch": 7.186499854524294, + "flow_matching_loss": 0.019702008552849294, + "grad_norm": 0.22188428044319153, + "learning_rate": 2.8052145243095706e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006934153661131859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031505475519225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017018854897469283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8990967631340027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24700, + "weighted_flow_matching_loss": 0.019702008552849294, + "weighted_rkd_loss": 0.015301575791090727 + }, + { + "epoch": 7.189409368635438, + "flow_matching_loss": 0.021359289158135654, + "grad_norm": 0.21755656599998474, + "learning_rate": 2.80501076756661e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007409495254978538, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033259137300774453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018144904263317586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8990179002285004, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24710, + "weighted_flow_matching_loss": 0.021359289158135654, + "weighted_rkd_loss": 0.01631260635331273 + }, + { + "epoch": 7.192318882746581, + "flow_matching_loss": 0.021013476140797138, + "grad_norm": 0.2112962305545807, + "learning_rate": 2.8048069117166942e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006795460684224963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032269414281472565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016817862913012505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8989389955997467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24720, + "weighted_flow_matching_loss": 0.021013476140797138, + "weighted_rkd_loss": 0.015118221379816532 + }, + { + "epoch": 7.195228396857725, + "flow_matching_loss": 0.019741719588637352, + "grad_norm": 0.2004958689212799, + "learning_rate": 2.8046029567753047e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007376238796859979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033399893436580896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018092466797679663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.898860079050064, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24730, + "weighted_flow_matching_loss": 0.019741719588637352, + "weighted_rkd_loss": 0.016262594237923624 + }, + { + "epoch": 7.198137910968868, + "flow_matching_loss": 0.02127796858549118, + "grad_norm": 0.24402543902397156, + "learning_rate": 2.80439890275793e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007448625564575195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032890428556129336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01818629400804639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8987811207771301, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24740, + "weighted_flow_matching_loss": 0.02127796858549118, + "weighted_rkd_loss": 0.01634548706933856 + }, + { + "epoch": 7.201047425080012, + "flow_matching_loss": 0.022240668442100285, + "grad_norm": 0.22429299354553223, + "learning_rate": 2.804194749680067e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007478194497525692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003329983237199485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01828637197613716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8987021446228027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24750, + "weighted_flow_matching_loss": 0.022240668442100285, + "weighted_rkd_loss": 0.016434040386229753 + }, + { + "epoch": 7.203956939191155, + "flow_matching_loss": 0.02002592496573925, + "grad_norm": 0.2525312006473541, + "learning_rate": 2.8039904975572193e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006868252204731107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003051021578721702, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016787525918334722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8986231446266174, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24760, + "weighted_flow_matching_loss": 0.02002592496573925, + "weighted_rkd_loss": 0.01508564678952098 + }, + { + "epoch": 7.206866453302299, + "flow_matching_loss": 0.021361051127314568, + "grad_norm": 0.22732983529567719, + "learning_rate": 2.8037861464048994e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008836137084290385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038391056703403594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02151137962937355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8985441088676452, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24770, + "weighted_flow_matching_loss": 0.021361051127314568, + "weighted_rkd_loss": 0.01932898731902242 + }, + { + "epoch": 7.209775967413442, + "flow_matching_loss": 0.020091792568564415, + "grad_norm": 0.23391132056713104, + "learning_rate": 2.803581696238626e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007790631987154484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00379003353882581, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01937129758298397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8984650492668151, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24780, + "weighted_flow_matching_loss": 0.020091792568564415, + "weighted_rkd_loss": 0.017404459044337274 + }, + { + "epoch": 7.212685481524585, + "flow_matching_loss": 0.02128077931702137, + "grad_norm": 0.2118961364030838, + "learning_rate": 2.8033771470739248e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007034212118014693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030341234290972353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017102547734975816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8983859598636628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24790, + "weighted_flow_matching_loss": 0.02128077931702137, + "weighted_rkd_loss": 0.015364692825824023 + }, + { + "epoch": 7.215594995635729, + "flow_matching_loss": 0.019778406620025633, + "grad_norm": 0.21580110490322113, + "learning_rate": 2.8031724989263315e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007234604423865676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003132482897490263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01760169193148613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8983068466186523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24800, + "weighted_flow_matching_loss": 0.019778406620025633, + "weighted_rkd_loss": 0.01581171378493309 + }, + { + "epoch": 7.2185045097468725, + "flow_matching_loss": 0.0203732724301517, + "grad_norm": 0.22808468341827393, + "learning_rate": 2.802967751811387e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006931215524673462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002953188889659941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016815619822591544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8982277035713195, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24810, + "weighted_flow_matching_loss": 0.0203732724301517, + "weighted_rkd_loss": 0.015104250982403755 + }, + { + "epoch": 7.221414023858015, + "flow_matching_loss": 0.01983488705009222, + "grad_norm": 0.3022323548793793, + "learning_rate": 2.8027629057446406e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00684276819229126, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029472869355231524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01663282364606857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8981485366821289, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24820, + "weighted_flow_matching_loss": 0.01983488705009222, + "weighted_rkd_loss": 0.014938734751194716 + }, + { + "epoch": 7.224323537969159, + "flow_matching_loss": 0.020522416569292545, + "grad_norm": 0.3556026220321655, + "learning_rate": 2.8025579607416495e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007912656525149941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035437064478173854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01936901956796646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8980693399906159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24830, + "weighted_flow_matching_loss": 0.020522416569292545, + "weighted_rkd_loss": 0.017394692450761796 + }, + { + "epoch": 7.227233052080303, + "flow_matching_loss": 0.02048003189265728, + "grad_norm": 0.21191942691802979, + "learning_rate": 2.8023529168179767e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006814835127443075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029806498903781177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016610320005565882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8979901194572448, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24840, + "weighted_flow_matching_loss": 0.02048003189265728, + "weighted_rkd_loss": 0.01491587720811367 + }, + { + "epoch": 7.2301425661914465, + "flow_matching_loss": 0.021841926500201225, + "grad_norm": 0.1920260339975357, + "learning_rate": 2.8021477739891946e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072191072627902034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031888349680230022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01762704933062196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.897910863161087, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24850, + "weighted_flow_matching_loss": 0.021841926500201225, + "weighted_rkd_loss": 0.01582750305533409 + }, + { + "epoch": 7.233052080302589, + "flow_matching_loss": 0.02048401115462184, + "grad_norm": 0.19851809740066528, + "learning_rate": 2.801942532270883e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069324942771345375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029967059846967457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016861694678664208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8978315889835358, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24860, + "weighted_flow_matching_loss": 0.02048401115462184, + "weighted_rkd_loss": 0.015138989873230457 + }, + { + "epoch": 7.235961594413733, + "flow_matching_loss": 0.021048931777477263, + "grad_norm": 0.20556943118572235, + "learning_rate": 2.8017371916786278e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006970917759463191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003231644444167614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01717347977682948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8977522850036621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24870, + "weighted_flow_matching_loss": 0.021048931777477263, + "weighted_rkd_loss": 0.01541753513738513 + }, + { + "epoch": 7.238871108524877, + "flow_matching_loss": 0.02238921485841274, + "grad_norm": 0.2067803591489792, + "learning_rate": 2.8015317522280237e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006799177872017026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031073810532689095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016705736797302962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8976729512214661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24880, + "weighted_flow_matching_loss": 0.02238921485841274, + "weighted_rkd_loss": 0.014996295142918826 + }, + { + "epoch": 7.2417806226360195, + "flow_matching_loss": 0.020973012130707503, + "grad_norm": 0.22740919888019562, + "learning_rate": 2.8013262139346723e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006959526287391782, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00313819725997746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017057250067591668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8975935876369476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24890, + "weighted_flow_matching_loss": 0.020973012130707503, + "weighted_rkd_loss": 0.01531050270423293 + }, + { + "epoch": 7.244690136747163, + "flow_matching_loss": 0.0213782231323421, + "grad_norm": 0.20355470478534698, + "learning_rate": 2.801120576814183e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006895028054714203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029568075435236095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016746863909065723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8975142002105713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24900, + "weighted_flow_matching_loss": 0.0213782231323421, + "weighted_rkd_loss": 0.01503056352958083 + }, + { + "epoch": 7.247599650858307, + "flow_matching_loss": 0.020575156807899474, + "grad_norm": 0.197248175740242, + "learning_rate": 2.8009148408821726e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007373132230713964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031106274342164396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017856891825795175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8974347889423371, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24910, + "weighted_flow_matching_loss": 0.020575156807899474, + "weighted_rkd_loss": 0.016025424655526876 + }, + { + "epoch": 7.25050916496945, + "flow_matching_loss": 0.019620054587721824, + "grad_norm": 0.1916624903678894, + "learning_rate": 2.8007090061542657e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007517708698287606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003246334148570895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018281751684844492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8973553478717804, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24920, + "weighted_flow_matching_loss": 0.019620054587721824, + "weighted_rkd_loss": 0.0164052601903677 + }, + { + "epoch": 7.2534186790805935, + "flow_matching_loss": 0.021945754624903203, + "grad_norm": 0.23883862793445587, + "learning_rate": 2.800503072646093e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007062292192131281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029796527000144125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017104236967861654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8972758769989013, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24930, + "weighted_flow_matching_loss": 0.021945754624903203, + "weighted_rkd_loss": 0.015347231645137072 + }, + { + "epoch": 7.256328193191737, + "flow_matching_loss": 0.021161599457263945, + "grad_norm": 0.21301531791687012, + "learning_rate": 2.800297040373295e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065131566952914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026951572275720538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015721470676362516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8971963822841644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24940, + "weighted_flow_matching_loss": 0.021161599457263945, + "weighted_rkd_loss": 0.014105271827429534 + }, + { + "epoch": 7.25923770730288, + "flow_matching_loss": 0.019679922237992288, + "grad_norm": 0.19579477608203888, + "learning_rate": 2.8000909093515178e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007065379153937101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003263002657331526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017393760941922666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8971168577671051, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24950, + "weighted_flow_matching_loss": 0.019679922237992288, + "weighted_rkd_loss": 0.01560425478965044 + }, + { + "epoch": 7.262147221414024, + "flow_matching_loss": 0.018531126901507378, + "grad_norm": 0.26706621050834656, + "learning_rate": 2.799884679596416e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007411468727514148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031834508758038283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018006388377398253, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8970372974872589, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24960, + "weighted_flow_matching_loss": 0.018531126901507378, + "weighted_rkd_loss": 0.016152367740869523 + }, + { + "epoch": 7.265056735525167, + "flow_matching_loss": 0.020150916650891304, + "grad_norm": 0.23883023858070374, + "learning_rate": 2.7996783511236513e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007308446569368243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030948513885959983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01771174483001232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8969577252864838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24970, + "weighted_flow_matching_loss": 0.020150916650891304, + "weighted_rkd_loss": 0.015886693261563778 + }, + { + "epoch": 7.267966249636311, + "flow_matching_loss": 0.021140961535274984, + "grad_norm": 0.19879277050495148, + "learning_rate": 2.7994719239488927e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066622348967939615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029953312827274204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01631980100646615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8968781232833862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24980, + "weighted_flow_matching_loss": 0.021140961535274984, + "weighted_rkd_loss": 0.01463689636439085 + }, + { + "epoch": 7.270875763747454, + "flow_matching_loss": 0.019663435500115158, + "grad_norm": 0.21825958788394928, + "learning_rate": 2.7992653980878175e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006880706269294023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030477520078420637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01680916463956237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8967984914779663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24990, + "weighted_flow_matching_loss": 0.019663435500115158, + "weighted_rkd_loss": 0.015074421279132366 + }, + { + "epoch": 7.273785277858598, + "flow_matching_loss": 0.02110330304130912, + "grad_norm": 0.1935225874185562, + "learning_rate": 2.7990587735561094e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007154904957860708, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033048329642042518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01761464271694422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.896718829870224, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25000, + "weighted_flow_matching_loss": 0.02110330304130912, + "weighted_rkd_loss": 0.015795391984283924 + }, + { + "epoch": 7.276694791969741, + "flow_matching_loss": 0.02106508780270815, + "grad_norm": 0.2470180243253708, + "learning_rate": 2.798852050369461e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007752446038648486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003691802779212594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019196694903075694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8966391503810882, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25010, + "weighted_flow_matching_loss": 0.02106508780270815, + "weighted_rkd_loss": 0.017212506104260682 + }, + { + "epoch": 7.279604306080884, + "flow_matching_loss": 0.021091966051608325, + "grad_norm": 0.20913472771644592, + "learning_rate": 2.7986452285435706e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007400690857321024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032238221727311613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018025204073637724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8965594291687011, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25020, + "weighted_flow_matching_loss": 0.021091966051608325, + "weighted_rkd_loss": 0.01616068398579955 + }, + { + "epoch": 7.282513820192028, + "flow_matching_loss": 0.020578378066420554, + "grad_norm": 0.19160489737987518, + "learning_rate": 2.7984383080941458e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00786658343859017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035611702827736736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01929433746263385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8964796960353851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25030, + "weighted_flow_matching_loss": 0.020578378066420554, + "weighted_rkd_loss": 0.01729698050767183 + }, + { + "epoch": 7.285423334303172, + "flow_matching_loss": 0.022070186026394368, + "grad_norm": 0.2308981865644455, + "learning_rate": 2.7982312890369e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007008095551282167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031362537294626234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017152444645762443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8963999271392822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25040, + "weighted_flow_matching_loss": 0.022070186026394368, + "weighted_rkd_loss": 0.015375441964715719 + }, + { + "epoch": 7.288332848414315, + "flow_matching_loss": 0.02007784377783537, + "grad_norm": 0.2077474594116211, + "learning_rate": 2.798024171387556e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067259869538247585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028922690311446787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016344243101775645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.896320128440857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25050, + "weighted_flow_matching_loss": 0.02007784377783537, + "weighted_rkd_loss": 0.014649703912436963 + }, + { + "epoch": 7.291242362525458, + "flow_matching_loss": 0.020371813140809537, + "grad_norm": 0.17640072107315063, + "learning_rate": 2.7978169551618414e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007239629980176687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031683987705036997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017647658381611107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8962402939796448, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25060, + "weighted_flow_matching_loss": 0.020371813140809537, + "weighted_rkd_loss": 0.015816571097820996 + }, + { + "epoch": 7.294151876636602, + "flow_matching_loss": 0.02215842679142952, + "grad_norm": 0.18979862332344055, + "learning_rate": 2.7976096403754948e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007189032947644591, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033141708467155696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017692236788570882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8961604714393616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25070, + "weighted_flow_matching_loss": 0.02215842679142952, + "weighted_rkd_loss": 0.01585506582632661 + }, + { + "epoch": 7.2970613907477455, + "flow_matching_loss": 0.0206104195676744, + "grad_norm": 0.21450521051883698, + "learning_rate": 2.7974022270442594e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00793913695961237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036313843447715044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019509658310562374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8960805773735047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25080, + "weighted_flow_matching_loss": 0.0206104195676744, + "weighted_rkd_loss": 0.017482287622988223 + }, + { + "epoch": 7.299970904858888, + "flow_matching_loss": 0.020095619186758996, + "grad_norm": 0.19390112161636353, + "learning_rate": 2.7971947151838863e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006780432118102908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002958158543333411, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016519022546708582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8960006773471832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25090, + "weighted_flow_matching_loss": 0.020095619186758996, + "weighted_rkd_loss": 0.014801069814711808 + }, + { + "epoch": 7.302880418970032, + "flow_matching_loss": 0.01810687417164445, + "grad_norm": 0.18218129873275757, + "learning_rate": 2.796987104810136e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00842556064017117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039484822424128655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020799603778868914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.895920741558075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25100, + "weighted_flow_matching_loss": 0.01810687417164445, + "weighted_rkd_loss": 0.018634829111397265 + }, + { + "epoch": 7.305789933081176, + "flow_matching_loss": 0.020675793755799532, + "grad_norm": 0.21794117987155914, + "learning_rate": 2.7967793959387742e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007341045467182994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003259249497205019, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017941340710967778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8958407878875733, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25110, + "weighted_flow_matching_loss": 0.020675793755799532, + "weighted_rkd_loss": 0.01607255218550563 + }, + { + "epoch": 7.308699447192319, + "flow_matching_loss": 0.02076775711029768, + "grad_norm": 0.18282227218151093, + "learning_rate": 2.7965715885855754e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006667976360768079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027683988912031053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016104351542890073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8957607984542847, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25120, + "weighted_flow_matching_loss": 0.02076775711029768, + "weighted_rkd_loss": 0.01442565005272627 + }, + { + "epoch": 7.311608961303462, + "flow_matching_loss": 0.01900488492101431, + "grad_norm": 0.24331827461719513, + "learning_rate": 2.7963636827663207e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006938950205221772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030500120483338833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01692791245877743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8956807851791382, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25130, + "weighted_flow_matching_loss": 0.01900488492101431, + "weighted_rkd_loss": 0.015161998104304075 + }, + { + "epoch": 7.314518475414606, + "flow_matching_loss": 0.019997198600322008, + "grad_norm": 0.23930279910564423, + "learning_rate": 2.7961556784968004e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006867963587865233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029387234011664987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0166746505536139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8956007480621337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25140, + "weighted_flow_matching_loss": 0.019997198600322008, + "weighted_rkd_loss": 0.014933800138533116 + }, + { + "epoch": 7.317427989525749, + "flow_matching_loss": 0.019682951085269452, + "grad_norm": 0.20229992270469666, + "learning_rate": 2.7959475757928094e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071867769584059715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003186235181055963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01755978902801871, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8955206751823426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25150, + "weighted_flow_matching_loss": 0.019682951085269452, + "weighted_rkd_loss": 0.015725145954638718 + }, + { + "epoch": 7.3203375036368925, + "flow_matching_loss": 0.02004459695890546, + "grad_norm": 0.2092834860086441, + "learning_rate": 2.795739374670153e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006752302823588252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002916310075670481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016420916002243758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8954405903816223, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25160, + "weighted_flow_matching_loss": 0.02004459695890546, + "weighted_rkd_loss": 0.014703929331153632 + }, + { + "epoch": 7.323247017748036, + "flow_matching_loss": 0.02028122264891863, + "grad_norm": 0.17080281674861908, + "learning_rate": 2.7955310751446418e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068626456893980505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002886026771739125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016611318197101356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8953604698181152, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25170, + "weighted_flow_matching_loss": 0.02028122264891863, + "weighted_rkd_loss": 0.014873113948851824 + }, + { + "epoch": 7.32615653185918, + "flow_matching_loss": 0.021378986164927484, + "grad_norm": 0.19869548082351685, + "learning_rate": 2.7953226772320955e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007045650202780962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031054669758304953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017196767404675482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8952803254127503, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25180, + "weighted_flow_matching_loss": 0.021378986164927484, + "weighted_rkd_loss": 0.01539588263258338 + }, + { + "epoch": 7.329066045970323, + "flow_matching_loss": 0.01944811185821891, + "grad_norm": 0.203749418258667, + "learning_rate": 2.7951141809483404e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006538868276402354, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00312841369304806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016206150129437446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8952001452445983, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25190, + "weighted_flow_matching_loss": 0.01944811185821891, + "weighted_rkd_loss": 0.0145077770575881 + }, + { + "epoch": 7.3319755600814664, + "flow_matching_loss": 0.0222668981179595, + "grad_norm": 0.21869155764579773, + "learning_rate": 2.79490558630921e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007195541309192777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031304241390898825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017521506547927855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8951199471950531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25200, + "weighted_flow_matching_loss": 0.0222668981179595, + "weighted_rkd_loss": 0.015683887992054224 + }, + { + "epoch": 7.33488507419261, + "flow_matching_loss": 0.021527619659900667, + "grad_norm": 0.20986086130142212, + "learning_rate": 2.7946968933305456e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006720233056694269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002928528189659119, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016368994675576686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8950397133827209, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25210, + "weighted_flow_matching_loss": 0.021527619659900667, + "weighted_rkd_loss": 0.01465093893930316 + }, + { + "epoch": 7.337794588303753, + "flow_matching_loss": 0.018690549582242966, + "grad_norm": 0.17634882032871246, + "learning_rate": 2.794488102028197e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006605774257332086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028875722200609744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01609912058338523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8949594497680664, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25220, + "weighted_flow_matching_loss": 0.018690549582242966, + "weighted_rkd_loss": 0.01440809527412057 + }, + { + "epoch": 7.340704102414897, + "flow_matching_loss": 0.019746249821037055, + "grad_norm": 0.18380218744277954, + "learning_rate": 2.7942792124180198e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006775641860440373, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029946594731882215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016545943170785903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8948791742324829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25230, + "weighted_flow_matching_loss": 0.019746249821037055, + "weighted_rkd_loss": 0.014806641358882188 + }, + { + "epoch": 7.34361361652604, + "flow_matching_loss": 0.01941441036760807, + "grad_norm": 0.20252487063407898, + "learning_rate": 2.794070224515878e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006789958383888006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00306906474288553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016648981347680093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.894798856973648, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25240, + "weighted_flow_matching_loss": 0.01941441036760807, + "weighted_rkd_loss": 0.014897528570145369 + }, + { + "epoch": 7.346523130637183, + "flow_matching_loss": 0.020035882666707038, + "grad_norm": 0.20054951310157776, + "learning_rate": 2.7938611383376423e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006512573221698403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002801863313652575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01582701001316309, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8947185277938843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25250, + "weighted_flow_matching_loss": 0.020035882666707038, + "weighted_rkd_loss": 0.014160702563822269 + }, + { + "epoch": 7.349432644748327, + "flow_matching_loss": 0.0186926300637424, + "grad_norm": 0.23499730229377747, + "learning_rate": 2.793651953899192e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065267957281321285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028636342962272466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015917225740849972, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8946381628513336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25260, + "weighted_flow_matching_loss": 0.0186926300637424, + "weighted_rkd_loss": 0.014240182004868984 + }, + { + "epoch": 7.352342158859471, + "flow_matching_loss": 0.02093365751206875, + "grad_norm": 0.18720944225788116, + "learning_rate": 2.7934426712164136e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006914324313402176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028791965916752814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016707845497876406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.894557774066925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25270, + "weighted_flow_matching_loss": 0.02093365751206875, + "weighted_rkd_loss": 0.014946133270859719 + }, + { + "epoch": 7.355251672970613, + "flow_matching_loss": 0.021976843103766443, + "grad_norm": 0.2043572962284088, + "learning_rate": 2.7932332903052e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006639264477416873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030520435189828276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016330572497099637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8944773375988007, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25280, + "weighted_flow_matching_loss": 0.021976843103766443, + "weighted_rkd_loss": 0.014607347268611192 + }, + { + "epoch": 7.358161187081757, + "flow_matching_loss": 0.02106471052393317, + "grad_norm": 0.1941019892692566, + "learning_rate": 2.7930238111814525e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006224642973393202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002838409796822816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015287695918232203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8943969249725342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25290, + "weighted_flow_matching_loss": 0.02106471052393317, + "weighted_rkd_loss": 0.013673275709152222 + }, + { + "epoch": 7.361070701192901, + "flow_matching_loss": 0.020098974369466303, + "grad_norm": 0.18498212099075317, + "learning_rate": 2.7928142338610802e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006772019062191248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003026163135655224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016570201236754657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8943164348602295, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25300, + "weighted_flow_matching_loss": 0.020098974369466303, + "weighted_rkd_loss": 0.014818978123366832 + }, + { + "epoch": 7.363980215304045, + "flow_matching_loss": 0.020819661766290666, + "grad_norm": 0.22174319624900818, + "learning_rate": 2.792604558359998e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007047701068222523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003015164332464337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01711056660860777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8942359387874603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25310, + "weighted_flow_matching_loss": 0.020819661766290666, + "weighted_rkd_loss": 0.015300833526998759 + }, + { + "epoch": 7.366889729415187, + "flow_matching_loss": 0.019368493929505348, + "grad_norm": 0.1928389072418213, + "learning_rate": 2.7923947846941306e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006670884368941188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028388485545292495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016180617175996302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8941554009914399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25320, + "weighted_flow_matching_loss": 0.019368493929505348, + "weighted_rkd_loss": 0.014468009490519762 + }, + { + "epoch": 7.369799243526331, + "flow_matching_loss": 0.021123512275516988, + "grad_norm": 0.19295655190944672, + "learning_rate": 2.7921849128794082e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006731057306751609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028934046858921647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01635551927611232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8940748572349548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25330, + "weighted_flow_matching_loss": 0.021123512275516988, + "weighted_rkd_loss": 0.014623031672090291 + }, + { + "epoch": 7.372708757637475, + "flow_matching_loss": 0.02013948205858469, + "grad_norm": 0.19954055547714233, + "learning_rate": 2.7919749429317694e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007108434056863189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031558544375002383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017372722458094358, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8939942717552185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25340, + "weighted_flow_matching_loss": 0.02013948205858469, + "weighted_rkd_loss": 0.015531154442578554 + }, + { + "epoch": 7.375618271748618, + "flow_matching_loss": 0.020703672990202905, + "grad_norm": 0.2507128119468689, + "learning_rate": 2.79176487486716e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071891781874001024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003080041380599141, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01745839770883322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8939136624336242, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25350, + "weighted_flow_matching_loss": 0.020703672990202905, + "weighted_rkd_loss": 0.015606280136853456 + }, + { + "epoch": 7.378527785859761, + "flow_matching_loss": 0.02270576674491167, + "grad_norm": 0.23060384392738342, + "learning_rate": 2.7915547087015328e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006943489285185933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002967516309581697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01685449481010437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8938330292701722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25360, + "weighted_flow_matching_loss": 0.02270576674491167, + "weighted_rkd_loss": 0.015065095759928227 + }, + { + "epoch": 7.381437299970905, + "flow_matching_loss": 0.022255540639162064, + "grad_norm": 0.16534726321697235, + "learning_rate": 2.7913444444508494e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006756712077185511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028234417317435146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01633686600252986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.893752372264862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25370, + "weighted_flow_matching_loss": 0.022255540639162064, + "weighted_rkd_loss": 0.014601161796599626 + }, + { + "epoch": 7.384346814082049, + "flow_matching_loss": 0.0199126610532403, + "grad_norm": 0.21394412219524384, + "learning_rate": 2.7911340821310777e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006750528048723936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002843236643821001, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01634429283440113, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8936716794967652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25380, + "weighted_flow_matching_loss": 0.0199126610532403, + "weighted_rkd_loss": 0.014606421440839767 + }, + { + "epoch": 7.387256328193192, + "flow_matching_loss": 0.02006089696660638, + "grad_norm": 0.22692517936229706, + "learning_rate": 2.790923621758193e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007193322479724884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030756853870116175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017462330218404532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8935909628868103, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25390, + "weighted_flow_matching_loss": 0.02006089696660638, + "weighted_rkd_loss": 0.015604159142822028 + }, + { + "epoch": 7.390165842304335, + "flow_matching_loss": 0.020912976749241353, + "grad_norm": 0.24535131454467773, + "learning_rate": 2.7907130633481787e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007037954591214657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032407503575086595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01731665963307023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8935102224349976, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25400, + "weighted_flow_matching_loss": 0.020912976749241353, + "weighted_rkd_loss": 0.015472578350454569 + }, + { + "epoch": 7.393075356415479, + "flow_matching_loss": 0.02001263592392206, + "grad_norm": 0.18792280554771423, + "learning_rate": 2.790502406917025e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064620448276400564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028687494108453394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015792839229106903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8934294581413269, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25410, + "weighted_flow_matching_loss": 0.02001263592392206, + "weighted_rkd_loss": 0.014109809510409832 + }, + { + "epoch": 7.395984870526622, + "flow_matching_loss": 0.019810461718589068, + "grad_norm": 0.21229571104049683, + "learning_rate": 2.79029165248073e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007164568686857819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003184000542387366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017513137962669133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.893348652124405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25420, + "weighted_flow_matching_loss": 0.019810461718589068, + "weighted_rkd_loss": 0.015645342133939266 + }, + { + "epoch": 7.3988943846377655, + "flow_matching_loss": 0.02266913065686822, + "grad_norm": 0.23539166152477264, + "learning_rate": 2.7900808000552998e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00722476439550519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033044451847672462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017753974162042142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.893267834186554, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25430, + "weighted_flow_matching_loss": 0.02266913065686822, + "weighted_rkd_loss": 0.015859018359333277 + }, + { + "epoch": 7.401803898748909, + "flow_matching_loss": 0.020609965175390245, + "grad_norm": 0.19430050253868103, + "learning_rate": 2.7898698496567457e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007359907869249583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00323947137221694, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01795928701758385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8931869745254517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25440, + "weighted_flow_matching_loss": 0.020609965175390245, + "weighted_rkd_loss": 0.016041044052690266 + }, + { + "epoch": 7.404713412860052, + "flow_matching_loss": 0.01963582318276167, + "grad_norm": 0.28516632318496704, + "learning_rate": 2.7896588013010894e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007605642918497324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034537695813924075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018665055464953183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8931061029434204, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25450, + "weighted_flow_matching_loss": 0.01963582318276167, + "weighted_rkd_loss": 0.016669803857803346 + }, + { + "epoch": 7.407622926971196, + "flow_matching_loss": 0.020570527482777833, + "grad_norm": 0.22578752040863037, + "learning_rate": 2.7894476550043585e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007231372129172087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003266185289248824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01772892950102687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8930252015590667, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25460, + "weighted_flow_matching_loss": 0.020570527482777833, + "weighted_rkd_loss": 0.015832383185625076 + }, + { + "epoch": 7.410532441082339, + "flow_matching_loss": 0.02153829326853156, + "grad_norm": 0.20958426594734192, + "learning_rate": 2.7892364107825874e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006851832894608378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028740762965753673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01657774243503809, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8929442703723908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25470, + "weighted_flow_matching_loss": 0.02153829326853156, + "weighted_rkd_loss": 0.01480299225077033 + }, + { + "epoch": 7.413441955193482, + "flow_matching_loss": 0.018296593241393565, + "grad_norm": 0.19776691496372223, + "learning_rate": 2.789025068651819e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006874358514323831, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003037162427790463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016785879246890546, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8928633093833923, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25480, + "weighted_flow_matching_loss": 0.018296593241393565, + "weighted_rkd_loss": 0.014987505041062832 + }, + { + "epoch": 7.416351469304626, + "flow_matching_loss": 0.02242100741714239, + "grad_norm": 0.23544931411743164, + "learning_rate": 2.788813628628104e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007214242871850729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033516566734761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017780142463743688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8927823185920716, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25490, + "weighted_flow_matching_loss": 0.02242100741714239, + "weighted_rkd_loss": 0.015873823966830968 + }, + { + "epoch": 7.41926098341577, + "flow_matching_loss": 0.021691148821264505, + "grad_norm": 0.22762605547904968, + "learning_rate": 2.788602090727499e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007325013540685177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032151473918929697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01786517444998026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8927013158798218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25500, + "weighted_flow_matching_loss": 0.021691148821264505, + "weighted_rkd_loss": 0.015948266815394165 + }, + { + "epoch": 7.422170497526913, + "flow_matching_loss": 0.018682208191603423, + "grad_norm": 0.19959481060504913, + "learning_rate": 2.7883904549660693e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006624871119856835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00311425116378814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016363993380218746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8926202654838562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25510, + "weighted_flow_matching_loss": 0.018682208191603423, + "weighted_rkd_loss": 0.014606810454279185 + }, + { + "epoch": 7.425080011638056, + "flow_matching_loss": 0.021101655997335912, + "grad_norm": 0.1816529780626297, + "learning_rate": 2.788178721359887e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007617198303341866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033454442163929343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018579840660095215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8925392031669617, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25520, + "weighted_flow_matching_loss": 0.021101655997335912, + "weighted_rkd_loss": 0.016583268251270055 + }, + { + "epoch": 7.4279895257492, + "flow_matching_loss": 0.02045115437358618, + "grad_norm": 0.23654907941818237, + "learning_rate": 2.787966889925032e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007265524659305811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034221242181956767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017953173350542785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8924581170082092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25530, + "weighted_flow_matching_loss": 0.02045115437358618, + "weighted_rkd_loss": 0.0160224424675107 + }, + { + "epoch": 7.430899039860344, + "flow_matching_loss": 0.01962193548679352, + "grad_norm": 0.2839536666870117, + "learning_rate": 2.7877549606775917e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068337388802319765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002959001576527953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016626479290425777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8923769950866699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25540, + "weighted_flow_matching_loss": 0.01962193548679352, + "weighted_rkd_loss": 0.014837043080478906 + }, + { + "epoch": 7.433808553971486, + "flow_matching_loss": 0.021269401349127293, + "grad_norm": 0.2115294188261032, + "learning_rate": 2.7875429336336607e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006530108116567135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028125714277848603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01587278749793768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8922958493232727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25550, + "weighted_flow_matching_loss": 0.021269401349127293, + "weighted_rkd_loss": 0.014163199719041586 + }, + { + "epoch": 7.43671806808263, + "flow_matching_loss": 0.02117727119475603, + "grad_norm": 0.25301891565322876, + "learning_rate": 2.787330808809341e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006496936036273837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026873870519921184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015681258961558343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8922146737575531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25560, + "weighted_flow_matching_loss": 0.02117727119475603, + "weighted_rkd_loss": 0.013991042226552963 + }, + { + "epoch": 7.439627582193774, + "flow_matching_loss": 0.01931297853589058, + "grad_norm": 0.17191599309444427, + "learning_rate": 2.7871185862207418e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007200540089979768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030760843539610503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017477164790034295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8921334743499756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25570, + "weighted_flow_matching_loss": 0.01931297853589058, + "weighted_rkd_loss": 0.015592016465961932 + }, + { + "epoch": 7.442537096304917, + "flow_matching_loss": 0.021643193997442724, + "grad_norm": 0.18144182860851288, + "learning_rate": 2.7869062658839804e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070949146058410404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031462857266888024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017336115054786206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8920522451400756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25580, + "weighted_flow_matching_loss": 0.021643193997442724, + "weighted_rkd_loss": 0.01546476809307933 + }, + { + "epoch": 7.44544661041606, + "flow_matching_loss": 0.020260816160589456, + "grad_norm": 0.21699757874011993, + "learning_rate": 2.7866938478151806e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006707239802926779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003192795650102198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016607275139540432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.891970980167389, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25590, + "weighted_flow_matching_loss": 0.020260816160589456, + "weighted_rkd_loss": 0.014813175797462464 + }, + { + "epoch": 7.448356124527204, + "flow_matching_loss": 0.01974896341562271, + "grad_norm": 0.19459757208824158, + "learning_rate": 2.7864813320304747e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00745012778788805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003193623083643615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01809387858957052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8918897151947022, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25600, + "weighted_flow_matching_loss": 0.01974896341562271, + "weighted_rkd_loss": 0.016137766186147928 + }, + { + "epoch": 7.451265638638348, + "flow_matching_loss": 0.019713624194264413, + "grad_norm": 0.26789116859436035, + "learning_rate": 2.786268718546002e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065442337654531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030771065736189484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016165573988109826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8918084025382995, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25610, + "weighted_flow_matching_loss": 0.019713624194264413, + "weighted_rkd_loss": 0.014416589494794607 + }, + { + "epoch": 7.454175152749491, + "flow_matching_loss": 0.02019447311758995, + "grad_norm": 0.2846088707447052, + "learning_rate": 2.7860560073779083e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00628466964699328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002742871968075633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015312211215496063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8917270600795746, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25620, + "weighted_flow_matching_loss": 0.02019447311758995, + "weighted_rkd_loss": 0.013654281105846166 + }, + { + "epoch": 7.457084666860634, + "flow_matching_loss": 0.019573323987424375, + "grad_norm": 0.25006192922592163, + "learning_rate": 2.7858431985423483e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074928598012775184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032438055146485566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018229525070637465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8916456997394562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25630, + "weighted_flow_matching_loss": 0.019573323987424375, + "weighted_rkd_loss": 0.016254258062690496 + }, + { + "epoch": 7.459994180971778, + "flow_matching_loss": 0.019533003680408, + "grad_norm": 0.26284128427505493, + "learning_rate": 2.785630292055483e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007687592320144177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034871332929469643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018862317968159913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8915643215179443, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25640, + "weighted_flow_matching_loss": 0.019533003680408, + "weighted_rkd_loss": 0.016816933266818525 + }, + { + "epoch": 7.462903695082921, + "flow_matching_loss": 0.01999995354562998, + "grad_norm": 0.1872052252292633, + "learning_rate": 2.7854172879334815e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007324506109580397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003206881834194064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017855893913656475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8914829015731811, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25650, + "weighted_flow_matching_loss": 0.01999995354562998, + "weighted_rkd_loss": 0.015918258018791674 + }, + { + "epoch": 7.4658132091940645, + "flow_matching_loss": 0.02152755409479141, + "grad_norm": 0.2128058671951294, + "learning_rate": 2.7852041861925205e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007232867600396276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003182169038336724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01764790415763855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8914014577865601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25660, + "weighted_flow_matching_loss": 0.02152755409479141, + "weighted_rkd_loss": 0.015731361322104932 + }, + { + "epoch": 7.468722723305208, + "flow_matching_loss": 0.018632668256759643, + "grad_norm": 0.15930485725402832, + "learning_rate": 2.7849909868487825e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006927868351340294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003015579911880195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016871316637843848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.891319990158081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25670, + "weighted_flow_matching_loss": 0.018632668256759643, + "weighted_rkd_loss": 0.015037753991782666 + }, + { + "epoch": 7.471632237416351, + "flow_matching_loss": 0.020067923702299595, + "grad_norm": 0.200091153383255, + "learning_rate": 2.7847776899184592e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007486216723918915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00318568826187402, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01815812150016427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8912384927272796, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25680, + "weighted_flow_matching_loss": 0.020067923702299595, + "weighted_rkd_loss": 0.016183263529092072 + }, + { + "epoch": 7.474541751527495, + "flow_matching_loss": 0.020417289342731237, + "grad_norm": 0.19444721937179565, + "learning_rate": 2.7845642954177497e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006672057742252946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002807410527020693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016151525918394326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8911569714546204, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25690, + "weighted_flow_matching_loss": 0.020417289342731237, + "weighted_rkd_loss": 0.01439356356859207 + }, + { + "epoch": 7.4774512656386385, + "flow_matching_loss": 0.020479364320635796, + "grad_norm": 0.18577533960342407, + "learning_rate": 2.784350803362859e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007010314567014575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029699056874960663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01699053505435586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8910754263401032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25700, + "weighted_flow_matching_loss": 0.020479364320635796, + "weighted_rkd_loss": 0.015139846224337816 + }, + { + "epoch": 7.480360779749782, + "flow_matching_loss": 0.020582934375852346, + "grad_norm": 0.1886492371559143, + "learning_rate": 2.7841372137700006e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006958171585574746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030503698159009217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01696671303361654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.890993845462799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25710, + "weighted_flow_matching_loss": 0.020582934375852346, + "weighted_rkd_loss": 0.015117255598306656 + }, + { + "epoch": 7.483270293860925, + "flow_matching_loss": 0.019083615206182004, + "grad_norm": 0.23810194432735443, + "learning_rate": 2.783923526655396e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067297161091119054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002894851053133607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01635428313165903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8909122467041015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25720, + "weighted_flow_matching_loss": 0.019083615206182004, + "weighted_rkd_loss": 0.014570234250277281 + }, + { + "epoch": 7.486179807972069, + "flow_matching_loss": 0.01990107474848628, + "grad_norm": 0.2589821517467499, + "learning_rate": 2.783709742035272e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071439588442444805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030850481474772097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017372965812683105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8908306062221527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25730, + "weighted_flow_matching_loss": 0.01990107474848628, + "weighted_rkd_loss": 0.015476350486278535 + }, + { + "epoch": 7.489089322083212, + "flow_matching_loss": 0.020401748642325402, + "grad_norm": 0.23249435424804688, + "learning_rate": 2.783495859925865e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006565089290961623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002975405869074166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016105584520846605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8907489478588104, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25740, + "weighted_flow_matching_loss": 0.020401748642325402, + "weighted_rkd_loss": 0.014345975406467915 + }, + { + "epoch": 7.491998836194355, + "flow_matching_loss": 0.019899977929890157, + "grad_norm": 0.255228191614151, + "learning_rate": 2.783281880343418e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00815464542247355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035346854478120805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019843976572155954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8906672775745392, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25750, + "weighted_flow_matching_loss": 0.019899977929890157, + "weighted_rkd_loss": 0.017674343939870597 + }, + { + "epoch": 7.494908350305499, + "flow_matching_loss": 0.021419234946370123, + "grad_norm": 0.19427745044231415, + "learning_rate": 2.783067803304181e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00665287422016263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002932158438488841, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016237907018512486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8905855596065522, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25760, + "weighted_flow_matching_loss": 0.021419234946370123, + "weighted_rkd_loss": 0.014461240358650684 + }, + { + "epoch": 7.497817864416643, + "flow_matching_loss": 0.020358507707715034, + "grad_norm": 0.2504926919937134, + "learning_rate": 2.7828536288244124e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007827095594257117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003561272961087525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01921546384692192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8905038237571716, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25770, + "weighted_flow_matching_loss": 0.020358507707715034, + "weighted_rkd_loss": 0.017111430782824753 + }, + { + "epoch": 7.5007273785277855, + "flow_matching_loss": 0.022266341373324394, + "grad_norm": 0.24449384212493896, + "learning_rate": 2.782639356920376e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006444621831178665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027829037280753255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015672147460281848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8904220581054687, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25780, + "weighted_flow_matching_loss": 0.022266341373324394, + "weighted_rkd_loss": 0.013954826723784208 + }, + { + "epoch": 7.503636892638929, + "flow_matching_loss": 0.021467151306569576, + "grad_norm": 0.18803228437900543, + "learning_rate": 2.7824249876083454e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073472709394991394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029160900274291636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01761063188314438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.890340268611908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25790, + "weighted_flow_matching_loss": 0.021467151306569576, + "weighted_rkd_loss": 0.015679466258734464 + }, + { + "epoch": 7.506546406750073, + "flow_matching_loss": 0.019245310313999654, + "grad_norm": 0.2234554886817932, + "learning_rate": 2.7822105209046003e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062745264265686275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027913330704905093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015340385865420103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8902584552764893, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25800, + "weighted_flow_matching_loss": 0.019245310313999654, + "weighted_rkd_loss": 0.01365690939128399 + }, + { + "epoch": 7.509455920861216, + "flow_matching_loss": 0.021422430500388147, + "grad_norm": 0.1859307885169983, + "learning_rate": 2.781995956825428e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006898481724783778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00306575030554086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016862713638693093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8901766121387482, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25810, + "weighted_flow_matching_loss": 0.021422430500388147, + "weighted_rkd_loss": 0.015010799374431371 + }, + { + "epoch": 7.512365434972359, + "flow_matching_loss": 0.019021824933588505, + "grad_norm": 0.18533557653427124, + "learning_rate": 2.781781295387123e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006793919578194618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029064284171909094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016494267620146275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8900947391986846, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25820, + "weighted_flow_matching_loss": 0.019021824933588505, + "weighted_rkd_loss": 0.01468147775158286 + }, + { + "epoch": 7.515274949083503, + "flow_matching_loss": 0.021694539301097392, + "grad_norm": 0.23669344186782837, + "learning_rate": 2.7815665366059884e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074057627469301225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003045736229978502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01785726174712181, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8900128424167633, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25830, + "weighted_flow_matching_loss": 0.021694539301097392, + "weighted_rkd_loss": 0.015893200039863588 + }, + { + "epoch": 7.518184463194647, + "flow_matching_loss": 0.02087471168488264, + "grad_norm": 0.19926834106445312, + "learning_rate": 2.781351680498332e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007006322918459773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002996798697859049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017009444255381822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.889930921792984, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25840, + "weighted_flow_matching_loss": 0.02087471168488264, + "weighted_rkd_loss": 0.015137217100709676 + }, + { + "epoch": 7.52109397730579, + "flow_matching_loss": 0.019173223339021207, + "grad_norm": 0.21108819544315338, + "learning_rate": 2.781136727080472e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006895855767652392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030434613348916174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016835172846913337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8898489713668823, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25850, + "weighted_flow_matching_loss": 0.019173223339021207, + "weighted_rkd_loss": 0.014980771020054818 + }, + { + "epoch": 7.524003491416933, + "flow_matching_loss": 0.018936616741120814, + "grad_norm": 0.27389654517173767, + "learning_rate": 2.7809216763687326e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007718811929225922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034350012661889195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018872625287622215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8897669851779938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25860, + "weighted_flow_matching_loss": 0.018936616741120814, + "weighted_rkd_loss": 0.0167922287248075 + }, + { + "epoch": 7.526913005528077, + "flow_matching_loss": 0.019470190349966287, + "grad_norm": 0.18606488406658173, + "learning_rate": 2.7807065283794446e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007027902361005545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036018275655806064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017657632101327182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8896849870681762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25870, + "weighted_flow_matching_loss": 0.019470190349966287, + "weighted_rkd_loss": 0.015709734056144953 + }, + { + "epoch": 7.52982251963922, + "flow_matching_loss": 0.019477045349776744, + "grad_norm": 0.19584375619888306, + "learning_rate": 2.7804912831289483e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006837358139455319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003139454312622547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016814170498400928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8896029472351075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25880, + "weighted_flow_matching_loss": 0.019477045349776744, + "weighted_rkd_loss": 0.014957908261567354 + }, + { + "epoch": 7.532732033750364, + "flow_matching_loss": 0.020209404267370702, + "grad_norm": 0.20600414276123047, + "learning_rate": 2.7802759406335893e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006760781677439809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002888274332508445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016409837640821934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8895208954811096, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25890, + "weighted_flow_matching_loss": 0.020209404267370702, + "weighted_rkd_loss": 0.01459687827154994 + }, + { + "epoch": 7.535641547861507, + "flow_matching_loss": 0.01885619182139635, + "grad_norm": 0.25984859466552734, + "learning_rate": 2.780060500909722e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007250661868602038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031279118033125997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017629235237836837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.889438807964325, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25900, + "weighted_flow_matching_loss": 0.01885619182139635, + "weighted_rkd_loss": 0.015680114179849623 + }, + { + "epoch": 7.538551061972651, + "flow_matching_loss": 0.019694967474788428, + "grad_norm": 0.2442576289176941, + "learning_rate": 2.779844963973707e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00724401636980474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032951726578176023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017783205211162566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8893566966056824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25910, + "weighted_flow_matching_loss": 0.019694967474788428, + "weighted_rkd_loss": 0.015815559308975935 + }, + { + "epoch": 7.541460576083794, + "flow_matching_loss": 0.020578479394316673, + "grad_norm": 0.22489941120147705, + "learning_rate": 2.779629329841913e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007322435732930899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031769031193107367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017821774631738663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8892745673656464, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25920, + "weighted_flow_matching_loss": 0.020578479394316673, + "weighted_rkd_loss": 0.01584843248128891 + }, + { + "epoch": 7.5443700901949375, + "flow_matching_loss": 0.018408212810754776, + "grad_norm": 0.2092534303665161, + "learning_rate": 2.7794135985307164e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007554219244048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003320340812206268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01842877930030227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8891923904418946, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25930, + "weighted_flow_matching_loss": 0.018408212810754776, + "weighted_rkd_loss": 0.016386752668768167 + }, + { + "epoch": 7.547279604306081, + "flow_matching_loss": 0.01982068121433258, + "grad_norm": 0.2508631646633148, + "learning_rate": 2.7791977700565007e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006228821445256471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026803431450389326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015137986093759537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8891102075576782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25940, + "weighted_flow_matching_loss": 0.01982068121433258, + "weighted_rkd_loss": 0.013459297083318233 + }, + { + "epoch": 7.550189118417224, + "flow_matching_loss": 0.021058076433837412, + "grad_norm": 0.19463419914245605, + "learning_rate": 2.7789818444356563e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006693597696721554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002876248606480658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016263443883508442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8890279948711395, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25950, + "weighted_flow_matching_loss": 0.021058076433837412, + "weighted_rkd_loss": 0.014458683598786593 + }, + { + "epoch": 7.553098632528368, + "flow_matching_loss": 0.0200118076056242, + "grad_norm": 0.20513944327831268, + "learning_rate": 2.778765821684581e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007015516841784119, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031681932043284176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017199226934462786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8889457583427429, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25960, + "weighted_flow_matching_loss": 0.0200118076056242, + "weighted_rkd_loss": 0.015289209131151438 + }, + { + "epoch": 7.5560081466395115, + "flow_matching_loss": 0.020959646627306937, + "grad_norm": 0.22251425683498383, + "learning_rate": 2.778549701819681e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006911077443510294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028499357867985964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01667209090664983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8888634741306305, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25970, + "weighted_flow_matching_loss": 0.020959646627306937, + "weighted_rkd_loss": 0.014819225668907166 + }, + { + "epoch": 7.558917660750654, + "flow_matching_loss": 0.017760434839874505, + "grad_norm": 0.19781945645809174, + "learning_rate": 2.778333484857369e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007496306160464883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003312920592725277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018305532820522784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8887811779975892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25980, + "weighted_flow_matching_loss": 0.017760434839874505, + "weighted_rkd_loss": 0.016269632149487734 + }, + { + "epoch": 7.561827174861798, + "flow_matching_loss": 0.021789848897606136, + "grad_norm": 0.1988188773393631, + "learning_rate": 2.7781171708140643e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006969462987035513, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003137637139298022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01707656281068921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8886988520622253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25990, + "weighted_flow_matching_loss": 0.021789848897606136, + "weighted_rkd_loss": 0.015175929106771947 + }, + { + "epoch": 7.564736688972942, + "flow_matching_loss": 0.020739755220711233, + "grad_norm": 0.25372743606567383, + "learning_rate": 2.777900759706196e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007679342431947589, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003564486908726394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01892317198216915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8886164963245392, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26000, + "weighted_flow_matching_loss": 0.020739755220711233, + "weighted_rkd_loss": 0.01681544501334429 + }, + { + "epoch": 7.5676462030840845, + "flow_matching_loss": 0.018412004876881837, + "grad_norm": 0.1944936215877533, + "learning_rate": 2.7776842515501985e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007375802518799901, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032781454268842936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018029750604182482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8885341227054596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26010, + "weighted_flow_matching_loss": 0.018412004876881837, + "weighted_rkd_loss": 0.016020058654248714 + }, + { + "epoch": 7.570555717195228, + "flow_matching_loss": 0.020287334360182284, + "grad_norm": 0.1851346641778946, + "learning_rate": 2.7774676463625135e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006287386594340205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025125087704509497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015087281912565231, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8884517192840576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26020, + "weighted_flow_matching_loss": 0.020287334360182284, + "weighted_rkd_loss": 0.013404339458793402 + }, + { + "epoch": 7.573465231306372, + "flow_matching_loss": 0.019551920238882303, + "grad_norm": 0.17185348272323608, + "learning_rate": 2.7772509441595914e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006283989595249295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002629193780012429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01519717313349247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8883692920207977, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26030, + "weighted_flow_matching_loss": 0.019551920238882303, + "weighted_rkd_loss": 0.013500733766704798 + }, + { + "epoch": 7.576374745417516, + "flow_matching_loss": 0.019515263941138984, + "grad_norm": 0.18642525374889374, + "learning_rate": 2.7770341449578898e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007350113615393639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003409487078897655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018109714426100255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.888286828994751, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26040, + "weighted_flow_matching_loss": 0.019515263941138984, + "weighted_rkd_loss": 0.016086602676659824 + }, + { + "epoch": 7.5792842595286585, + "flow_matching_loss": 0.017754416819661854, + "grad_norm": 0.209192156791687, + "learning_rate": 2.7768172487738718e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007486985623836517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032979037379845975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018271875008940695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8882043480873107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26050, + "weighted_flow_matching_loss": 0.017754416819661854, + "weighted_rkd_loss": 0.016229115426540375 + }, + { + "epoch": 7.582193773639802, + "flow_matching_loss": 0.020385641418397427, + "grad_norm": 0.225779727101326, + "learning_rate": 2.7766002556240106e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007292998675256968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003219409310258925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017805406544357538, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8881218373775482, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26060, + "weighted_flow_matching_loss": 0.020385641418397427, + "weighted_rkd_loss": 0.015813363809138538 + }, + { + "epoch": 7.585103287750946, + "flow_matching_loss": 0.020503556355834007, + "grad_norm": 0.2172376811504364, + "learning_rate": 2.7763831655247847e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007503118878230452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031737572979182007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018179994914680718, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8880392909049988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26070, + "weighted_flow_matching_loss": 0.020503556355834007, + "weighted_rkd_loss": 0.016144562605768444 + }, + { + "epoch": 7.588012801862089, + "flow_matching_loss": 0.02076407428830862, + "grad_norm": 0.2363288402557373, + "learning_rate": 2.7761659784926808e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006444559898227453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002746413997374475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015635533537715674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8879567325115204, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26080, + "weighted_flow_matching_loss": 0.02076407428830862, + "weighted_rkd_loss": 0.013883684948086738 + }, + { + "epoch": 7.590922315973232, + "flow_matching_loss": 0.019897293485701083, + "grad_norm": 0.21737965941429138, + "learning_rate": 2.7759486945441922e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007167059229686856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031534045585431157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017487523145973682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8878741443157196, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26090, + "weighted_flow_matching_loss": 0.019897293485701083, + "weighted_rkd_loss": 0.015526714734733105 + }, + { + "epoch": 7.593831830084376, + "flow_matching_loss": 0.018653584830462932, + "grad_norm": 0.19214631617069244, + "learning_rate": 2.7757313136958216e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007034925790503621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029933845973573623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01706323614344001, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8877915203571319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26100, + "weighted_flow_matching_loss": 0.018653584830462932, + "weighted_rkd_loss": 0.01514859301969409 + }, + { + "epoch": 7.59674134419552, + "flow_matching_loss": 0.018563408777117728, + "grad_norm": 0.17699331045150757, + "learning_rate": 2.7755138359640766e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006591521762311458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028927672654390337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01607581079006195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8877088844776153, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26110, + "weighted_flow_matching_loss": 0.018563408777117728, + "weighted_rkd_loss": 0.014270645659416914 + }, + { + "epoch": 7.599650858306663, + "flow_matching_loss": 0.02026739539578557, + "grad_norm": 0.23625068366527557, + "learning_rate": 2.7752962613654733e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00614655721001327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028049535350874065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015098067931830883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8876262068748474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26120, + "weighted_flow_matching_loss": 0.02026739539578557, + "weighted_rkd_loss": 0.013401432055979967 + }, + { + "epoch": 7.602560372417806, + "flow_matching_loss": 0.019139551650732757, + "grad_norm": 0.26651760935783386, + "learning_rate": 2.775078589916535e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072589096613228325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033103065565228463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017828125786036254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8875435173511506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26130, + "weighted_flow_matching_loss": 0.019139551650732757, + "weighted_rkd_loss": 0.015823188424110412 + }, + { + "epoch": 7.605469886528949, + "flow_matching_loss": 0.01935020536184311, + "grad_norm": 0.21067124605178833, + "learning_rate": 2.774860821633793e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006552693108096719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002847629296593368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015953015256673097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8874607920646668, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26140, + "weighted_flow_matching_loss": 0.01935020536184311, + "weighted_rkd_loss": 0.014157644286751748 + }, + { + "epoch": 7.608379400640093, + "flow_matching_loss": 0.019912843592464924, + "grad_norm": 0.20366613566875458, + "learning_rate": 2.7746429565337845e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007104508997872472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032281109131872656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017437129002064466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8873780369758606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26150, + "weighted_flow_matching_loss": 0.019912843592464924, + "weighted_rkd_loss": 0.015473332721740007 + }, + { + "epoch": 7.611288914751237, + "flow_matching_loss": 0.021714345179498197, + "grad_norm": 0.2073804885149002, + "learning_rate": 2.7744249946330557e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005969332065433264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025781141594052315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014516778383404016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.887295264005661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26160, + "weighted_flow_matching_loss": 0.021714345179498197, + "weighted_rkd_loss": 0.012880654074251652 + }, + { + "epoch": 7.61419842886238, + "flow_matching_loss": 0.02075784932821989, + "grad_norm": 0.21716046333312988, + "learning_rate": 2.7742069359481588e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005708820372819901, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024559684447012843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013873609062284232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.887212461233139, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26170, + "weighted_flow_matching_loss": 0.02075784932821989, + "weighted_rkd_loss": 0.012308868858963252 + }, + { + "epoch": 7.617107942973523, + "flow_matching_loss": 0.021218868717551232, + "grad_norm": 0.23167258501052856, + "learning_rate": 2.773988780495654e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006986594898626209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028905387269333007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01686372859403491, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8871296346187592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26180, + "weighted_flow_matching_loss": 0.021218868717551232, + "weighted_rkd_loss": 0.014960374310612678 + }, + { + "epoch": 7.620017457084667, + "flow_matching_loss": 0.019639690220355988, + "grad_norm": 0.20580121874809265, + "learning_rate": 2.7737705282921094e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006886856211349368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003068241919390857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01684195436537266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8870467841625214, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26190, + "weighted_flow_matching_loss": 0.019639690220355988, + "weighted_rkd_loss": 0.014939635805785657 + }, + { + "epoch": 7.6229269711958105, + "flow_matching_loss": 0.021014678943902253, + "grad_norm": 0.2079407423734665, + "learning_rate": 2.7735521793540994e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006198417907580733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002729526208713651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015126362070441246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8869638860225677, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26200, + "weighted_flow_matching_loss": 0.021014678943902253, + "weighted_rkd_loss": 0.013416560739278794 + }, + { + "epoch": 7.625836485306953, + "flow_matching_loss": 0.01851885989308357, + "grad_norm": 0.21305163204669952, + "learning_rate": 2.773333733698205e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006152512086555362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027231162413954733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015028140414506197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8868809819221497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26210, + "weighted_flow_matching_loss": 0.01851885989308357, + "weighted_rkd_loss": 0.013328169379383326 + }, + { + "epoch": 7.628745999418097, + "flow_matching_loss": 0.02086949087679386, + "grad_norm": 0.2481965273618698, + "learning_rate": 2.7731151913410182e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006615098752081394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002849967370275408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01608016462996602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8867980420589447, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26220, + "weighted_flow_matching_loss": 0.02086949087679386, + "weighted_rkd_loss": 0.014259829092770815 + }, + { + "epoch": 7.631655513529241, + "flow_matching_loss": 0.019991533365100622, + "grad_norm": 0.268252968788147, + "learning_rate": 2.7728965522991338e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007581556076183915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032281492138281466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018391260970383884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8867150843143463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26230, + "weighted_flow_matching_loss": 0.019991533365100622, + "weighted_rkd_loss": 0.016307791881263255 + }, + { + "epoch": 7.6345650276403845, + "flow_matching_loss": 0.01966134924441576, + "grad_norm": 0.2111676037311554, + "learning_rate": 2.7726778165891565e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006594089744612575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028133032261393963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016001482494175434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8866320967674255, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26240, + "weighted_flow_matching_loss": 0.01966134924441576, + "weighted_rkd_loss": 0.014187461510300637 + }, + { + "epoch": 7.637474541751527, + "flow_matching_loss": 0.019613177608698608, + "grad_norm": 0.18024003505706787, + "learning_rate": 2.772458984227698e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006878697546198964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030070837819948792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01676447857171297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8865490794181824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26250, + "weighted_flow_matching_loss": 0.019613177608698608, + "weighted_rkd_loss": 0.01486255256459117 + }, + { + "epoch": 7.640384055862671, + "flow_matching_loss": 0.02136280508711934, + "grad_norm": 0.16745910048484802, + "learning_rate": 2.7722400552313777e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00711356052197516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030435518361628055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017270672786980867, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8864660382270813, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26260, + "weighted_flow_matching_loss": 0.02136280508711934, + "weighted_rkd_loss": 0.015309869311749935 + }, + { + "epoch": 7.643293569973815, + "flow_matching_loss": 0.021000039763748646, + "grad_norm": 0.20388881862163544, + "learning_rate": 2.7720210296168204e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007540891924872994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003485740884207189, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018567524664103986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8863829672336578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26270, + "weighted_flow_matching_loss": 0.021000039763748646, + "weighted_rkd_loss": 0.01645790319889784 + }, + { + "epoch": 7.6462030840849575, + "flow_matching_loss": 0.021207277104258536, + "grad_norm": 0.1983148604631424, + "learning_rate": 2.7718019074006614e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006768238265067339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002949834009632468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01648631040006876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8862998723983765, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26280, + "weighted_flow_matching_loss": 0.021207277104258536, + "weighted_rkd_loss": 0.01461185421794653 + }, + { + "epoch": 7.649112598196101, + "flow_matching_loss": 0.0198508407920599, + "grad_norm": 0.23478646576404572, + "learning_rate": 2.77158268859954e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007761565595865249, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033978741150349377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018921005353331567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8862167537212372, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26290, + "weighted_flow_matching_loss": 0.0198508407920599, + "weighted_rkd_loss": 0.01676808260381222 + }, + { + "epoch": 7.652022112307245, + "flow_matching_loss": 0.021532865799963476, + "grad_norm": 0.23843419551849365, + "learning_rate": 2.7713633732301056e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00610965364612639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002819630037993193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015038937237113715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.88613361120224, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26300, + "weighted_flow_matching_loss": 0.021532865799963476, + "weighted_rkd_loss": 0.013326526805758476 + }, + { + "epoch": 7.654931626418388, + "flow_matching_loss": 0.02185765504837036, + "grad_norm": 0.22885555028915405, + "learning_rate": 2.7711439613090132e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007622556714341045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033001862931996582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018545299582183362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8860504269599915, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26310, + "weighted_flow_matching_loss": 0.02185765504837036, + "weighted_rkd_loss": 0.016432082373648883 + }, + { + "epoch": 7.6578411405295315, + "flow_matching_loss": 0.018754244968295097, + "grad_norm": 0.18798772990703583, + "learning_rate": 2.7709244528529253e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00679065128788352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028915251488797365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01647282764315605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8859672307968139, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26320, + "weighted_flow_matching_loss": 0.018754244968295097, + "weighted_rkd_loss": 0.014594344515353441 + }, + { + "epoch": 7.660750654640675, + "flow_matching_loss": 0.021099978405982257, + "grad_norm": 0.19596177339553833, + "learning_rate": 2.7707048478785125e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006828823452815413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027651467360556127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016422793548554183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8858839988708496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26330, + "weighted_flow_matching_loss": 0.021099978405982257, + "weighted_rkd_loss": 0.014548742678016424 + }, + { + "epoch": 7.663660168751818, + "flow_matching_loss": 0.020572294294834138, + "grad_norm": 0.21283362805843353, + "learning_rate": 2.7704851464024536e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007179294712841511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003298175008967519, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017656764574348928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8858007490634918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26340, + "weighted_flow_matching_loss": 0.020572294294834138, + "weighted_rkd_loss": 0.015640371944755316 + }, + { + "epoch": 7.666569682862962, + "flow_matching_loss": 0.019778193347156046, + "grad_norm": 0.1753387451171875, + "learning_rate": 2.7702653484414314e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007305800542235374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003374454751610756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017986056115478277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8857174634933471, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26350, + "weighted_flow_matching_loss": 0.019778193347156046, + "weighted_rkd_loss": 0.01593053378164768 + }, + { + "epoch": 7.669479196974105, + "flow_matching_loss": 0.019287065230309965, + "grad_norm": 0.2338278889656067, + "learning_rate": 2.7700454540121388e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006735448772087693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029204547172412276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01639135228469968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8856341660022735, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26360, + "weighted_flow_matching_loss": 0.019287065230309965, + "weighted_rkd_loss": 0.014516733027994633 + }, + { + "epoch": 7.672388711085249, + "flow_matching_loss": 0.019584254082292318, + "grad_norm": 0.2013867050409317, + "learning_rate": 2.769825463131276e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006764321215450763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029030935373157265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016431735921651126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8855508267879486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26370, + "weighted_flow_matching_loss": 0.019584254082292318, + "weighted_rkd_loss": 0.014551134686917067 + }, + { + "epoch": 7.675298225196392, + "flow_matching_loss": 0.019771358277648688, + "grad_norm": 0.25055786967277527, + "learning_rate": 2.7696053758155493e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006339962687343359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028674775967374444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015547403041273356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8854674696922302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26380, + "weighted_flow_matching_loss": 0.019771358277648688, + "weighted_rkd_loss": 0.01376669192686677 + }, + { + "epoch": 7.678207739307536, + "flow_matching_loss": 0.019889947213232517, + "grad_norm": 0.21920229494571686, + "learning_rate": 2.7693851920816725e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00575529714114964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025759241078048946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01408651852980256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8853840827941895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26390, + "weighted_flow_matching_loss": 0.019889947213232517, + "weighted_rkd_loss": 0.012471962347626686 + }, + { + "epoch": 7.681117253418679, + "flow_matching_loss": 0.020473104901611805, + "grad_norm": 0.21012304723262787, + "learning_rate": 2.7691649119463682e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007260880339890719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003007612540386617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01752937277778983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8853006839752198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26400, + "weighted_flow_matching_loss": 0.020473104901611805, + "weighted_rkd_loss": 0.015518775023519994 + }, + { + "epoch": 7.684026767529822, + "flow_matching_loss": 0.020614637434482573, + "grad_norm": 0.20728875696659088, + "learning_rate": 2.7689445354263642e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006413385737687349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027425649925135074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015569336246699095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8852172493934631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26410, + "weighted_flow_matching_loss": 0.020614637434482573, + "weighted_rkd_loss": 0.013782243710011244 + }, + { + "epoch": 7.686936281640966, + "flow_matching_loss": 0.02114332402125001, + "grad_norm": 0.221833735704422, + "learning_rate": 2.7687240625383975e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007355358405038715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033033784478902818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01801409525796771, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8851337671279907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26420, + "weighted_flow_matching_loss": 0.02114332402125001, + "weighted_rkd_loss": 0.015944871678948404 + }, + { + "epoch": 7.68984579575211, + "flow_matching_loss": 0.017808484472334386, + "grad_norm": 0.1622527837753296, + "learning_rate": 2.7685034932992107e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007099727308377624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031789373606443405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01737839188426733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8850502789020538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26430, + "weighted_flow_matching_loss": 0.017808484472334386, + "weighted_rkd_loss": 0.015380797442048788 + }, + { + "epoch": 7.692755309863253, + "flow_matching_loss": 0.0210052072070539, + "grad_norm": 0.24278420209884644, + "learning_rate": 2.7682828277255553e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006454122019931674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00296749013941735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015875734202563763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8849667608737946, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26440, + "weighted_flow_matching_loss": 0.0210052072070539, + "weighted_rkd_loss": 0.01404948504641652 + }, + { + "epoch": 7.695664823974396, + "flow_matching_loss": 0.020401838701218367, + "grad_norm": 0.24189160764217377, + "learning_rate": 2.7680620658341886e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007134820288047195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003229467221535742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01749910768121481, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8848832130432129, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26450, + "weighted_flow_matching_loss": 0.020401838701218367, + "weighted_rkd_loss": 0.015484665241092443 + }, + { + "epoch": 7.69857433808554, + "flow_matching_loss": 0.020519634149968623, + "grad_norm": 0.2103319764137268, + "learning_rate": 2.7678412076418772e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007015851233154536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032426118385046722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017274314351379872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8847996473312378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26460, + "weighted_flow_matching_loss": 0.020519634149968623, + "weighted_rkd_loss": 0.015284322947263718 + }, + { + "epoch": 7.7014838521966835, + "flow_matching_loss": 0.019495286419987677, + "grad_norm": 0.19886836409568787, + "learning_rate": 2.7676202531653924e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070876930840313435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003083950001746416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017259336169809104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8847160458564758, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26470, + "weighted_flow_matching_loss": 0.019495286419987677, + "weighted_rkd_loss": 0.015269599948078394 + }, + { + "epoch": 7.704393366307826, + "flow_matching_loss": 0.019989414513111113, + "grad_norm": 0.23984411358833313, + "learning_rate": 2.7673992024215152e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00704285972751677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033000034978613257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017385723255574703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8846324205398559, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26480, + "weighted_flow_matching_loss": 0.019989414513111113, + "weighted_rkd_loss": 0.015379932802170515 + }, + { + "epoch": 7.70730288041897, + "flow_matching_loss": 0.02101733470335603, + "grad_norm": 0.20473571121692657, + "learning_rate": 2.767178055427033e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006753361644223332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028801513370126488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016386874578893186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8845487713813782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26490, + "weighted_flow_matching_loss": 0.02101733470335603, + "weighted_rkd_loss": 0.014494992699474097 + }, + { + "epoch": 7.710212394530114, + "flow_matching_loss": 0.020575355645269154, + "grad_norm": 0.18044641613960266, + "learning_rate": 2.7669568121987394e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006246200017631054, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026893941685557365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015181794110685588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8844650983810425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26500, + "weighted_flow_matching_loss": 0.020575355645269154, + "weighted_rkd_loss": 0.013427780754864217 + }, + { + "epoch": 7.713121908641257, + "flow_matching_loss": 0.01949276914820075, + "grad_norm": 0.2269313484430313, + "learning_rate": 2.766735472753438e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006812628218904137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003030816768296063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01665607327595353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8843814015388489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26510, + "weighted_flow_matching_loss": 0.01949276914820075, + "weighted_rkd_loss": 0.01473030699416995 + }, + { + "epoch": 7.7160314227524, + "flow_matching_loss": 0.02008445281535387, + "grad_norm": 0.19302384555339813, + "learning_rate": 2.7665140371079364e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006974504329264164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029560341499745846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016905042715370654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8842976689338684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26520, + "weighted_flow_matching_loss": 0.02008445281535387, + "weighted_rkd_loss": 0.014949050731956959 + }, + { + "epoch": 7.718940936863544, + "flow_matching_loss": 0.02124138902872801, + "grad_norm": 0.21712709963321686, + "learning_rate": 2.7662925052790523e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006956893485039473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003044774476438761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01695856135338545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8842139124870301, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26530, + "weighted_flow_matching_loss": 0.02124138902872801, + "weighted_rkd_loss": 0.014994948729872703 + }, + { + "epoch": 7.721850450974687, + "flow_matching_loss": 0.019823879189789295, + "grad_norm": 0.23365472257137299, + "learning_rate": 2.7660708772836093e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007229732535779476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003137688781134784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017597154062241317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8841301381587983, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26540, + "weighted_flow_matching_loss": 0.019823879189789295, + "weighted_rkd_loss": 0.015558183938264847 + }, + { + "epoch": 7.7247599650858305, + "flow_matching_loss": 0.017937025614082815, + "grad_norm": 0.19282472133636475, + "learning_rate": 2.7658491531384383e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005916362535208464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002531944320071489, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014364669471979142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.884046322107315, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26550, + "weighted_flow_matching_loss": 0.017937025614082815, + "weighted_rkd_loss": 0.012699034344404937 + }, + { + "epoch": 7.727669479196974, + "flow_matching_loss": 0.01915489612147212, + "grad_norm": 0.20315898954868317, + "learning_rate": 2.7656273328603783e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006499472307041287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002775503299199045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01577444775030017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8839624941349029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26560, + "weighted_flow_matching_loss": 0.01915489612147212, + "weighted_rkd_loss": 0.013944053184241056 + }, + { + "epoch": 7.730578993308118, + "flow_matching_loss": 0.01951632434502244, + "grad_norm": 0.24905455112457275, + "learning_rate": 2.765405416466274e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006496050814166665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029130720533430575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01590517358854413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8838786244392395, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26570, + "weighted_flow_matching_loss": 0.01951632434502244, + "weighted_rkd_loss": 0.01405822141095996 + }, + { + "epoch": 7.733488507419261, + "flow_matching_loss": 0.018994152173399927, + "grad_norm": 0.19840528070926666, + "learning_rate": 2.76518340397298e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007295293966308236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030620758654549716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01765266377478838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8837947428226471, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26580, + "weighted_flow_matching_loss": 0.018994152173399927, + "weighted_rkd_loss": 0.015601341053843499 + }, + { + "epoch": 7.7363980215304045, + "flow_matching_loss": 0.021058071032166482, + "grad_norm": 0.20441550016403198, + "learning_rate": 2.7649612953973557e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007414845563471317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003263131808489561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01809282284229994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8837108314037323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26590, + "weighted_flow_matching_loss": 0.021058071032166482, + "weighted_rkd_loss": 0.015988859720528124 + }, + { + "epoch": 7.739307535641548, + "flow_matching_loss": 0.018790699541568756, + "grad_norm": 0.22037114202976227, + "learning_rate": 2.7647390907562692e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006481297127902508, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003041271888650954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016003866121172904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8836268901824951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26600, + "weighted_flow_matching_loss": 0.018790699541568756, + "weighted_rkd_loss": 0.014141433406621218 + }, + { + "epoch": 7.742217049752691, + "flow_matching_loss": 0.02088112160563469, + "grad_norm": 0.25458407402038574, + "learning_rate": 2.7645167900665953e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006214358075521886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002804285555612296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015233001764863729, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8835429251194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26610, + "weighted_flow_matching_loss": 0.02088112160563469, + "weighted_rkd_loss": 0.013459027232602238 + }, + { + "epoch": 7.745126563863835, + "flow_matching_loss": 0.01988449990749359, + "grad_norm": 0.22965067625045776, + "learning_rate": 2.764294393345216e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006051764171570539, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002932459395378828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015035987924784422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8834589421749115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26620, + "weighted_flow_matching_loss": 0.01988449990749359, + "weighted_rkd_loss": 0.013283708970993758 + }, + { + "epoch": 7.748036077974978, + "flow_matching_loss": 0.022473756782710554, + "grad_norm": 0.16214750707149506, + "learning_rate": 2.7640719006090214e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006550211412832141, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002847310923971236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015947734005749226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8833748996257782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26630, + "weighted_flow_matching_loss": 0.022473756782710554, + "weighted_rkd_loss": 0.01408782796934247 + }, + { + "epoch": 7.750945592086122, + "flow_matching_loss": 0.019121474120765924, + "grad_norm": 0.22814087569713593, + "learning_rate": 2.7638493118749083e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006732541555538774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029088825918734073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016373965423554182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8832908570766449, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26640, + "weighted_flow_matching_loss": 0.019121474120765924, + "weighted_rkd_loss": 0.014462944865226746 + }, + { + "epoch": 7.753855106197265, + "flow_matching_loss": 0.021032383665442467, + "grad_norm": 0.21564394235610962, + "learning_rate": 2.7636266271597806e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006938615348190069, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003032717411406338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016909947991371153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8832068085670471, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26650, + "weighted_flow_matching_loss": 0.021032383665442467, + "weighted_rkd_loss": 0.014934979751706124 + }, + { + "epoch": 7.756764620308409, + "flow_matching_loss": 0.019695131853222848, + "grad_norm": 0.21358944475650787, + "learning_rate": 2.7634038464805495e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006619694409891963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029936418402940037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016233030706644058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8831227004528046, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26660, + "weighted_flow_matching_loss": 0.019695131853222848, + "weighted_rkd_loss": 0.014335721638053656 + }, + { + "epoch": 7.7596741344195515, + "flow_matching_loss": 0.020278369169682266, + "grad_norm": 0.24490003287792206, + "learning_rate": 2.7631809698541336e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006919349171221256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032363303238525988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017075028736144304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8830385804176331, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26670, + "weighted_flow_matching_loss": 0.020278369169682266, + "weighted_rkd_loss": 0.015077907871454955 + }, + { + "epoch": 7.762583648530695, + "flow_matching_loss": 0.020176061242818833, + "grad_norm": 0.2194359302520752, + "learning_rate": 2.76295799729746e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006749904621392489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028193244128488005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01631913362070918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8829544305801391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26680, + "weighted_flow_matching_loss": 0.020176061242818833, + "weighted_rkd_loss": 0.014409103896468878 + }, + { + "epoch": 7.765493162641839, + "flow_matching_loss": 0.01947484165430069, + "grad_norm": 0.18390586972236633, + "learning_rate": 2.7627349288274612e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006591809680685401, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027603803435340525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01594399968162179, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8828702569007874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26690, + "weighted_flow_matching_loss": 0.01947484165430069, + "weighted_rkd_loss": 0.01407649489119649 + }, + { + "epoch": 7.768402676752983, + "flow_matching_loss": 0.01835456006228924, + "grad_norm": 0.22504524886608124, + "learning_rate": 2.7625117644610776e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006360320188105106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028570502530783416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01557769076898694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8827860534191132, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26700, + "weighted_flow_matching_loss": 0.01835456006228924, + "weighted_rkd_loss": 0.01375176403671503 + }, + { + "epoch": 7.771312190864125, + "flow_matching_loss": 0.018979125563055275, + "grad_norm": 0.23955148458480835, + "learning_rate": 2.7622885042152576e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006988217541947961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031542716780677436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017130706645548343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8827018320560456, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26710, + "weighted_flow_matching_loss": 0.018979125563055275, + "weighted_rkd_loss": 0.015121296513825655 + }, + { + "epoch": 7.774221704975269, + "flow_matching_loss": 0.01920549087226391, + "grad_norm": 0.20302870869636536, + "learning_rate": 2.762065148106956e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007078257389366627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029767209431156515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017133235931396484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8826175689697265, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26720, + "weighted_flow_matching_loss": 0.01920549087226391, + "weighted_rkd_loss": 0.01512208953499794 + }, + { + "epoch": 7.777131219086413, + "flow_matching_loss": 0.021175054274499415, + "grad_norm": 0.2445734441280365, + "learning_rate": 2.761841696153135e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006187749514356256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002759402059018612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015134900994598865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8825332939624786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26730, + "weighted_flow_matching_loss": 0.021175054274499415, + "weighted_rkd_loss": 0.01335704354569316 + }, + { + "epoch": 7.780040733197556, + "flow_matching_loss": 0.016969445254653694, + "grad_norm": 0.19362182915210724, + "learning_rate": 2.761618148370765e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006013643881306052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025622367858886717, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014589524548500776, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8824489831924438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26740, + "weighted_flow_matching_loss": 0.016969445254653694, + "weighted_rkd_loss": 0.012874510046094656 + }, + { + "epoch": 7.782950247308699, + "flow_matching_loss": 0.02127635069191456, + "grad_norm": 0.21158058941364288, + "learning_rate": 2.7613945047768226e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007001634407788515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029821591218933464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01698542796075344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8823646605014801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26750, + "weighted_flow_matching_loss": 0.02127635069191456, + "weighted_rkd_loss": 0.014987343922257423 + }, + { + "epoch": 7.785859761419843, + "flow_matching_loss": 0.020375197194516657, + "grad_norm": 0.2144315093755722, + "learning_rate": 2.761170765388292e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006962406635284424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031586804427206515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017083493806421757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8822802901268005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26760, + "weighted_flow_matching_loss": 0.020375197194516657, + "weighted_rkd_loss": 0.015072455257177353 + }, + { + "epoch": 7.788769275530987, + "flow_matching_loss": 0.02049261350184679, + "grad_norm": 0.18847684562206268, + "learning_rate": 2.760946930222165e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006722527788951993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003087911824695766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0165329672396183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8821959018707275, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26770, + "weighted_flow_matching_loss": 0.02049261350184679, + "weighted_rkd_loss": 0.014585324842482806 + }, + { + "epoch": 7.79167878964213, + "flow_matching_loss": 0.020364274457097053, + "grad_norm": 0.26375097036361694, + "learning_rate": 2.7607229992954394e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006724111223593354, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030066583771258593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01645488105714321, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8821114957332611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26780, + "weighted_flow_matching_loss": 0.020364274457097053, + "weighted_rkd_loss": 0.014515059255063533 + }, + { + "epoch": 7.794588303753273, + "flow_matching_loss": 0.02120428401976824, + "grad_norm": 0.19478991627693176, + "learning_rate": 2.7604989726251227e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006157359015196562, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002665591088589281, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014980309270322322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8820270597934723, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26790, + "weighted_flow_matching_loss": 0.02120428401976824, + "weighted_rkd_loss": 0.013213034346699714 + }, + { + "epoch": 7.797497817864417, + "flow_matching_loss": 0.01927999798208475, + "grad_norm": 0.18425889313220978, + "learning_rate": 2.7602748502282275e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064198358450084925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029106631642207502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01575033478438854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8819425940513611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26800, + "weighted_flow_matching_loss": 0.01927999798208475, + "weighted_rkd_loss": 0.01389092393219471 + }, + { + "epoch": 7.80040733197556, + "flow_matching_loss": 0.021126158721745013, + "grad_norm": 0.19675521552562714, + "learning_rate": 2.7600506321217744e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006879616808146238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003223983175121248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016983216814696788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.881858104467392, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26810, + "weighted_flow_matching_loss": 0.021126158721745013, + "weighted_rkd_loss": 0.014976786449551582 + }, + { + "epoch": 7.8033168460867035, + "flow_matching_loss": 0.020376593060791492, + "grad_norm": 0.2037823498249054, + "learning_rate": 2.7598263183227916e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006989227281883359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031640775501728057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017142532300204037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8817735850811005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26820, + "weighted_flow_matching_loss": 0.020376593060791492, + "weighted_rkd_loss": 0.015115902572870255 + }, + { + "epoch": 7.806226360197847, + "flow_matching_loss": 0.020395622961223124, + "grad_norm": 0.2485721856355667, + "learning_rate": 2.759601908848314e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00647465861402452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029213206376880406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015870638005435467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.881689041852951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26830, + "weighted_flow_matching_loss": 0.020395622961223124, + "weighted_rkd_loss": 0.013992962054908275 + }, + { + "epoch": 7.80913587430899, + "flow_matching_loss": 0.01930918488651514, + "grad_norm": 0.20140135288238525, + "learning_rate": 2.759377403715384e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006906373659148812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030746629694476725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01688741035759449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8816044747829437, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26840, + "weighted_flow_matching_loss": 0.01930918488651514, + "weighted_rkd_loss": 0.01488800598308444 + }, + { + "epoch": 7.812045388420134, + "flow_matching_loss": 0.01864925315603614, + "grad_norm": 0.19129644334316254, + "learning_rate": 2.759152802941051e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006536703230813146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029300982132554054, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016003504768013953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.881519877910614, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26850, + "weighted_flow_matching_loss": 0.01864925315603614, + "weighted_rkd_loss": 0.01410741712898016 + }, + { + "epoch": 7.8149549025312774, + "flow_matching_loss": 0.019100509863346816, + "grad_norm": 0.20352329313755035, + "learning_rate": 2.758928106542373e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068470426369458435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029539896408095957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016648075077682734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8814352750778198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26860, + "weighted_flow_matching_loss": 0.019100509863346816, + "weighted_rkd_loss": 0.014674175158143043 + }, + { + "epoch": 7.81786441664242, + "flow_matching_loss": 0.020859198924154044, + "grad_norm": 0.2039870321750641, + "learning_rate": 2.7587033145364127e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006183349434286356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002740839833859354, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015107538644224405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8813506245613099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26870, + "weighted_flow_matching_loss": 0.020859198924154044, + "weighted_rkd_loss": 0.013315016124397516 + }, + { + "epoch": 7.820773930753564, + "flow_matching_loss": 0.020069557055830956, + "grad_norm": 0.27398309111595154, + "learning_rate": 2.7584784269402427e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007353033078834415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003183287219144404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017889353353530167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8812659382820129, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26880, + "weighted_flow_matching_loss": 0.020069557055830956, + "weighted_rkd_loss": 0.015765238367021085 + }, + { + "epoch": 7.823683444864708, + "flow_matching_loss": 0.019382782746106386, + "grad_norm": 0.23569869995117188, + "learning_rate": 2.7582534437709414e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007032500999048352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031112533295527102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017176255211234094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8811812400817871, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26890, + "weighted_flow_matching_loss": 0.019382782746106386, + "weighted_rkd_loss": 0.015135375969111919 + }, + { + "epoch": 7.826592958975851, + "flow_matching_loss": 0.01992348683997989, + "grad_norm": 0.20737522840499878, + "learning_rate": 2.758028365045595e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007774301711469889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003246724512428045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018795327842235567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8810965120792389, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26900, + "weighted_flow_matching_loss": 0.01992348683997989, + "weighted_rkd_loss": 0.016560465935617684 + }, + { + "epoch": 7.829502473086994, + "flow_matching_loss": 0.019875619560480118, + "grad_norm": 0.3282453715801239, + "learning_rate": 2.7578031907812966e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006735070887953043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031395241850987075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01660966584458947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8810117661952972, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26910, + "weighted_flow_matching_loss": 0.019875619560480118, + "weighted_rkd_loss": 0.014633250236511231 + }, + { + "epoch": 7.832411987198138, + "flow_matching_loss": 0.020077122934162618, + "grad_norm": 0.2303827553987503, + "learning_rate": 2.7575779209951465e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006776126055046916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002908137859776616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01646038992330432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8809269785881042, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26920, + "weighted_flow_matching_loss": 0.020077122934162618, + "weighted_rkd_loss": 0.014500358793884515 + }, + { + "epoch": 7.835321501309282, + "flow_matching_loss": 0.022293326444923878, + "grad_norm": 0.2218470275402069, + "learning_rate": 2.757352555704253e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00699752401560545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003050830774009228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017045878898352386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8808421790599823, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26930, + "weighted_flow_matching_loss": 0.022293326444923878, + "weighted_rkd_loss": 0.015014729555696249 + }, + { + "epoch": 7.838231015420424, + "flow_matching_loss": 0.020253402926027776, + "grad_norm": 0.23723319172859192, + "learning_rate": 2.7571270949257297e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006812996184453368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030425544129684566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016668547037988903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.880757349729538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26940, + "weighted_flow_matching_loss": 0.020253402926027776, + "weighted_rkd_loss": 0.014680892135947943 + }, + { + "epoch": 7.841140529531568, + "flow_matching_loss": 0.018573592696338893, + "grad_norm": 0.22069671750068665, + "learning_rate": 2.7569015386767008e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006357470247894526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002661693829577416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015376634430140257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8806724905967712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26950, + "weighted_flow_matching_loss": 0.018573592696338893, + "weighted_rkd_loss": 0.013541788328438997 + }, + { + "epoch": 7.844050043642712, + "flow_matching_loss": 0.022867992334067822, + "grad_norm": 0.25159692764282227, + "learning_rate": 2.7566758869742944e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007159325294196606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003215344250202179, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01753399483859539, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8805876135826111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26960, + "weighted_flow_matching_loss": 0.022867992334067822, + "weighted_rkd_loss": 0.015440214797854424 + }, + { + "epoch": 7.846959557753856, + "flow_matching_loss": 0.019231258891522884, + "grad_norm": 0.19221772253513336, + "learning_rate": 2.756450139835648e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006064268667250872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027751185931265353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014903656113892794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8805027008056641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26970, + "weighted_flow_matching_loss": 0.019231258891522884, + "weighted_rkd_loss": 0.013122720550745726 + }, + { + "epoch": 7.849869071864998, + "flow_matching_loss": 0.01990563329309225, + "grad_norm": 0.18079277873039246, + "learning_rate": 2.7562242972779054e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006818237667903304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029038371867500246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01654031276702881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8804177641868591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26980, + "weighted_flow_matching_loss": 0.01990563329309225, + "weighted_rkd_loss": 0.014562381617724895 + }, + { + "epoch": 7.852778585976142, + "flow_matching_loss": 0.02106198500841856, + "grad_norm": 0.2376967966556549, + "learning_rate": 2.755998359318217e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007549103209748864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034956952091306447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01859390139579773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8803328096866607, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26990, + "weighted_flow_matching_loss": 0.02106198500841856, + "weighted_rkd_loss": 0.016368778981268407 + }, + { + "epoch": 7.855688100087286, + "flow_matching_loss": 0.019655567407608033, + "grad_norm": 0.1986340880393982, + "learning_rate": 2.755772325973743e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005823278520256281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002501532924361527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014148089848458768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.88024782538414, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27000, + "weighted_flow_matching_loss": 0.019655567407608033, + "weighted_rkd_loss": 0.012453806307166814 + }, + { + "epoch": 7.858597614198429, + "flow_matching_loss": 0.02080685682594776, + "grad_norm": 0.2521944046020508, + "learning_rate": 2.755546197261648e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006958832498639822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030866799759678544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017004344891756774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8801628112792969, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27010, + "weighted_flow_matching_loss": 0.02080685682594776, + "weighted_rkd_loss": 0.014966607838869096 + }, + { + "epoch": 7.861507128309572, + "flow_matching_loss": 0.01902599148452282, + "grad_norm": 0.19769054651260376, + "learning_rate": 2.7553199731991046e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006246518203988671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025747679639607666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015067804232239723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8800777733325958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27020, + "weighted_flow_matching_loss": 0.01902599148452282, + "weighted_rkd_loss": 0.013260847143828868 + }, + { + "epoch": 7.864416642420716, + "flow_matching_loss": 0.02112683802843094, + "grad_norm": 0.1732524037361145, + "learning_rate": 2.7550936538032942e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064841752871871, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027020119363442064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015670362301170825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8799927055835723, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27030, + "weighted_flow_matching_loss": 0.02112683802843094, + "weighted_rkd_loss": 0.013789795618504287 + }, + { + "epoch": 7.867326156531859, + "flow_matching_loss": 0.019737811293452978, + "grad_norm": 0.199121356010437, + "learning_rate": 2.7548672390914036e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005923819215968251, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002447931794449687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014295570272952317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8799076199531555, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27040, + "weighted_flow_matching_loss": 0.019737811293452978, + "weighted_rkd_loss": 0.012578760366886854 + }, + { + "epoch": 7.870235670643003, + "flow_matching_loss": 0.018798320647329093, + "grad_norm": 0.1895212084054947, + "learning_rate": 2.7546407290806272e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00615251618437469, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002562292804941535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014867324940860271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8798224985599518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27050, + "weighted_flow_matching_loss": 0.018798320647329093, + "weighted_rkd_loss": 0.013080627005547286 + }, + { + "epoch": 7.873145184754146, + "flow_matching_loss": 0.018602807354182006, + "grad_norm": 0.21644777059555054, + "learning_rate": 2.754414123788168e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00647709690965712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028144166339188815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015768610499799253, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8797373592853546, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27060, + "weighted_flow_matching_loss": 0.018602807354182006, + "weighted_rkd_loss": 0.013872232474386691 + }, + { + "epoch": 7.876054698865289, + "flow_matching_loss": 0.02010271567851305, + "grad_norm": 0.16314014792442322, + "learning_rate": 2.7541874232312342e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006985112885013222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003101336129475385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017071561980992557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8796521842479705, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27070, + "weighted_flow_matching_loss": 0.02010271567851305, + "weighted_rkd_loss": 0.015017021726816893 + }, + { + "epoch": 7.878964212976433, + "flow_matching_loss": 0.019672859087586404, + "grad_norm": 0.18953253328800201, + "learning_rate": 2.7539606274270427e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006695042410865426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003013575938530266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01640366083011031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8795669972896576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27080, + "weighted_flow_matching_loss": 0.019672859087586404, + "weighted_rkd_loss": 0.014428134448826313 + }, + { + "epoch": 7.8818737270875765, + "flow_matching_loss": 0.02113109566271305, + "grad_norm": 0.20796161890029907, + "learning_rate": 2.7537337363928166e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007094301423057914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031570857390761375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01734568877145648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8794817626476288, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27090, + "weighted_flow_matching_loss": 0.02113109566271305, + "weighted_rkd_loss": 0.015255250222980975 + }, + { + "epoch": 7.88478324119872, + "flow_matching_loss": 0.020312292501330376, + "grad_norm": 0.20439431071281433, + "learning_rate": 2.7535067501457877e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006533617246896028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00280488186981529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01587211648002267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8793965280056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27100, + "weighted_flow_matching_loss": 0.020312292501330376, + "weighted_rkd_loss": 0.013957873173058032 + }, + { + "epoch": 7.887692755309863, + "flow_matching_loss": 0.019675127882510423, + "grad_norm": 0.2525406777858734, + "learning_rate": 2.7532796687031928e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007006528694182635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003299847012385726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017312904261052607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8793112516403199, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27110, + "weighted_flow_matching_loss": 0.019675127882510423, + "weighted_rkd_loss": 0.015223448816686868 + }, + { + "epoch": 7.890602269421007, + "flow_matching_loss": 0.018288374412804842, + "grad_norm": 0.19949635863304138, + "learning_rate": 2.7530524920822784e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006128611415624619, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024849264184013007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014742149226367473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8792259573936463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27120, + "weighted_flow_matching_loss": 0.018288374412804842, + "weighted_rkd_loss": 0.012961640581488609 + }, + { + "epoch": 7.89351178353215, + "flow_matching_loss": 0.02057018717750907, + "grad_norm": 0.20579232275485992, + "learning_rate": 2.752825220300297e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006346550025045872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028482871712185442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015541387163102627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8791406214237213, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27130, + "weighted_flow_matching_loss": 0.02057018717750907, + "weighted_rkd_loss": 0.01366309830918908 + }, + { + "epoch": 7.896421297643293, + "flow_matching_loss": 0.01940315030515194, + "grad_norm": 0.18616801500320435, + "learning_rate": 2.7525978533745073e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006449741125106811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002698607207275927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015598089341074229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.879055279493332, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27140, + "weighted_flow_matching_loss": 0.01940315030515194, + "weighted_rkd_loss": 0.01371156582608819 + }, + { + "epoch": 7.899330811754437, + "flow_matching_loss": 0.021438286826014518, + "grad_norm": 0.22122831642627716, + "learning_rate": 2.7523703913221776e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070655122864991425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00317384812515229, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01730487300083041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8789699018001557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27150, + "weighted_flow_matching_loss": 0.021438286826014518, + "weighted_rkd_loss": 0.015210431441664696 + }, + { + "epoch": 7.902240325865581, + "flow_matching_loss": 0.018332111835479736, + "grad_norm": 0.1948995292186737, + "learning_rate": 2.7521428341605813e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006771867536008358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030162906856276094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016560025699436664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.878884494304657, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27160, + "weighted_flow_matching_loss": 0.018332111835479736, + "weighted_rkd_loss": 0.01455439766868949 + }, + { + "epoch": 7.905149839976724, + "flow_matching_loss": 0.019364650174975396, + "grad_norm": 0.21392472088336945, + "learning_rate": 2.7519151819070003e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006126230768859387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002582729794085026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014835191424936056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8787990748882294, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27170, + "weighted_flow_matching_loss": 0.019364650174975396, + "weighted_rkd_loss": 0.013037119153887033 + }, + { + "epoch": 7.908059354087867, + "flow_matching_loss": 0.020112835988402366, + "grad_norm": 0.23586593568325043, + "learning_rate": 2.7516874345787234e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00668805530294776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002883869572542608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016259980015456678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8787136137485504, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27180, + "weighted_flow_matching_loss": 0.020112835988402366, + "weighted_rkd_loss": 0.014287886768579483 + }, + { + "epoch": 7.910968868199011, + "flow_matching_loss": 0.019231380708515645, + "grad_norm": 0.18601273000240326, + "learning_rate": 2.7514595921930455e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006406179303303361, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027809415478259325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015593299921602011, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.878628134727478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27190, + "weighted_flow_matching_loss": 0.019231380708515645, + "weighted_rkd_loss": 0.01370074050500989 + }, + { + "epoch": 7.913878382310154, + "flow_matching_loss": 0.021075196750462055, + "grad_norm": 0.201370969414711, + "learning_rate": 2.7512316547672714e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006977211544290185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031666285824030636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01712105143815279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8785426318645477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27200, + "weighted_flow_matching_loss": 0.021075196750462055, + "weighted_rkd_loss": 0.01504160026088357 + }, + { + "epoch": 7.916787896421297, + "flow_matching_loss": 0.021377057023346423, + "grad_norm": 0.2215641885995865, + "learning_rate": 2.7510036223187098e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007035475084558129, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032075623050332068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017278512381017207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.878457099199295, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27210, + "weighted_flow_matching_loss": 0.021377057023346423, + "weighted_rkd_loss": 0.015178432501852513 + }, + { + "epoch": 7.919697410532441, + "flow_matching_loss": 0.01838543312624097, + "grad_norm": 0.1913525015115738, + "learning_rate": 2.7507754948646796e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006618220591917634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028070216299965977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0160434627905488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.87837153673172, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27220, + "weighted_flow_matching_loss": 0.01838543312624097, + "weighted_rkd_loss": 0.014092115312814712 + }, + { + "epoch": 7.922606924643585, + "flow_matching_loss": 0.019643554743379355, + "grad_norm": 0.16960008442401886, + "learning_rate": 2.7505472724225048e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00681699407286942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028208899544551968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01645487817004323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.878285950422287, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27230, + "weighted_flow_matching_loss": 0.019643554743379355, + "weighted_rkd_loss": 0.014452100079506636 + }, + { + "epoch": 7.925516438754728, + "flow_matching_loss": 0.018209200445562602, + "grad_norm": 0.1999446153640747, + "learning_rate": 2.7503189550095174e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006861012149602175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030222744913771748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016744298581033944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8782003402709961, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27240, + "weighted_flow_matching_loss": 0.018209200445562602, + "weighted_rkd_loss": 0.01470481911674142 + }, + { + "epoch": 7.928425952865871, + "flow_matching_loss": 0.019816221948713065, + "grad_norm": 0.16950668394565582, + "learning_rate": 2.750090542643057e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006149777490645647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027545962249860166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015054151136428117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8781147062778473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27250, + "weighted_flow_matching_loss": 0.019816221948713065, + "weighted_rkd_loss": 0.013219262752681971 + }, + { + "epoch": 7.931335466977015, + "flow_matching_loss": 0.01864370396360755, + "grad_norm": 0.2151743769645691, + "learning_rate": 2.7498620353404696e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006906169140711427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030004347674548624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016812773048877717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8780290484428406, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27260, + "weighted_flow_matching_loss": 0.01864370396360755, + "weighted_rkd_loss": 0.01476211529225111 + }, + { + "epoch": 7.934244981088158, + "flow_matching_loss": 0.021543245390057565, + "grad_norm": 0.22629022598266602, + "learning_rate": 2.7496334331191088e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007068552961573005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030219344887882473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01715904027223587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.877943354845047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27270, + "weighted_flow_matching_loss": 0.021543245390057565, + "weighted_rkd_loss": 0.015064662136137486 + }, + { + "epoch": 7.937154495199302, + "flow_matching_loss": 0.020232477132230996, + "grad_norm": 0.24082015454769135, + "learning_rate": 2.7494047359963358e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006563521455973387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029723744723014534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01609941739588976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.87785764336586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27280, + "weighted_flow_matching_loss": 0.020232477132230996, + "weighted_rkd_loss": 0.01413298286497593 + }, + { + "epoch": 7.940064009310445, + "flow_matching_loss": 0.019125460274517538, + "grad_norm": 0.22925889492034912, + "learning_rate": 2.7491759439895185e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006529149133712053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002635936951264739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015694235544651748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8777718961238861, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27290, + "weighted_flow_matching_loss": 0.019125460274517538, + "weighted_rkd_loss": 0.013775966037064791 + }, + { + "epoch": 7.942973523421589, + "flow_matching_loss": 0.02031026892364025, + "grad_norm": 0.2106458842754364, + "learning_rate": 2.748947057116032e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007126229582354426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003102607256732881, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01735506607219577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8776861369609833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27300, + "weighted_flow_matching_loss": 0.02031026892364025, + "weighted_rkd_loss": 0.015232317242771387 + }, + { + "epoch": 7.945883037532732, + "flow_matching_loss": 0.020940948836505414, + "grad_norm": 0.1853950470685959, + "learning_rate": 2.7487180753932587e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007169695105403662, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034570338437333705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017796423844993116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.877600347995758, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27310, + "weighted_flow_matching_loss": 0.020940948836505414, + "weighted_rkd_loss": 0.015618170239031315 + }, + { + "epoch": 7.9487925516438755, + "flow_matching_loss": 0.018922340031713246, + "grad_norm": 0.2049700915813446, + "learning_rate": 2.7484889988385885e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065354312537238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029566533863544463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01602751584723592, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8775145411491394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27320, + "weighted_flow_matching_loss": 0.018922340031713246, + "weighted_rkd_loss": 0.014064408559352159 + }, + { + "epoch": 7.951702065755019, + "flow_matching_loss": 0.01850172784179449, + "grad_norm": 0.22521738708019257, + "learning_rate": 2.7482598274694185e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006294262362644076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027348308358341455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015323355514556169, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8774286866188049, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27330, + "weighted_flow_matching_loss": 0.01850172784179449, + "weighted_rkd_loss": 0.01344514824450016 + }, + { + "epoch": 7.954611579866162, + "flow_matching_loss": 0.01996447630226612, + "grad_norm": 0.33730265498161316, + "learning_rate": 2.748030561303152e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006540166586637497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029125717002898453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015992905013263226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.877342814207077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27340, + "weighted_flow_matching_loss": 0.01996447630226612, + "weighted_rkd_loss": 0.01403127396479249 + }, + { + "epoch": 7.957521093977306, + "flow_matching_loss": 0.020392895862460136, + "grad_norm": 0.2071230560541153, + "learning_rate": 2.747801200357201e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007143423100933433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003136446280404925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017423292528837918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8772569119930267, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27350, + "weighted_flow_matching_loss": 0.020392895862460136, + "weighted_rkd_loss": 0.015284714754670858 + }, + { + "epoch": 7.9604306080884495, + "flow_matching_loss": 0.020259991474449634, + "grad_norm": 0.19416683912277222, + "learning_rate": 2.747571744648983e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062024723272770645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027102955617010595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015115240402519704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8771709978580475, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27360, + "weighted_flow_matching_loss": 0.020259991474449634, + "weighted_rkd_loss": 0.013258637953549623 + }, + { + "epoch": 7.963340122199592, + "flow_matching_loss": 0.017459698021411896, + "grad_norm": 0.1931561976671219, + "learning_rate": 2.747342194195925e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007344435388222337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031943015521392224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01788317197933793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8770850479602814, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27370, + "weighted_flow_matching_loss": 0.017459698021411896, + "weighted_rkd_loss": 0.01568507757037878 + }, + { + "epoch": 7.966249636310736, + "flow_matching_loss": 0.02024609884247184, + "grad_norm": 0.18203724920749664, + "learning_rate": 2.747112549015459e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007081709243357182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003095422452315688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01725884098559618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8769990801811218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27380, + "weighted_flow_matching_loss": 0.02024609884247184, + "weighted_rkd_loss": 0.015136019792407751 + }, + { + "epoch": 7.96915915042188, + "flow_matching_loss": 0.02024766579270363, + "grad_norm": 0.18455955386161804, + "learning_rate": 2.7468828091250243e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006512659369036555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028534503420814873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015878768917173148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8769130766391754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27390, + "weighted_flow_matching_loss": 0.02024766579270363, + "weighted_rkd_loss": 0.01392431315034628 + }, + { + "epoch": 7.9720686645330225, + "flow_matching_loss": 0.019831240829080343, + "grad_norm": 0.20346792042255402, + "learning_rate": 2.7466529745420698e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006408629380166531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00277694349642843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01559420209378004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8768270492553711, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27400, + "weighted_flow_matching_loss": 0.019831240829080343, + "weighted_rkd_loss": 0.013673442229628564 + }, + { + "epoch": 7.974978178644166, + "flow_matching_loss": 0.020313138887286187, + "grad_norm": 0.23699896037578583, + "learning_rate": 2.7464230452840486e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007330938102677464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030831692973151805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01774504529312253, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8767410039901733, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27410, + "weighted_flow_matching_loss": 0.020313138887286187, + "weighted_rkd_loss": 0.015557759534567594 + }, + { + "epoch": 7.97788769275531, + "flow_matching_loss": 0.02214568927884102, + "grad_norm": 0.21439149975776672, + "learning_rate": 2.7461930213684228e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00718633527867496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003155696182511747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01752836676314473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8766549289226532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27420, + "weighted_flow_matching_loss": 0.02214568927884102, + "weighted_rkd_loss": 0.015366325061768294 + }, + { + "epoch": 7.980797206866454, + "flow_matching_loss": 0.01884512146934867, + "grad_norm": 0.21356715261936188, + "learning_rate": 2.7459629028126613e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00694138091057539, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031649710843339564, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017047732975333928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8765688240528107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27430, + "weighted_flow_matching_loss": 0.01884512146934867, + "weighted_rkd_loss": 0.01494352389127016 + }, + { + "epoch": 7.9837067209775965, + "flow_matching_loss": 0.019287934713065624, + "grad_norm": 0.20270544290542603, + "learning_rate": 2.7457326896342404e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069278541021049024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033041117247194053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017159819975495337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8764826953411102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27440, + "weighted_flow_matching_loss": 0.019287934713065624, + "weighted_rkd_loss": 0.015040256921201945 + }, + { + "epoch": 7.98661623508874, + "flow_matching_loss": 0.01928673256188631, + "grad_norm": 0.2334970086812973, + "learning_rate": 2.7455023818506423e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068404557649046184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002950735518243164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016631647385656832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8763965487480163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27450, + "weighted_flow_matching_loss": 0.01928673256188631, + "weighted_rkd_loss": 0.01457587517797947 + }, + { + "epoch": 7.989525749199884, + "flow_matching_loss": 0.01895973887294531, + "grad_norm": 0.20377779006958008, + "learning_rate": 2.745271979479358e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006648565363138914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030314977397210897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016328628361225127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8763103663921357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27460, + "weighted_flow_matching_loss": 0.01895973887294531, + "weighted_rkd_loss": 0.014308920968323946 + }, + { + "epoch": 7.992435263311027, + "flow_matching_loss": 0.019891276210546493, + "grad_norm": 0.18348288536071777, + "learning_rate": 2.745041482537885e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062284951098263265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027651139418594537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01522210417315364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.876224160194397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27470, + "weighted_flow_matching_loss": 0.019891276210546493, + "weighted_rkd_loss": 0.013337990269064903 + }, + { + "epoch": 7.99534477742217, + "flow_matching_loss": 0.02000203337520361, + "grad_norm": 0.19679200649261475, + "learning_rate": 2.744810891043728e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007433135015890002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00315994240809232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018026212602853774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8761379301548005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27480, + "weighted_flow_matching_loss": 0.02000203337520361, + "weighted_rkd_loss": 0.015793433599174022 + }, + { + "epoch": 7.998254291533314, + "flow_matching_loss": 0.019954474177211523, + "grad_norm": 0.20289559662342072, + "learning_rate": 2.7445802050143987e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066543300170451404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028579307952895762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016166590619832277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8760516762733459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27490, + "weighted_flow_matching_loss": 0.019954474177211523, + "weighted_rkd_loss": 0.014162771496921777 + }, + { + "epoch": 8.001163805644458, + "flow_matching_loss": 0.01801422080025077, + "grad_norm": 0.17961403727531433, + "learning_rate": 2.7443494244674167e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006186900148168206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026281367056071757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015001936815679073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8759653925895691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27500, + "weighted_flow_matching_loss": 0.01801422080025077, + "weighted_rkd_loss": 0.013141202833503484 + }, + { + "epoch": 8.004073319755602, + "flow_matching_loss": 0.02051801234483719, + "grad_norm": 0.24824869632720947, + "learning_rate": 2.7441185494203077e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071432969067245725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00320137282833457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017487966828048228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8758790850639343, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27510, + "weighted_flow_matching_loss": 0.02051801234483719, + "weighted_rkd_loss": 0.015317349415272474 + }, + { + "epoch": 8.006982833866743, + "flow_matching_loss": 0.019720847345888616, + "grad_norm": 0.15938831865787506, + "learning_rate": 2.7438875798906054e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006308371992781758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002923346613533795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01554009085521102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8757927477359772, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27520, + "weighted_flow_matching_loss": 0.019720847345888616, + "weighted_rkd_loss": 0.013609905634075404 + }, + { + "epoch": 8.009892347977887, + "flow_matching_loss": 0.021157023776322604, + "grad_norm": 0.21515603363513947, + "learning_rate": 2.74365651589585e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007282743183895946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032194134080782534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017784900031983852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8757063925266266, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27530, + "weighted_flow_matching_loss": 0.021157023776322604, + "weighted_rkd_loss": 0.015574368182569742 + }, + { + "epoch": 8.012801862089031, + "flow_matching_loss": 0.01830006595700979, + "grad_norm": 0.20113500952720642, + "learning_rate": 2.74342535745359e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007357367221266032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003227111115120351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017941845767199992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8756200015544892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27540, + "weighted_flow_matching_loss": 0.01830006595700979, + "weighted_rkd_loss": 0.015710229706019162 + }, + { + "epoch": 8.015711376200175, + "flow_matching_loss": 0.019688067957758904, + "grad_norm": 0.2232484370470047, + "learning_rate": 2.7431941045813803e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006100364169105888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026454960578121247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014846224244683981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8755336105823517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27550, + "weighted_flow_matching_loss": 0.019688067957758904, + "weighted_rkd_loss": 0.0129983551800251 + }, + { + "epoch": 8.018620890311318, + "flow_matching_loss": 0.02038878044113517, + "grad_norm": 0.21630366146564484, + "learning_rate": 2.742962757296782e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00708528021350503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003255590354092419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017426150757819413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8754471600055694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27560, + "weighted_flow_matching_loss": 0.02038878044113517, + "weighted_rkd_loss": 0.015255657397210598 + }, + { + "epoch": 8.021530404422462, + "flow_matching_loss": 0.021752911247313024, + "grad_norm": 0.21418854594230652, + "learning_rate": 2.7427313156173658e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007088274974375963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002937993244268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017114543169736863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8753606975078583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27570, + "weighted_flow_matching_loss": 0.021752911247313024, + "weighted_rkd_loss": 0.014981360081583262 + }, + { + "epoch": 8.024439918533606, + "flow_matching_loss": 0.01903289882466197, + "grad_norm": 0.24287143349647522, + "learning_rate": 2.742499779560707e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006504543405026197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002805560687556863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015814647357910872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8752741992473603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27580, + "weighted_flow_matching_loss": 0.01903289882466197, + "weighted_rkd_loss": 0.013842111825942994 + }, + { + "epoch": 8.027349432644748, + "flow_matching_loss": 0.018100554775446653, + "grad_norm": 0.23143772780895233, + "learning_rate": 2.7422681491443893e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006671661976724863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029240492032840846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016267373133450745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8751876950263977, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27590, + "weighted_flow_matching_loss": 0.018100554775446653, + "weighted_rkd_loss": 0.014237046893686055 + }, + { + "epoch": 8.030258946755891, + "flow_matching_loss": 0.021131197176873683, + "grad_norm": 0.2574550211429596, + "learning_rate": 2.7420364243860048e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00753794745542109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033521021250635386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01842799698933959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8751011490821838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27600, + "weighted_flow_matching_loss": 0.021131197176873683, + "weighted_rkd_loss": 0.016126327961683274 + }, + { + "epoch": 8.033168460867035, + "flow_matching_loss": 0.020068153087049723, + "grad_norm": 0.20126447081565857, + "learning_rate": 2.7418046053031503e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007177394488826394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002947495807893574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0173022847622633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8750145852565765, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27610, + "weighted_flow_matching_loss": 0.020068153087049723, + "weighted_rkd_loss": 0.015139775257557631 + }, + { + "epoch": 8.036077974978179, + "flow_matching_loss": 0.019739916548132896, + "grad_norm": 0.17928482592105865, + "learning_rate": 2.7415726919134313e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00639104014262557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002804643241688609, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015586723573505878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8749279916286469, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27620, + "weighted_flow_matching_loss": 0.019739916548132896, + "weighted_rkd_loss": 0.013637273013591767 + }, + { + "epoch": 8.038987489089322, + "flow_matching_loss": 0.018089197017252444, + "grad_norm": 0.214387446641922, + "learning_rate": 2.74134068423446e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006465374957770109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028077048948034642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015738454554229973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8748413741588592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27630, + "weighted_flow_matching_loss": 0.018089197017252444, + "weighted_rkd_loss": 0.013768637087196112 + }, + { + "epoch": 8.041897003200466, + "flow_matching_loss": 0.019211567286401988, + "grad_norm": 0.2278691828250885, + "learning_rate": 2.7411085822838554e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070599410217255356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003152638883329928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017272520717233418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8747547328472137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27640, + "weighted_flow_matching_loss": 0.019211567286401988, + "weighted_rkd_loss": 0.015109227783977986 + }, + { + "epoch": 8.044806517311608, + "flow_matching_loss": 0.020048882253468035, + "grad_norm": 0.22795234620571136, + "learning_rate": 2.740876386079245e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066811803728342055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002851037064101547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016213397774845362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8746680676937103, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27650, + "weighted_flow_matching_loss": 0.020048882253468035, + "weighted_rkd_loss": 0.014181363116949796 + }, + { + "epoch": 8.047716031422752, + "flow_matching_loss": 0.020626751892268657, + "grad_norm": 0.2728878855705261, + "learning_rate": 2.7406440956382626e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006415008381009102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030911100446246566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015921126678586008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.87458136677742, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27660, + "weighted_flow_matching_loss": 0.020626751892268657, + "weighted_rkd_loss": 0.013924292847514152 + }, + { + "epoch": 8.050625545533896, + "flow_matching_loss": 0.021183013543486597, + "grad_norm": 0.19293631613254547, + "learning_rate": 2.7404117109785482e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007416685903444886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033871039282530545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01822047559544444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8744946420192719, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27670, + "weighted_flow_matching_loss": 0.021183013543486597, + "weighted_rkd_loss": 0.015933694317936897 + }, + { + "epoch": 8.05353505964504, + "flow_matching_loss": 0.01936487350612879, + "grad_norm": 0.18662309646606445, + "learning_rate": 2.7401792321177503e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006730271736159921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028413195628672837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016301862988620998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8744079053401947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27680, + "weighted_flow_matching_loss": 0.01936487350612879, + "weighted_rkd_loss": 0.01425453694537282 + }, + { + "epoch": 8.056444573756183, + "flow_matching_loss": 0.01804658267647028, + "grad_norm": 0.21608945727348328, + "learning_rate": 2.7399466590735243e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005993397254496813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026863171719014645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01467311168089509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8743211388587951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27690, + "weighted_flow_matching_loss": 0.01804658267647028, + "weighted_rkd_loss": 0.01282899770885706 + }, + { + "epoch": 8.059354087867327, + "flow_matching_loss": 0.018541238270699978, + "grad_norm": 0.20590434968471527, + "learning_rate": 2.7397139918635325e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007038708170875907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030396123649552463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017117028404027224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8742343425750733, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27700, + "weighted_flow_matching_loss": 0.018541238270699978, + "weighted_rkd_loss": 0.01496431529521942 + }, + { + "epoch": 8.06226360197847, + "flow_matching_loss": 0.020514513831585646, + "grad_norm": 0.16640907526016235, + "learning_rate": 2.739481230505444e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006924764066934585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030634598108008502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016912987735122443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8741475164890289, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27710, + "weighted_flow_matching_loss": 0.020514513831585646, + "weighted_rkd_loss": 0.014784453622996806 + }, + { + "epoch": 8.065173116089612, + "flow_matching_loss": 0.02104636523872614, + "grad_norm": 0.17830848693847656, + "learning_rate": 2.7392483750169362e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006178533378988505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002643642423208803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0150007089599967, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8740606725215911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27720, + "weighted_flow_matching_loss": 0.02104636523872614, + "weighted_rkd_loss": 0.013111545238643885 + }, + { + "epoch": 8.068082630200756, + "flow_matching_loss": 0.019054932333528996, + "grad_norm": 0.18788479268550873, + "learning_rate": 2.739015425415693e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006303231278434396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002830358664505184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01543682124465704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8739737927913666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27730, + "weighted_flow_matching_loss": 0.019054932333528996, + "weighted_rkd_loss": 0.013491381052881479 + }, + { + "epoch": 8.0709921443119, + "flow_matching_loss": 0.0204637810587883, + "grad_norm": 0.1934954971075058, + "learning_rate": 2.738782381719405e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006603773450478911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027872053906321526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015994752384722234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8738869011402131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27740, + "weighted_flow_matching_loss": 0.0204637810587883, + "weighted_rkd_loss": 0.013977574277669191 + }, + { + "epoch": 8.073901658423043, + "flow_matching_loss": 0.018377150502055885, + "grad_norm": 0.21031539142131805, + "learning_rate": 2.73854924394577e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00647634151391685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027695351745933294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01572221824899316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8737999677658081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27750, + "weighted_flow_matching_loss": 0.018377150502055885, + "weighted_rkd_loss": 0.013738054595887662 + }, + { + "epoch": 8.076811172534187, + "flow_matching_loss": 0.021583994291722773, + "grad_norm": 0.18941214680671692, + "learning_rate": 2.7383160121124935e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006357237603515387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002850087615661323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01556456284597516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8737130165100098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27760, + "weighted_flow_matching_loss": 0.021583994291722773, + "weighted_rkd_loss": 0.013598955143243075 + }, + { + "epoch": 8.07972068664533, + "flow_matching_loss": 0.020214739255607128, + "grad_norm": 0.21325939893722534, + "learning_rate": 2.7380826862372885e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006687052035704255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002952707337681204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016326811537146567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8736260533332825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27770, + "weighted_flow_matching_loss": 0.020214739255607128, + "weighted_rkd_loss": 0.014263525418937207 + }, + { + "epoch": 8.082630200756475, + "flow_matching_loss": 0.022666032053530218, + "grad_norm": 0.21686826646327972, + "learning_rate": 2.7378492663378735e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006445169588550926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002812852256465703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015703191328793765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8735390305519104, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27780, + "weighted_flow_matching_loss": 0.022666032053530218, + "weighted_rkd_loss": 0.013717313297092915 + }, + { + "epoch": 8.085539714867616, + "flow_matching_loss": 0.018308105506002904, + "grad_norm": 0.22954058647155762, + "learning_rate": 2.7376157524319764e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066130409017205235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003024633519817144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016250715032219885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8734520077705383, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27790, + "weighted_flow_matching_loss": 0.018308105506002904, + "weighted_rkd_loss": 0.01419419962912798 + }, + { + "epoch": 8.08844922897876, + "flow_matching_loss": 0.01850384073331952, + "grad_norm": 0.1900399625301361, + "learning_rate": 2.73738214453733e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006231069378554821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028957536444067953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015357892215251922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8733649671077728, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27800, + "weighted_flow_matching_loss": 0.01850384073331952, + "weighted_rkd_loss": 0.013413027860224247 + }, + { + "epoch": 8.091358743089904, + "flow_matching_loss": 0.01985397655516863, + "grad_norm": 0.21077580749988556, + "learning_rate": 2.7371484426716764e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005767636606469751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002547436859458685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0140827102586627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8732778847217559, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27810, + "weighted_flow_matching_loss": 0.01985397655516863, + "weighted_rkd_loss": 0.012298098998144269 + }, + { + "epoch": 8.094268257201048, + "flow_matching_loss": 0.020013681333512067, + "grad_norm": 0.1813381314277649, + "learning_rate": 2.7369146468527626e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061961057595908645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003059933823533356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015452145133167505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8731907844543457, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27820, + "weighted_flow_matching_loss": 0.020013681333512067, + "weighted_rkd_loss": 0.013492671772837639 + }, + { + "epoch": 8.097177771312191, + "flow_matching_loss": 0.01918209744617343, + "grad_norm": 0.2639859616756439, + "learning_rate": 2.7366807570983442e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006536425277590752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030703415279276667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016143191885203124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8731036484241486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27830, + "weighted_flow_matching_loss": 0.01918209744617343, + "weighted_rkd_loss": 0.014094629511237144 + }, + { + "epoch": 8.100087285423335, + "flow_matching_loss": 0.021220959536731244, + "grad_norm": 0.18992973864078522, + "learning_rate": 2.736446773426184e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006485396390780806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028165159048512576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015787308756262063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8730164885520935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27840, + "weighted_flow_matching_loss": 0.021220959536731244, + "weighted_rkd_loss": 0.01378257256001234 + }, + { + "epoch": 8.102996799534477, + "flow_matching_loss": 0.020090225525200366, + "grad_norm": 0.19387106597423553, + "learning_rate": 2.7362126958540513e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006969983549788594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00286147641018033, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016801443137228488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8729293167591095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27850, + "weighted_flow_matching_loss": 0.020090225525200366, + "weighted_rkd_loss": 0.014666479267179965 + }, + { + "epoch": 8.10590631364562, + "flow_matching_loss": 0.019442720990628003, + "grad_norm": 0.19463267922401428, + "learning_rate": 2.735978524399722e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059966950211673975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026769471587613225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014670337084680796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8728421032428741, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27860, + "weighted_flow_matching_loss": 0.019442720990628003, + "weighted_rkd_loss": 0.012804931029677391 + }, + { + "epoch": 8.108815827756764, + "flow_matching_loss": 0.020136240310966967, + "grad_norm": 0.23672837018966675, + "learning_rate": 2.735744259080981e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067905894946306946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029453333001583813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0165265123359859, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8727548778057098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27870, + "weighted_flow_matching_loss": 0.020136240310966967, + "weighted_rkd_loss": 0.014423610270023346 + }, + { + "epoch": 8.111725341867908, + "flow_matching_loss": 0.01897885203361511, + "grad_norm": 0.20856443047523499, + "learning_rate": 2.7355098999156185e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005938695790246129, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002615510404575616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014492901973426342, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8726676225662231, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27880, + "weighted_flow_matching_loss": 0.01897885203361511, + "weighted_rkd_loss": 0.012647482939064503 + }, + { + "epoch": 8.114634855979052, + "flow_matching_loss": 0.019121828395873307, + "grad_norm": 0.21383336186408997, + "learning_rate": 2.735275446921433e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006580340396612883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002804130385629833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01596481129527092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8725803375244141, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27890, + "weighted_flow_matching_loss": 0.019121828395873307, + "weighted_rkd_loss": 0.013930587004870177 + }, + { + "epoch": 8.117544370090195, + "flow_matching_loss": 0.01955691948533058, + "grad_norm": 0.20344780385494232, + "learning_rate": 2.735040900116229e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065142848063260315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031966192182153463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016225188877433537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8724930286407471, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27900, + "weighted_flow_matching_loss": 0.01955691948533058, + "weighted_rkd_loss": 0.014156411215662957 + }, + { + "epoch": 8.12045388420134, + "flow_matching_loss": 0.020471197366714478, + "grad_norm": 0.2596740126609802, + "learning_rate": 2.7348062595178196e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007306305319070816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034280216321349146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018040632363408804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8724056899547576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27910, + "weighted_flow_matching_loss": 0.020471197366714478, + "weighted_rkd_loss": 0.01573873935267329 + }, + { + "epoch": 8.123363398312481, + "flow_matching_loss": 0.01841569719836116, + "grad_norm": 0.26293155550956726, + "learning_rate": 2.734571525144024e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065947241149842736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028313116636127234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01602075956761837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8723183333873749, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27920, + "weighted_flow_matching_loss": 0.01841569719836116, + "weighted_rkd_loss": 0.013975123316049576 + }, + { + "epoch": 8.126272912423625, + "flow_matching_loss": 0.01953132152557373, + "grad_norm": 0.19913657009601593, + "learning_rate": 2.734336697012668e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006245383061468601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026266063330695035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015117372665554286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8722309470176697, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27930, + "weighted_flow_matching_loss": 0.01953132152557373, + "weighted_rkd_loss": 0.013185831904411315 + }, + { + "epoch": 8.129182426534769, + "flow_matching_loss": 0.02111178133636713, + "grad_norm": 0.20738455653190613, + "learning_rate": 2.734101775141586e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060688723810017105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002520165010355413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0146579097956419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.872143542766571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27940, + "weighted_flow_matching_loss": 0.02111178133636713, + "weighted_rkd_loss": 0.01278380686417222 + }, + { + "epoch": 8.132091940645912, + "flow_matching_loss": 0.01957071628421545, + "grad_norm": 0.2736552953720093, + "learning_rate": 2.7338667595486188e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00700021293014288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002996934251859784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016997359972447158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8720561027526855, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27950, + "weighted_flow_matching_loss": 0.01957071628421545, + "weighted_rkd_loss": 0.014822685066610575 + }, + { + "epoch": 8.135001454757056, + "flow_matching_loss": 0.01914576031267643, + "grad_norm": 0.24393054842948914, + "learning_rate": 2.7336316502516138e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006595445843413472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029260678915306927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016116959601640703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8719686448574067, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27960, + "weighted_flow_matching_loss": 0.01914576031267643, + "weighted_rkd_loss": 0.014053499326109886 + }, + { + "epoch": 8.1379109688682, + "flow_matching_loss": 0.019021904002875088, + "grad_norm": 0.196969136595726, + "learning_rate": 2.7333964472684257e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062227834481745955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002750080788973719, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015195647813379764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8718811511993408, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27970, + "weighted_flow_matching_loss": 0.019021904002875088, + "weighted_rkd_loss": 0.013248821068555117 + }, + { + "epoch": 8.140820482979342, + "flow_matching_loss": 0.02015702612698078, + "grad_norm": 0.18293282389640808, + "learning_rate": 2.733161150616918e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006054862029850483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002655180497094989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014764904603362083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8717936336994171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27980, + "weighted_flow_matching_loss": 0.02015702612698078, + "weighted_rkd_loss": 0.012871972098946572 + }, + { + "epoch": 8.143729997090485, + "flow_matching_loss": 0.021267618518322706, + "grad_norm": 0.2490990310907364, + "learning_rate": 2.7329257603149585e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00673838839866221, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002952171175274998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016428948007524015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8717061042785644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27990, + "weighted_flow_matching_loss": 0.021267618518322706, + "weighted_rkd_loss": 0.014321168512105941 + }, + { + "epoch": 8.146639511201629, + "flow_matching_loss": 0.021195058524608613, + "grad_norm": 0.17868009209632874, + "learning_rate": 2.7326902763804245e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006048851972445845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025203742552548647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014618078153580428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8716185390949249, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28000, + "weighted_flow_matching_loss": 0.021195058524608613, + "weighted_rkd_loss": 0.012741404213011264 + }, + { + "epoch": 8.149549025312773, + "flow_matching_loss": 0.019995448272675277, + "grad_norm": 0.2047051191329956, + "learning_rate": 2.7324546988311986e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006092763319611549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00275521605508402, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014940742682665587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8715309500694275, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28010, + "weighted_flow_matching_loss": 0.019995448272675277, + "weighted_rkd_loss": 0.013021330907940864 + }, + { + "epoch": 8.152458539423916, + "flow_matching_loss": 0.020572574343532324, + "grad_norm": 0.20053088665008545, + "learning_rate": 2.7322190276851723e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006212965352460742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002696961455512792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015122891962528228, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8714433550834656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28020, + "weighted_flow_matching_loss": 0.020572574343532324, + "weighted_rkd_loss": 0.013178718462586402 + }, + { + "epoch": 8.15536805353506, + "flow_matching_loss": 0.02045581182464957, + "grad_norm": 0.38675180077552795, + "learning_rate": 2.7319832629602425e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063577815890312195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002885912067722529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015601475443691015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8713556826114655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28030, + "weighted_flow_matching_loss": 0.02045581182464957, + "weighted_rkd_loss": 0.013594420440495015 + }, + { + "epoch": 8.158277567646204, + "flow_matching_loss": 0.021243900898844002, + "grad_norm": 0.19293907284736633, + "learning_rate": 2.7317474046743143e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006356342323124408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026169642922468485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015329649113118649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8712680280208588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28040, + "weighted_flow_matching_loss": 0.021243900898844002, + "weighted_rkd_loss": 0.013356260769069195 + }, + { + "epoch": 8.161187081757346, + "flow_matching_loss": 0.019873747788369655, + "grad_norm": 0.20471638441085815, + "learning_rate": 2.7315114528452994e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068408453837037085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030416369903832675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016723327804356813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8711803436279297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28050, + "weighted_flow_matching_loss": 0.019873747788369655, + "weighted_rkd_loss": 0.014569012820720673 + }, + { + "epoch": 8.16409659586849, + "flow_matching_loss": 0.019372293073683978, + "grad_norm": 0.23893271386623383, + "learning_rate": 2.731275407491117e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064306686166673895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026399666676297784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015501303598284721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8710926234722137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28060, + "weighted_flow_matching_loss": 0.019372293073683978, + "weighted_rkd_loss": 0.013503036834299565 + }, + { + "epoch": 8.167006109979633, + "flow_matching_loss": 0.02066247370094061, + "grad_norm": 0.1961141973733902, + "learning_rate": 2.7310392686296936e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007036502519622445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030355100287124516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01710851499810815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8710048794746399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28070, + "weighted_flow_matching_loss": 0.02066247370094061, + "weighted_rkd_loss": 0.01490167872980237 + }, + { + "epoch": 8.169915624090777, + "flow_matching_loss": 0.02224689219146967, + "grad_norm": 0.21953576803207397, + "learning_rate": 2.7308030362789614e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00597010925412178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027634650468826294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01470368355512619, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8709171116352081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28080, + "weighted_flow_matching_loss": 0.02224689219146967, + "weighted_rkd_loss": 0.012805684749037028 + }, + { + "epoch": 8.17282513820192, + "flow_matching_loss": 0.01876609195023775, + "grad_norm": 0.19149957597255707, + "learning_rate": 2.7305667104568614e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006513939937576652, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002764901192858815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015792781300842763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8708293259143829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28090, + "weighted_flow_matching_loss": 0.01876609195023775, + "weighted_rkd_loss": 0.013752776104956866 + }, + { + "epoch": 8.175734652313064, + "flow_matching_loss": 0.020504665561020374, + "grad_norm": 0.2006070464849472, + "learning_rate": 2.7303302911813402e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006669784523546695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002783730928786099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01612330013886094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8707415044307709, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28100, + "weighted_flow_matching_loss": 0.020504665561020374, + "weighted_rkd_loss": 0.014039227925240993 + }, + { + "epoch": 8.178644166424208, + "flow_matching_loss": 0.019613811932504176, + "grad_norm": 0.20916783809661865, + "learning_rate": 2.7300937784703536e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00678233071230352, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003084124776069075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01664878623560071, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8706536591053009, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28110, + "weighted_flow_matching_loss": 0.019613811932504176, + "weighted_rkd_loss": 0.014495362248271703 + }, + { + "epoch": 8.18155368053535, + "flow_matching_loss": 0.019980312511324882, + "grad_norm": 0.22735056281089783, + "learning_rate": 2.729857172341862e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006897085160017013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033469255547970535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01714109592139721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8705657899379731, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28120, + "weighted_flow_matching_loss": 0.019980312511324882, + "weighted_rkd_loss": 0.014922471158206463 + }, + { + "epoch": 8.184463194646494, + "flow_matching_loss": 0.02205710522830486, + "grad_norm": 0.15803444385528564, + "learning_rate": 2.7296204728138347e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006892201770097017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032047613989561797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016989164892584084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8704778850078583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28130, + "weighted_flow_matching_loss": 0.02205710522830486, + "weighted_rkd_loss": 0.01478866981342435 + }, + { + "epoch": 8.187372708757637, + "flow_matching_loss": 0.01920759603381157, + "grad_norm": 0.2315058410167694, + "learning_rate": 2.7293836799042474e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006532578868791461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028664465062320232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015931604243814946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8703899681568146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28140, + "weighted_flow_matching_loss": 0.01920759603381157, + "weighted_rkd_loss": 0.01386669659987092 + }, + { + "epoch": 8.190282222868781, + "flow_matching_loss": 0.022107106726616622, + "grad_norm": 0.16856469213962555, + "learning_rate": 2.7291467936310824e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006677821930497885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028746935771778227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01623033722862601, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8703020334243774, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28150, + "weighted_flow_matching_loss": 0.022107106726616622, + "weighted_rkd_loss": 0.01412533912807703 + }, + { + "epoch": 8.193191736979925, + "flow_matching_loss": 0.020700084511190653, + "grad_norm": 0.24900132417678833, + "learning_rate": 2.728909814012331e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070683171041309835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029905134346336127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01712714759632945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8702140629291535, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28160, + "weighted_flow_matching_loss": 0.020700084511190653, + "weighted_rkd_loss": 0.014904256258159877 + }, + { + "epoch": 8.196101251091068, + "flow_matching_loss": 0.02124746926128864, + "grad_norm": 0.24303309619426727, + "learning_rate": 2.7286727410659884e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055075255688279865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023884592694230376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013403510488569737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8701260566711426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28170, + "weighted_flow_matching_loss": 0.02124746926128864, + "weighted_rkd_loss": 0.011662750225514173 + }, + { + "epoch": 8.19901076520221, + "flow_matching_loss": 0.018920320365577937, + "grad_norm": 0.20686691999435425, + "learning_rate": 2.7284355748100596e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00631385650485754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026851062662899493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015312819369137287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8700380444526672, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28180, + "weighted_flow_matching_loss": 0.018920320365577937, + "weighted_rkd_loss": 0.013322726171463729 + }, + { + "epoch": 8.201920279313354, + "flow_matching_loss": 0.020560551341623066, + "grad_norm": 0.25634363293647766, + "learning_rate": 2.7281983152625563e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006450249161571264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002854786976240575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015755285136401655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.869949996471405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28190, + "weighted_flow_matching_loss": 0.020560551341623066, + "weighted_rkd_loss": 0.013706254679709673 + }, + { + "epoch": 8.204829793424498, + "flow_matching_loss": 0.019659495167434215, + "grad_norm": 0.28921207785606384, + "learning_rate": 2.7279609624414962e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006986741442233324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031917446525767446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017165227327495813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8698619246482849, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28200, + "weighted_flow_matching_loss": 0.019659495167434215, + "weighted_rkd_loss": 0.014931360818445682 + }, + { + "epoch": 8.207739307535642, + "flow_matching_loss": 0.018157634139060973, + "grad_norm": 0.19435296952724457, + "learning_rate": 2.7277235163649052e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006396574061363935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00264530333224684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01543845171108842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8697738349437714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28210, + "weighted_flow_matching_loss": 0.018157634139060973, + "weighted_rkd_loss": 0.013427933305501938 + }, + { + "epoch": 8.210648821646785, + "flow_matching_loss": 0.019234766159206627, + "grad_norm": 0.23006165027618408, + "learning_rate": 2.727485977050815e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006079060072079301, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002678188309073448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014836308546364307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8696857035160065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28220, + "weighted_flow_matching_loss": 0.019234766159206627, + "weighted_rkd_loss": 0.012902917712926865 + }, + { + "epoch": 8.213558335757929, + "flow_matching_loss": 0.02000089259818196, + "grad_norm": 0.19141177833080292, + "learning_rate": 2.7272483445172657e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006568047450855374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002865983545780182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016002078354358674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8695975542068481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28230, + "weighted_flow_matching_loss": 0.02000089259818196, + "weighted_rkd_loss": 0.013915396202355624 + }, + { + "epoch": 8.216467849869073, + "flow_matching_loss": 0.020344395376741885, + "grad_norm": 0.1861410140991211, + "learning_rate": 2.7270106187823035e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059680560603737835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025474363123066724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014483548514544965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8695093989372253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28240, + "weighted_flow_matching_loss": 0.020344395376741885, + "weighted_rkd_loss": 0.01259359773248434 + }, + { + "epoch": 8.219377363980215, + "flow_matching_loss": 0.018024738412350416, + "grad_norm": 0.16334131360054016, + "learning_rate": 2.7267727998639827e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006171996938064694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002844477503094822, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01518847132101655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8694211840629578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28250, + "weighted_flow_matching_loss": 0.018024738412350416, + "weighted_rkd_loss": 0.013205225113779306 + }, + { + "epoch": 8.222286878091358, + "flow_matching_loss": 0.017375443875789643, + "grad_norm": 0.2113490104675293, + "learning_rate": 2.7265348877803634e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006102723302319646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026020403602160513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01480748699977994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8693329691886902, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28260, + "weighted_flow_matching_loss": 0.017375443875789643, + "weighted_rkd_loss": 0.012872652802616357 + }, + { + "epoch": 8.225196392202502, + "flow_matching_loss": 0.019207036308944225, + "grad_norm": 0.15834158658981323, + "learning_rate": 2.7262968825495145e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005528077157214284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002449690690264106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013505844958126545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8692447125911713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28270, + "weighted_flow_matching_loss": 0.019207036308944225, + "weighted_rkd_loss": 0.011739891394972802 + }, + { + "epoch": 8.228105906313646, + "flow_matching_loss": 0.019157120306044818, + "grad_norm": 0.24845920503139496, + "learning_rate": 2.72605878418951e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055272913537919525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024173163110390306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013471899088472128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8691564440727234, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28280, + "weighted_flow_matching_loss": 0.019157120306044818, + "weighted_rkd_loss": 0.01170916622504592 + }, + { + "epoch": 8.23101542042479, + "flow_matching_loss": 0.02042880980297923, + "grad_norm": 0.164522185921669, + "learning_rate": 2.7258205927184314e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006440130947157741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002890381251927465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01577064311131835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8690681338310242, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28290, + "weighted_flow_matching_loss": 0.02042880980297923, + "weighted_rkd_loss": 0.013705763407051564 + }, + { + "epoch": 8.233924934535933, + "flow_matching_loss": 0.020043129194527866, + "grad_norm": 0.18907958269119263, + "learning_rate": 2.7255823081543693e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007211082940921187, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003416379343252629, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01783854505047202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8689798176288605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28300, + "weighted_flow_matching_loss": 0.020043129194527866, + "weighted_rkd_loss": 0.015501400269567967 + }, + { + "epoch": 8.236834448647077, + "flow_matching_loss": 0.020593446772545577, + "grad_norm": 0.19410640001296997, + "learning_rate": 2.7253439305154186e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006631199270486832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002987484633922577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016249882895499466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.86889146566391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28310, + "weighted_flow_matching_loss": 0.020593446772545577, + "weighted_rkd_loss": 0.014119394216686487 + }, + { + "epoch": 8.239743962758219, + "flow_matching_loss": 0.017649259138852357, + "grad_norm": 0.21191419661045074, + "learning_rate": 2.7251054598196827e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064422913361340765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029080684762448072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015792651008814572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8688030898571014, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28320, + "weighted_flow_matching_loss": 0.017649259138852357, + "weighted_rkd_loss": 0.01372072398662567 + }, + { + "epoch": 8.242653476869362, + "flow_matching_loss": 0.019568288512527943, + "grad_norm": 0.19157232344150543, + "learning_rate": 2.724866896085273e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006664983183145523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028408330399543047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016170799359679223, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8687146782875061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28330, + "weighted_flow_matching_loss": 0.019568288512527943, + "weighted_rkd_loss": 0.014047758653759956 + }, + { + "epoch": 8.245562990980506, + "flow_matching_loss": 0.018928624875843524, + "grad_norm": 0.18000571429729462, + "learning_rate": 2.7246282393303053e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006012580078095198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027139533427543937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014739113580435514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8686262607574463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28340, + "weighted_flow_matching_loss": 0.018928624875843524, + "weighted_rkd_loss": 0.012802788335829974 + }, + { + "epoch": 8.24847250509165, + "flow_matching_loss": 0.018898638803511857, + "grad_norm": 0.2549968957901001, + "learning_rate": 2.7243894895729046e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005615986324846745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002540079480968416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013772052247077227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8685378074645996, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28350, + "weighted_flow_matching_loss": 0.018898638803511857, + "weighted_rkd_loss": 0.01196153536438942 + }, + { + "epoch": 8.251382019202794, + "flow_matching_loss": 0.019058072939515114, + "grad_norm": 0.1864999234676361, + "learning_rate": 2.7241506468312028e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006216961052268743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002747913566417992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015181835927069186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.868449330329895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28360, + "weighted_flow_matching_loss": 0.019058072939515114, + "weighted_rkd_loss": 0.013184654153883457 + }, + { + "epoch": 8.254291533313937, + "flow_matching_loss": 0.01851974232122302, + "grad_norm": 0.17744410037994385, + "learning_rate": 2.7239117111233374e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006826101010665298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029957347316667436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016647936776280403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.868360823392868, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28370, + "weighted_flow_matching_loss": 0.01851974232122302, + "weighted_rkd_loss": 0.014456384629011155 + }, + { + "epoch": 8.25720104742508, + "flow_matching_loss": 0.02107167299836874, + "grad_norm": 0.220230832695961, + "learning_rate": 2.7236726824674558e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062153932172805074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003202075813896954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015632862225174904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8682723045349121, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28380, + "weighted_flow_matching_loss": 0.02107167299836874, + "weighted_rkd_loss": 0.013573605194687843 + }, + { + "epoch": 8.260110561536223, + "flow_matching_loss": 0.01943932259455323, + "grad_norm": 0.22983534634113312, + "learning_rate": 2.7234335608817088e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006323318183422089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027789945714175703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01542563121765852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8681837499141694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28390, + "weighted_flow_matching_loss": 0.01943932259455323, + "weighted_rkd_loss": 0.013392264302819968 + }, + { + "epoch": 8.263020075647367, + "flow_matching_loss": 0.019179133605211972, + "grad_norm": 0.19706177711486816, + "learning_rate": 2.723194346384257e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059945517219603065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024898017523810267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014478905219584703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8680951714515686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28400, + "weighted_flow_matching_loss": 0.019179133605211972, + "weighted_rkd_loss": 0.012569074425846339 + }, + { + "epoch": 8.26592958975851, + "flow_matching_loss": 0.02097277920693159, + "grad_norm": 0.23970438539981842, + "learning_rate": 2.7229550389932665e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00722128814086318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003292597597464919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017735173739492894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8680065691471099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28410, + "weighted_flow_matching_loss": 0.02097277920693159, + "weighted_rkd_loss": 0.015394266974180938 + }, + { + "epoch": 8.268839103869654, + "flow_matching_loss": 0.018642285838723183, + "grad_norm": 0.18167510628700256, + "learning_rate": 2.722715638726912e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005738078942522406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025334394769743083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014009597431868314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8679179430007935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28420, + "weighted_flow_matching_loss": 0.018642285838723183, + "weighted_rkd_loss": 0.012159196007996798 + }, + { + "epoch": 8.271748617980798, + "flow_matching_loss": 0.018381100334227084, + "grad_norm": 0.198482483625412, + "learning_rate": 2.722476145603374e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006082655815407634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027820639428682626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014947375562041999, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.867829293012619, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28430, + "weighted_flow_matching_loss": 0.018381100334227084, + "weighted_rkd_loss": 0.012971767596900463 + }, + { + "epoch": 8.274658132091941, + "flow_matching_loss": 0.019799395743757487, + "grad_norm": 0.1924360990524292, + "learning_rate": 2.72223655964084e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006829796265810728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031578070716932415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016817399580031634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8677406072616577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28440, + "weighted_flow_matching_loss": 0.019799395743757487, + "weighted_rkd_loss": 0.01459312541410327 + }, + { + "epoch": 8.277567646203083, + "flow_matching_loss": 0.019154466968029738, + "grad_norm": 0.13024333119392395, + "learning_rate": 2.721996880857506e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006360354041680694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028897287673316898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015610436908900738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8676519095897675, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28450, + "weighted_flow_matching_loss": 0.019154466968029738, + "weighted_rkd_loss": 0.013544435147196054 + }, + { + "epoch": 8.280477160314227, + "flow_matching_loss": 0.019802328012883664, + "grad_norm": 0.23680922389030457, + "learning_rate": 2.7217571092715727e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00707364697009325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032186669763177634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017365960869938135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8675631821155548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28460, + "weighted_flow_matching_loss": 0.019802328012883664, + "weighted_rkd_loss": 0.015066121239215136 + }, + { + "epoch": 8.28338667442537, + "flow_matching_loss": 0.0192225425504148, + "grad_norm": 0.21039623022079468, + "learning_rate": 2.7215172449012504e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005607863515615463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002350417582783848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013566144462674855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8674744188785553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28470, + "weighted_flow_matching_loss": 0.0192225425504148, + "weighted_rkd_loss": 0.011768273077905178 + }, + { + "epoch": 8.286296188536515, + "flow_matching_loss": 0.01880639297887683, + "grad_norm": 0.2286294400691986, + "learning_rate": 2.7212772877647542e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006731376983225346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002981100487522781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01644385438412428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8673856556415558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28480, + "weighted_flow_matching_loss": 0.01880639297887683, + "weighted_rkd_loss": 0.014263168163597583 + }, + { + "epoch": 8.289205702647658, + "flow_matching_loss": 0.019916130974888802, + "grad_norm": 0.2501591444015503, + "learning_rate": 2.721037237880308e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006271728593856096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026106823468580844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01515413960441947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8672968447208405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28490, + "weighted_flow_matching_loss": 0.019916130974888802, + "weighted_rkd_loss": 0.013143089599907398 + }, + { + "epoch": 8.292115216758802, + "flow_matching_loss": 0.018310187850147486, + "grad_norm": 0.17370161414146423, + "learning_rate": 2.7207970952661417e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00618593292310834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002574358647689223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014946224633604288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8672080159187316, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28500, + "weighted_flow_matching_loss": 0.018310187850147486, + "weighted_rkd_loss": 0.012961483839899302 + }, + { + "epoch": 8.295024730869944, + "flow_matching_loss": 0.018701092526316642, + "grad_norm": 0.22439639270305634, + "learning_rate": 2.7205568599404927e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00645536482334137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00278996869456023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015700698550790548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8671191573143006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28510, + "weighted_flow_matching_loss": 0.018701092526316642, + "weighted_rkd_loss": 0.013614360988140107 + }, + { + "epoch": 8.297934244981088, + "flow_matching_loss": 0.02025266094133258, + "grad_norm": 0.20905324816703796, + "learning_rate": 2.7203165319216047e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005877111200243234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002616921067237854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014371143374592065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8670302748680114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28520, + "weighted_flow_matching_loss": 0.02025266094133258, + "weighted_rkd_loss": 0.012460216972976923 + }, + { + "epoch": 8.300843759092231, + "flow_matching_loss": 0.018613900523632765, + "grad_norm": 0.17962509393692017, + "learning_rate": 2.72007611122773e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005841761408373713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023700016550719737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014053524564951658, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8669413685798645, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28530, + "weighted_flow_matching_loss": 0.018613900523632765, + "weighted_rkd_loss": 0.01218356117606163 + }, + { + "epoch": 8.303753273203375, + "flow_matching_loss": 0.019129542354494333, + "grad_norm": 0.23304767906665802, + "learning_rate": 2.7198355978771264e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006381835136562586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002800502604804933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015564173087477684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8668524384498596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28540, + "weighted_flow_matching_loss": 0.019129542354494333, + "weighted_rkd_loss": 0.01349185910075903 + }, + { + "epoch": 8.306662787314519, + "flow_matching_loss": 0.020788996387273072, + "grad_norm": 0.21731531620025635, + "learning_rate": 2.719594991888059e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065450385678559545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002858033333905041, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0159481106325984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8667634844779968, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28550, + "weighted_flow_matching_loss": 0.020788996387273072, + "weighted_rkd_loss": 0.013823232520371676 + }, + { + "epoch": 8.309572301425662, + "flow_matching_loss": 0.02005237601697445, + "grad_norm": 0.23168762028217316, + "learning_rate": 2.7193542932788008e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006831345846876502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028129610465839503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016475652903318407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8666745007038117, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28560, + "weighted_flow_matching_loss": 0.02005237601697445, + "weighted_rkd_loss": 0.014278980437666178 + }, + { + "epoch": 8.312481815536806, + "flow_matching_loss": 0.020135836489498616, + "grad_norm": 0.22288259863853455, + "learning_rate": 2.719113502067631e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006593878916464746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028794320882298053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016067189909517766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8665854930877686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28570, + "weighted_flow_matching_loss": 0.020135836489498616, + "weighted_rkd_loss": 0.013923558220267297 + }, + { + "epoch": 8.315391329647948, + "flow_matching_loss": 0.01988502610474825, + "grad_norm": 0.2700008153915405, + "learning_rate": 2.718872618272836e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007063162699341774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030832896241918205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01720961481332779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.866496467590332, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28580, + "weighted_flow_matching_loss": 0.01988502610474825, + "weighted_rkd_loss": 0.014912021718919278 + }, + { + "epoch": 8.318300843759092, + "flow_matching_loss": 0.019730457942932843, + "grad_norm": 0.2521508038043976, + "learning_rate": 2.71863164191271e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006149818375706673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002653079736046493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014952716697007418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8664074122905732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28590, + "weighted_flow_matching_loss": 0.019730457942932843, + "weighted_rkd_loss": 0.012955124862492085 + }, + { + "epoch": 8.321210357870235, + "flow_matching_loss": 0.01977886902168393, + "grad_norm": 0.19708780944347382, + "learning_rate": 2.7183905730055528e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006092954194173217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026514865458011627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014837395027279853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8663183331489563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28600, + "weighted_flow_matching_loss": 0.01977886902168393, + "weighted_rkd_loss": 0.012853913474828005 + }, + { + "epoch": 8.32411987198138, + "flow_matching_loss": 0.020002527069300412, + "grad_norm": 0.2752445638179779, + "learning_rate": 2.7181494115696722e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006605761963874102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030827053473331035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016294229310005903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8662292301654816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28610, + "weighted_flow_matching_loss": 0.020002527069300412, + "weighted_rkd_loss": 0.014114473573863506 + }, + { + "epoch": 8.327029386092523, + "flow_matching_loss": 0.017608735151588915, + "grad_norm": 0.2266458123922348, + "learning_rate": 2.717908157623383e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006758229155093431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002907497971318662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016423956397920846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.86614009141922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28620, + "weighted_flow_matching_loss": 0.017608735151588915, + "weighted_rkd_loss": 0.014225476235151292 + }, + { + "epoch": 8.329938900203667, + "flow_matching_loss": 0.019931337516754865, + "grad_norm": 0.1562991589307785, + "learning_rate": 2.7176668111850066e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005892377579584718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002641811885405332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01442656721919775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8660509347915649, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28630, + "weighted_flow_matching_loss": 0.019931337516754865, + "weighted_rkd_loss": 0.012494149431586266 + }, + { + "epoch": 8.33284841431481, + "flow_matching_loss": 0.020312981214374304, + "grad_norm": 0.1671592742204666, + "learning_rate": 2.717425372272872e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006813903711736203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031380216125398873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01676582908257842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8659617602825165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28640, + "weighted_flow_matching_loss": 0.020312981214374304, + "weighted_rkd_loss": 0.014518567733466626 + }, + { + "epoch": 8.335757928425952, + "flow_matching_loss": 0.020803857035934924, + "grad_norm": 0.1937045454978943, + "learning_rate": 2.7171838409053142e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005477643571794033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024005663814023136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013355853408575058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8658725500106812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28650, + "weighted_flow_matching_loss": 0.020803857035934924, + "weighted_rkd_loss": 0.011564468592405319 + }, + { + "epoch": 8.338667442537096, + "flow_matching_loss": 0.020183939766138793, + "grad_norm": 0.20135125517845154, + "learning_rate": 2.7169422171006768e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006540153548121453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002884623338468373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015964930411428213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8657833158969879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28660, + "weighted_flow_matching_loss": 0.020183939766138793, + "weighted_rkd_loss": 0.013822142221033572 + }, + { + "epoch": 8.34157695664824, + "flow_matching_loss": 0.020070656761527063, + "grad_norm": 0.21006032824516296, + "learning_rate": 2.7167005008773093e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006759746558964252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029323393246158957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01645183265209198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8656940579414367, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28670, + "weighted_flow_matching_loss": 0.020070656761527063, + "weighted_rkd_loss": 0.014242264349013567 + }, + { + "epoch": 8.344486470759383, + "flow_matching_loss": 0.02089329995214939, + "grad_norm": 0.21445870399475098, + "learning_rate": 2.7164586922535678e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005974593735300004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002757116872817278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014706304483115672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8656047761440278, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28680, + "weighted_flow_matching_loss": 0.02089329995214939, + "weighted_rkd_loss": 0.01272984966635704 + }, + { + "epoch": 8.347395984870527, + "flow_matching_loss": 0.01844272846356034, + "grad_norm": 0.19455884397029877, + "learning_rate": 2.7162167912478167e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005574475973844528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002431140071712434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013580091949552297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8655154705047607, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28690, + "weighted_flow_matching_loss": 0.01844272846356034, + "weighted_rkd_loss": 0.011753797624260187 + }, + { + "epoch": 8.35030549898167, + "flow_matching_loss": 0.019919618591666223, + "grad_norm": 0.20498685538768768, + "learning_rate": 2.715974797878427e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006184007227420807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027695939759723844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015137608628720045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8654261410236359, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28700, + "weighted_flow_matching_loss": 0.019919618591666223, + "weighted_rkd_loss": 0.013100488856434823 + }, + { + "epoch": 8.353215013092813, + "flow_matching_loss": 0.019947803020477294, + "grad_norm": 0.21179550886154175, + "learning_rate": 2.715732712163776e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00629433854483068, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027836870169267057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015372363943606615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8653367757797241, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28710, + "weighted_flow_matching_loss": 0.019947803020477294, + "weighted_rkd_loss": 0.013302224222570658 + }, + { + "epoch": 8.356124527203956, + "flow_matching_loss": 0.017794670443981885, + "grad_norm": 0.21077664196491241, + "learning_rate": 2.7154905341222483e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006423428701236844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028724662377499046, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01571932351216674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8652474105358123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28720, + "weighted_flow_matching_loss": 0.017794670443981885, + "weighted_rkd_loss": 0.013601124193519354 + }, + { + "epoch": 8.3590340413151, + "flow_matching_loss": 0.018888263311237097, + "grad_norm": 0.20204409956932068, + "learning_rate": 2.7152482637722363e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006112234573811293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002543673128820956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014768142346292734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8651579976081848, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28730, + "weighted_flow_matching_loss": 0.018888263311237097, + "weighted_rkd_loss": 0.012776755727827549 + }, + { + "epoch": 8.361943555426244, + "flow_matching_loss": 0.018762756884098054, + "grad_norm": 0.18003058433532715, + "learning_rate": 2.7150059011321385e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005824799556285143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025482317665591835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014197830948978663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8650685548782349, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28740, + "weighted_flow_matching_loss": 0.018762756884098054, + "weighted_rkd_loss": 0.012282097898423671 + }, + { + "epoch": 8.364853069537387, + "flow_matching_loss": 0.02093473430722952, + "grad_norm": 0.26486504077911377, + "learning_rate": 2.714763446220361e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006650342978537083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002843706822022796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016144392918795346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8649791061878205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28750, + "weighted_flow_matching_loss": 0.02093473430722952, + "weighted_rkd_loss": 0.013964561466127633 + }, + { + "epoch": 8.367762583648531, + "flow_matching_loss": 0.018117800075560807, + "grad_norm": 0.17558395862579346, + "learning_rate": 2.7145208990553164e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007561877695843577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003390413289889693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018514168541878463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8648896217346191, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28760, + "weighted_flow_matching_loss": 0.018117800075560807, + "weighted_rkd_loss": 0.016012719087302684 + }, + { + "epoch": 8.370672097759675, + "flow_matching_loss": 0.01838029520586133, + "grad_norm": 0.23539239168167114, + "learning_rate": 2.714278259655425e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057274349965155125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025533861247822643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014008256047964096, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.86480011343956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28770, + "weighted_flow_matching_loss": 0.01838029520586133, + "weighted_rkd_loss": 0.012114334851503372 + }, + { + "epoch": 8.373581611870817, + "flow_matching_loss": 0.019488852005451918, + "grad_norm": 0.20404134690761566, + "learning_rate": 2.714035528039113e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006669843290001154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002944520954042673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016284207440912725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8647105813026428, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28780, + "weighted_flow_matching_loss": 0.019488852005451918, + "weighted_rkd_loss": 0.014081109687685967 + }, + { + "epoch": 8.37649112598196, + "flow_matching_loss": 0.01925689969211817, + "grad_norm": 0.23673667013645172, + "learning_rate": 2.7137927042248144e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006222725752741099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002751345746219158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015196797251701356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8646210253238678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28790, + "weighted_flow_matching_loss": 0.01925689969211817, + "weighted_rkd_loss": 0.01313944635912776 + }, + { + "epoch": 8.379400640093104, + "flow_matching_loss": 0.017987411469221115, + "grad_norm": 0.19487793743610382, + "learning_rate": 2.7135497882309704e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006246456457301974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024791107163764535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014972023572772742, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8645314455032349, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28800, + "weighted_flow_matching_loss": 0.017987411469221115, + "weighted_rkd_loss": 0.012943787965923548 + }, + { + "epoch": 8.382310154204248, + "flow_matching_loss": 0.01970006413757801, + "grad_norm": 0.1847398579120636, + "learning_rate": 2.713306780076029e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007008744543418289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003133322694338858, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017150811664760113, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.864441841840744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28810, + "weighted_flow_matching_loss": 0.01970006413757801, + "weighted_rkd_loss": 0.01482584085315466 + }, + { + "epoch": 8.385219668315392, + "flow_matching_loss": 0.019731539860367776, + "grad_norm": 0.19783534109592438, + "learning_rate": 2.7130636797784445e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006587431905791164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002926080906763673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01610094467177987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8643522024154663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28820, + "weighted_flow_matching_loss": 0.019731539860367776, + "weighted_rkd_loss": 0.01391694499179721 + }, + { + "epoch": 8.388129182426535, + "flow_matching_loss": 0.017348217964172363, + "grad_norm": 0.2233390212059021, + "learning_rate": 2.712820487356679e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006894966820254922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029237986076623202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016713732201606036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8642625451087952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28830, + "weighted_flow_matching_loss": 0.017348217964172363, + "weighted_rkd_loss": 0.014445035438984632 + }, + { + "epoch": 8.391038696537677, + "flow_matching_loss": 0.019112709816545247, + "grad_norm": 0.19542770087718964, + "learning_rate": 2.7125772028292012e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006381914112716913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026186746777966618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015382502973079682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8641728699207306, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28840, + "weighted_flow_matching_loss": 0.019112709816545247, + "weighted_rkd_loss": 0.013293131161481142 + }, + { + "epoch": 8.393948210648821, + "flow_matching_loss": 0.018837887234985827, + "grad_norm": 0.20852887630462646, + "learning_rate": 2.7123338262144878e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060123618924990295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002730968501418829, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014755692332983017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8640831589698792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28850, + "weighted_flow_matching_loss": 0.018837887234985827, + "weighted_rkd_loss": 0.012750175409018993 + }, + { + "epoch": 8.396857724759965, + "flow_matching_loss": 0.017899296153336762, + "grad_norm": 0.16974034905433655, + "learning_rate": 2.7120903575310207e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006353719299659133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002747778152115643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015455216728150844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8639934241771698, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28860, + "weighted_flow_matching_loss": 0.017899296153336762, + "weighted_rkd_loss": 0.013353229966014624 + }, + { + "epoch": 8.399767238871108, + "flow_matching_loss": 0.020117984898388386, + "grad_norm": 0.16866303980350494, + "learning_rate": 2.71184679679729e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005980793479830026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026111736660823226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014572760462760926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.863903671503067, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28870, + "weighted_flow_matching_loss": 0.020117984898388386, + "weighted_rkd_loss": 0.01258946182206273 + }, + { + "epoch": 8.402676752982252, + "flow_matching_loss": 0.018104763142764567, + "grad_norm": 0.19027678668498993, + "learning_rate": 2.7116031440317925e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006239067763090134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00285831238143146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015336447954177856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8638138890266418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28880, + "weighted_flow_matching_loss": 0.018104763142764567, + "weighted_rkd_loss": 0.013247833680361509 + }, + { + "epoch": 8.405586267093396, + "flow_matching_loss": 0.01966632083058357, + "grad_norm": 0.21502329409122467, + "learning_rate": 2.7113593992530322e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006816660752519965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032324664178304376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016865788120776416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8637240827083588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28890, + "weighted_flow_matching_loss": 0.01966632083058357, + "weighted_rkd_loss": 0.014567353297024965 + }, + { + "epoch": 8.40849578120454, + "flow_matching_loss": 0.01957721309736371, + "grad_norm": 0.2212693840265274, + "learning_rate": 2.7111155624795198e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007557970937341452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033628908451646565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01847883276641369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8636342525482178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28900, + "weighted_flow_matching_loss": 0.01957721309736371, + "weighted_rkd_loss": 0.01595892049372196 + }, + { + "epoch": 8.411405295315681, + "flow_matching_loss": 0.01693766498938203, + "grad_norm": 0.2065090835094452, + "learning_rate": 2.7108716337297733e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006671422300860286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030563509790226817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016399195604026317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8635443866252899, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28910, + "weighted_flow_matching_loss": 0.01693766498938203, + "weighted_rkd_loss": 0.014161414187401532 + }, + { + "epoch": 8.414314809426825, + "flow_matching_loss": 0.020004493091255425, + "grad_norm": 0.1953296810388565, + "learning_rate": 2.7106276130223167e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062815284356474875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002593500085640699, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015156556852161884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8634545087814331, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28920, + "weighted_flow_matching_loss": 0.020004493091255425, + "weighted_rkd_loss": 0.013086992222815752 + }, + { + "epoch": 8.417224323537969, + "flow_matching_loss": 0.019423331040889023, + "grad_norm": 0.2150023877620697, + "learning_rate": 2.7103835003756827e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00620653023943305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026342209661379457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015047281421720982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8633646011352539, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28930, + "weighted_flow_matching_loss": 0.019423331040889023, + "weighted_rkd_loss": 0.012991272378712893 + }, + { + "epoch": 8.420133837649113, + "flow_matching_loss": 0.01804345566779375, + "grad_norm": 0.19713984429836273, + "learning_rate": 2.71013929580841e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062385460827499625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002624577307142317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015101669356226921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8632746696472168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28940, + "weighted_flow_matching_loss": 0.01804345566779375, + "weighted_rkd_loss": 0.013036897033452987 + }, + { + "epoch": 8.423043351760256, + "flow_matching_loss": 0.01865355661138892, + "grad_norm": 0.20324799418449402, + "learning_rate": 2.7098949993390435e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00670357714407146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002929649921134114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016336804255843163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8631847202777863, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28950, + "weighted_flow_matching_loss": 0.01865355661138892, + "weighted_rkd_loss": 0.014101671427488327 + }, + { + "epoch": 8.4259528658714, + "flow_matching_loss": 0.018658216390758754, + "grad_norm": 0.2335265725851059, + "learning_rate": 2.7096506109861363e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006788813136518001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002910175058059394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01648780135437846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8630947172641754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28960, + "weighted_flow_matching_loss": 0.018658216390758754, + "weighted_rkd_loss": 0.01423049308359623 + }, + { + "epoch": 8.428862379982544, + "flow_matching_loss": 0.019994525145739316, + "grad_norm": 0.1665036827325821, + "learning_rate": 2.7094061307682484e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005930407531559467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024759780615568163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014336793124675751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8630047142505646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28970, + "weighted_flow_matching_loss": 0.019994525145739316, + "weighted_rkd_loss": 0.012372712232172489 + }, + { + "epoch": 8.431771894093686, + "flow_matching_loss": 0.017709142435342072, + "grad_norm": 0.19121618568897247, + "learning_rate": 2.7091615587039468e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007097100233659148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002979965112172067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017174165602773427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8629147052764893, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28980, + "weighted_flow_matching_loss": 0.017709142435342072, + "weighted_rkd_loss": 0.01481991382315755 + }, + { + "epoch": 8.43468140820483, + "flow_matching_loss": 0.019018519297242166, + "grad_norm": 0.19998015463352203, + "learning_rate": 2.708916894811804e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006415929784998298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002701308252289891, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01553316805511713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8628246426582337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28990, + "weighted_flow_matching_loss": 0.019018519297242166, + "weighted_rkd_loss": 0.013402405008673669 + }, + { + "epoch": 8.437590922315973, + "flow_matching_loss": 0.020882943645119667, + "grad_norm": 0.22766689956188202, + "learning_rate": 2.7086721391104016e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006368104973807931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027942488552071152, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015530458558350802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8627345621585846, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29000, + "weighted_flow_matching_loss": 0.020882943645119667, + "weighted_rkd_loss": 0.013398689962923527 + }, + { + "epoch": 8.440500436427117, + "flow_matching_loss": 0.019874971732497214, + "grad_norm": 0.21145492792129517, + "learning_rate": 2.708427291618326e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006345114530995488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031061118002980947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01579634062945843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8626444578170777, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29010, + "weighted_flow_matching_loss": 0.019874971732497214, + "weighted_rkd_loss": 0.013626634702086448 + }, + { + "epoch": 8.44340995053826, + "flow_matching_loss": 0.019125676620751618, + "grad_norm": 0.2947019040584564, + "learning_rate": 2.7081823523541736e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006078867521136999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026531660812906923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014810901321470737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8625543296337128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29020, + "weighted_flow_matching_loss": 0.019125676620751618, + "weighted_rkd_loss": 0.01277520740404725 + }, + { + "epoch": 8.446319464649404, + "flow_matching_loss": 0.018509464617818595, + "grad_norm": 0.1798201948404312, + "learning_rate": 2.707937321336544e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005895450152456761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00255982109811157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014350721519440413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.862464165687561, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29030, + "weighted_flow_matching_loss": 0.018509464617818595, + "weighted_rkd_loss": 0.012377005722373723 + }, + { + "epoch": 8.449228978760548, + "flow_matching_loss": 0.01936515821143985, + "grad_norm": 0.18406470119953156, + "learning_rate": 2.7076921985840467e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006429777992889285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002811212744563818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015670768544077875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8623739898204803, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29040, + "weighted_flow_matching_loss": 0.01936515821143985, + "weighted_rkd_loss": 0.01351406928151846 + }, + { + "epoch": 8.45213849287169, + "flow_matching_loss": 0.019023192301392555, + "grad_norm": 0.18400467932224274, + "learning_rate": 2.7074469841152975e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006016541714780032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002583815052639693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014616898261010647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8622837841510773, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29050, + "weighted_flow_matching_loss": 0.019023192301392555, + "weighted_rkd_loss": 0.012603928707540036 + }, + { + "epoch": 8.455048006982834, + "flow_matching_loss": 0.018183587305247783, + "grad_norm": 0.23948374390602112, + "learning_rate": 2.7072016779489176e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006074079545214772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026370859588496385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01478524524718523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8621935546398163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29060, + "weighted_flow_matching_loss": 0.018183587305247783, + "weighted_rkd_loss": 0.012747711967676878 + }, + { + "epoch": 8.457957521093977, + "flow_matching_loss": 0.020181604847311973, + "grad_norm": 0.15793314576148987, + "learning_rate": 2.7069562801035374e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00581997474655509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025702493963763117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0142101988196373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8621033012866974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29070, + "weighted_flow_matching_loss": 0.020181604847311973, + "weighted_rkd_loss": 0.012250711303204298 + }, + { + "epoch": 8.460867035205121, + "flow_matching_loss": 0.020524200052022934, + "grad_norm": 0.21015147864818573, + "learning_rate": 2.706710790597793e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006258849264122545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002701911807525903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015219610556960106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8620130121707916, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29080, + "weighted_flow_matching_loss": 0.020524200052022934, + "weighted_rkd_loss": 0.01311952481046319 + }, + { + "epoch": 8.463776549316265, + "flow_matching_loss": 0.018609500490128994, + "grad_norm": 0.18770484626293182, + "learning_rate": 2.7064652094503277e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006443994538858533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025268251658417283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015414814371615649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8619227111339569, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29090, + "weighted_flow_matching_loss": 0.018609500490128994, + "weighted_rkd_loss": 0.013286360446363688 + }, + { + "epoch": 8.466686063427408, + "flow_matching_loss": 0.018380671739578247, + "grad_norm": 0.2321418821811676, + "learning_rate": 2.7062195366797917e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006535264384001494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002798985387198627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015869513992220162, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8618323922157287, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29100, + "weighted_flow_matching_loss": 0.018380671739578247, + "weighted_rkd_loss": 0.013676817622035741 + }, + { + "epoch": 8.46959557753855, + "flow_matching_loss": 0.017209593672305346, + "grad_norm": 0.20950907468795776, + "learning_rate": 2.705973772304842e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007015542802400887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003241245134267956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017272330727428196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8617420315742492, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29110, + "weighted_flow_matching_loss": 0.017209593672305346, + "weighted_rkd_loss": 0.01488424399867654 + }, + { + "epoch": 8.472505091649694, + "flow_matching_loss": 0.021136876102536915, + "grad_norm": 0.2932283878326416, + "learning_rate": 2.7057279163441433e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006199165666475892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002653918508440256, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01505224984139204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8616516530513764, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29120, + "weighted_flow_matching_loss": 0.021136876102536915, + "weighted_rkd_loss": 0.012969777081161737 + }, + { + "epoch": 8.475414605760838, + "flow_matching_loss": 0.01973234545439482, + "grad_norm": 0.2714148461818695, + "learning_rate": 2.7054819688163664e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006758270878344774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029010112397372723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016417553275823595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8615612506866455, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29130, + "weighted_flow_matching_loss": 0.01973234545439482, + "weighted_rkd_loss": 0.01414470924064517 + }, + { + "epoch": 8.478324119871981, + "flow_matching_loss": 0.01984187588095665, + "grad_norm": 0.20934052765369415, + "learning_rate": 2.7052359297401897e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006170432921499014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002637016959488392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014977882802486419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8614708244800567, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29140, + "weighted_flow_matching_loss": 0.01984187588095665, + "weighted_rkd_loss": 0.012902977876365185 + }, + { + "epoch": 8.481233633983125, + "flow_matching_loss": 0.0196000081487, + "grad_norm": 0.20473478734493256, + "learning_rate": 2.7049897991342976e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006046655774116516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00290159429423511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014994906168431044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8613803744316101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29150, + "weighted_flow_matching_loss": 0.0196000081487, + "weighted_rkd_loss": 0.012916321121156215 + }, + { + "epoch": 8.484143148094269, + "flow_matching_loss": 0.019360191002488137, + "grad_norm": 0.1786530613899231, + "learning_rate": 2.7047435770173825e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006985925603657961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031525643309578298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017124415654689074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.861289894580841, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29160, + "weighted_flow_matching_loss": 0.019360191002488137, + "weighted_rkd_loss": 0.014749087020754814 + }, + { + "epoch": 8.487052662205413, + "flow_matching_loss": 0.0184734714217484, + "grad_norm": 0.22025351226329803, + "learning_rate": 2.7044972634081434e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005947039695456624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002595901966560632, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014489981438964605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8611993968486786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29170, + "weighted_flow_matching_loss": 0.0184734714217484, + "weighted_rkd_loss": 0.01247870996594429 + }, + { + "epoch": 8.489962176316554, + "flow_matching_loss": 0.01812228122726083, + "grad_norm": 0.2197939157485962, + "learning_rate": 2.7042508583252863e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006812921073287725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031207035761326553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01674654595553875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8611088633537293, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29180, + "weighted_flow_matching_loss": 0.01812228122726083, + "weighted_rkd_loss": 0.014420589338988066 + }, + { + "epoch": 8.492871690427698, + "flow_matching_loss": 0.018038470298051834, + "grad_norm": 0.21334145963191986, + "learning_rate": 2.7040043617875237e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006321565713733435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002818811289034784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015461942832916974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8610183000564575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29190, + "weighted_flow_matching_loss": 0.018038470298051834, + "weighted_rkd_loss": 0.013313009217381478 + }, + { + "epoch": 8.495781204538842, + "flow_matching_loss": 0.019193854276090862, + "grad_norm": 0.20181886851787567, + "learning_rate": 2.7037577738135756e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005721291992813349, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023960522608831524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0138386364094913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8609277546405792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29200, + "weighted_flow_matching_loss": 0.019193854276090862, + "weighted_rkd_loss": 0.01191402431577444 + }, + { + "epoch": 8.498690718649986, + "flow_matching_loss": 0.01978927068412304, + "grad_norm": 0.3220616281032562, + "learning_rate": 2.7035110944221686e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006187785184010863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00278061181306839, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015156182087957859, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8608371436595916, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29210, + "weighted_flow_matching_loss": 0.01978927068412304, + "weighted_rkd_loss": 0.0130469661206007 + }, + { + "epoch": 8.50160023276113, + "flow_matching_loss": 0.018830831069499253, + "grad_norm": 0.19610969722270966, + "learning_rate": 2.703264323632037e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065881532151252035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030488722142763434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016225178446620702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8607465028762817, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29220, + "weighted_flow_matching_loss": 0.018830831069499253, + "weighted_rkd_loss": 0.013965834304690361 + }, + { + "epoch": 8.504509746872273, + "flow_matching_loss": 0.018839992582798004, + "grad_norm": 0.2153119146823883, + "learning_rate": 2.7030174614619202e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006230329908430576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025105511071160437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01497121099382639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8606558561325073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29230, + "weighted_flow_matching_loss": 0.018839992582798004, + "weighted_rkd_loss": 0.012885058298707008 + }, + { + "epoch": 8.507419260983415, + "flow_matching_loss": 0.02002890342846513, + "grad_norm": 0.20303839445114136, + "learning_rate": 2.7027705079305668e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006084883166477084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026420745300129054, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014811841025948524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8605651795864105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29240, + "weighted_flow_matching_loss": 0.02002890342846513, + "weighted_rkd_loss": 0.012746592052280903 + }, + { + "epoch": 8.510328775094559, + "flow_matching_loss": 0.01920518893748522, + "grad_norm": 0.20721207559108734, + "learning_rate": 2.702523463056731e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006396175688132643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002891677641309798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015684028808027505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8604744851589203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29250, + "weighted_flow_matching_loss": 0.01920518893748522, + "weighted_rkd_loss": 0.013495721016079187 + }, + { + "epoch": 8.513238289205702, + "flow_matching_loss": 0.021124326344579458, + "grad_norm": 0.17476288974285126, + "learning_rate": 2.702276326859174e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006629142956808209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002970088820438832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016228374931961297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8603837549686432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29260, + "weighted_flow_matching_loss": 0.021124326344579458, + "weighted_rkd_loss": 0.013962697144597769 + }, + { + "epoch": 8.516147803316846, + "flow_matching_loss": 0.01781719774007797, + "grad_norm": 0.20389816164970398, + "learning_rate": 2.7020290993566637e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061736516654491425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026626643724739552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015009967889636756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8602930068969726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29270, + "weighted_flow_matching_loss": 0.01781719774007797, + "weighted_rkd_loss": 0.012912958208471537 + }, + { + "epoch": 8.51905731742799, + "flow_matching_loss": 0.019851594045758248, + "grad_norm": 0.2059796303510666, + "learning_rate": 2.701781780567977e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006210006820037961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026882918318733573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015108305774629117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8602022349834442, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29280, + "weighted_flow_matching_loss": 0.019851594045758248, + "weighted_rkd_loss": 0.012996206618845462 + }, + { + "epoch": 8.521966831539133, + "flow_matching_loss": 0.01842493386939168, + "grad_norm": 0.17588435113430023, + "learning_rate": 2.7015343705118946e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005710372235625982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025679090758785605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013988653663545847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8601114332675934, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29290, + "weighted_flow_matching_loss": 0.01842493386939168, + "weighted_rkd_loss": 0.012031817808747292 + }, + { + "epoch": 8.524876345650277, + "flow_matching_loss": 0.01953373719006777, + "grad_norm": 0.21459591388702393, + "learning_rate": 2.7012868692072064e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007579786889255047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003211087384261191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018370660953223705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8600206136703491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29300, + "weighted_flow_matching_loss": 0.01953373719006777, + "weighted_rkd_loss": 0.01579910609871149 + }, + { + "epoch": 8.52778585976142, + "flow_matching_loss": 0.021200418658554553, + "grad_norm": 0.20685400068759918, + "learning_rate": 2.7010392766727078e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063543982803821565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027945936657488347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015503390226513147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.859929758310318, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29310, + "weighted_flow_matching_loss": 0.021200418658554553, + "weighted_rkd_loss": 0.013331844750791789 + }, + { + "epoch": 8.530695373872563, + "flow_matching_loss": 0.020204611402004958, + "grad_norm": 0.19325891137123108, + "learning_rate": 2.700791592927203e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006265314808115363, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028461327543482184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015376762207597493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.859838891029358, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29320, + "weighted_flow_matching_loss": 0.020204611402004958, + "weighted_rkd_loss": 0.01322147697210312 + }, + { + "epoch": 8.533604887983707, + "flow_matching_loss": 0.019466884899884462, + "grad_norm": 0.17411497235298157, + "learning_rate": 2.700543817989501e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006844024360179901, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031315648462623357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016819613892585038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8597479879856109, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29330, + "weighted_flow_matching_loss": 0.019466884899884462, + "weighted_rkd_loss": 0.014460639376193286 + }, + { + "epoch": 8.53651440209485, + "flow_matching_loss": 0.019175849854946136, + "grad_norm": 0.1707477569580078, + "learning_rate": 2.7002959518784183e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006461214087903499, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002786428760737181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01570885693654418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8596570670604706, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29340, + "weighted_flow_matching_loss": 0.019175849854946136, + "weighted_rkd_loss": 0.013504249230027198 + }, + { + "epoch": 8.539423916205994, + "flow_matching_loss": 0.0191127885133028, + "grad_norm": 0.19061324000358582, + "learning_rate": 2.7000479946127805e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005971187725663185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026620812946930527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014604456815868616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8595661222934723, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29350, + "weighted_flow_matching_loss": 0.0191127885133028, + "weighted_rkd_loss": 0.012553504202514887 + }, + { + "epoch": 8.542333430317138, + "flow_matching_loss": 0.020775934122502802, + "grad_norm": 0.19680631160736084, + "learning_rate": 2.699799946211416e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006025253934785724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026080611161887644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014658568799495697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8594751477241516, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29360, + "weighted_flow_matching_loss": 0.020775934122502802, + "weighted_rkd_loss": 0.012598714977502822 + }, + { + "epoch": 8.545242944428281, + "flow_matching_loss": 0.02135053090751171, + "grad_norm": 0.19499415159225464, + "learning_rate": 2.6995518066931648e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00688704033382237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032577642006799577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01703184489160776, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8593841552734375, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29370, + "weighted_flow_matching_loss": 0.02135053090751171, + "weighted_rkd_loss": 0.014636874292045831 + }, + { + "epoch": 8.548152458539423, + "flow_matching_loss": 0.02042866200208664, + "grad_norm": 0.17095111310482025, + "learning_rate": 2.6993035760768692e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006211159145459532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027649671072140338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015187285654246808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.859293133020401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29380, + "weighted_flow_matching_loss": 0.02042866200208664, + "weighted_rkd_loss": 0.01305032754316926 + }, + { + "epoch": 8.551061972650567, + "flow_matching_loss": 0.019563399627804755, + "grad_norm": 0.19469612836837769, + "learning_rate": 2.699055254381382e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005871541984379292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002495383704081178, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01423846771940589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8592020809650421, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29390, + "weighted_flow_matching_loss": 0.019563399627804755, + "weighted_rkd_loss": 0.012233742512762547 + }, + { + "epoch": 8.55397148676171, + "flow_matching_loss": 0.018459732178598642, + "grad_norm": 0.17909373342990875, + "learning_rate": 2.6988068416255615e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00596096133813262, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002647147118113935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014569069910794496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8591110110282898, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29400, + "weighted_flow_matching_loss": 0.018459732178598642, + "weighted_rkd_loss": 0.012516444642096758 + }, + { + "epoch": 8.556881000872854, + "flow_matching_loss": 0.01842951476573944, + "grad_norm": 0.19372405111789703, + "learning_rate": 2.6985583378282727e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006023216107860207, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027691900031641126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014815622009336948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8590199232101441, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29410, + "weighted_flow_matching_loss": 0.01842951476573944, + "weighted_rkd_loss": 0.012726911436766386 + }, + { + "epoch": 8.559790514983998, + "flow_matching_loss": 0.020445468183606864, + "grad_norm": 0.17489558458328247, + "learning_rate": 2.6983097430083877e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005974809499457478, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023764274083077908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014326046500355006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8589287996292114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29420, + "weighted_flow_matching_loss": 0.020445468183606864, + "weighted_rkd_loss": 0.012305036839097738 + }, + { + "epoch": 8.562700029095142, + "flow_matching_loss": 0.019120394997298718, + "grad_norm": 0.19261795282363892, + "learning_rate": 2.698061057184786e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005826376844197512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025058646453544496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01415861826390028, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8588376522064209, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29430, + "weighted_flow_matching_loss": 0.019120394997298718, + "weighted_rkd_loss": 0.01215995280072093 + }, + { + "epoch": 8.565609543206284, + "flow_matching_loss": 0.019958303216844796, + "grad_norm": 0.21908487379550934, + "learning_rate": 2.6978122803763538e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005942392535507679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002531835320405662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014416620414704085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8587464988231659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29440, + "weighted_flow_matching_loss": 0.019958303216844796, + "weighted_rkd_loss": 0.01238022642210126 + }, + { + "epoch": 8.568519057317427, + "flow_matching_loss": 0.019629364926368, + "grad_norm": 0.17798399925231934, + "learning_rate": 2.6975634126019833e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006341360462829471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026895090704783797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015372230019420386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8586553037166595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29450, + "weighted_flow_matching_loss": 0.019629364926368, + "weighted_rkd_loss": 0.013199438620358706 + }, + { + "epoch": 8.571428571428571, + "flow_matching_loss": 0.01847354555502534, + "grad_norm": 0.23171141743659973, + "learning_rate": 2.6973144538805753e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00666679902933538, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029178993660025298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016251497343182562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8585640788078308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29460, + "weighted_flow_matching_loss": 0.01847354555502534, + "weighted_rkd_loss": 0.013952924497425556 + }, + { + "epoch": 8.574338085539715, + "flow_matching_loss": 0.017509301379323005, + "grad_norm": 0.24477367103099823, + "learning_rate": 2.697065404231036e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006402275525033474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028339656884782015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01563851647078991, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8584728360176086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29470, + "weighted_flow_matching_loss": 0.017509301379323005, + "weighted_rkd_loss": 0.013425251096487045 + }, + { + "epoch": 8.577247599650859, + "flow_matching_loss": 0.017985961586236953, + "grad_norm": 0.1834772378206253, + "learning_rate": 2.696816263672279e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006586514413356781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028594078263267874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01603243676945567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8583815693855286, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29480, + "weighted_flow_matching_loss": 0.017985961586236953, + "weighted_rkd_loss": 0.013761923369020223 + }, + { + "epoch": 8.580157113762002, + "flow_matching_loss": 0.019881158228963615, + "grad_norm": 0.24072901904582977, + "learning_rate": 2.6965670322232254e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006339619401842356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028011509450152515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015480389818549156, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8582902729511261, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29490, + "weighted_flow_matching_loss": 0.019881158228963615, + "weighted_rkd_loss": 0.013286649063229562 + }, + { + "epoch": 8.583066627873146, + "flow_matching_loss": 0.019595681875944137, + "grad_norm": 0.20307953655719757, + "learning_rate": 2.696317709902802e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006034873658791184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025710386806167664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014640786033123732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8581989586353302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29500, + "weighted_flow_matching_loss": 0.019595681875944137, + "weighted_rkd_loss": 0.012564720585942268 + }, + { + "epoch": 8.585976141984288, + "flow_matching_loss": 0.020208040717989206, + "grad_norm": 0.2433297038078308, + "learning_rate": 2.6960682967299437e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005584185943007469, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023073095246218143, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013475681468844414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8581076204776764, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29510, + "weighted_flow_matching_loss": 0.020208040717989206, + "weighted_rkd_loss": 0.0115635734051466 + }, + { + "epoch": 8.588885656095432, + "flow_matching_loss": 0.01834977101534605, + "grad_norm": 0.24599885940551758, + "learning_rate": 2.6958187927235915e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062149058561772105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002790387801360339, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01522019961848855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8580162525177002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29520, + "weighted_flow_matching_loss": 0.01834977101534605, + "weighted_rkd_loss": 0.013059145957231521 + }, + { + "epoch": 8.591795170206575, + "flow_matching_loss": 0.017329176515340806, + "grad_norm": 0.20591622591018677, + "learning_rate": 2.695569197902694e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070142236072570086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003166892216540873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01719533931463957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.857924860715866, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29530, + "weighted_flow_matching_loss": 0.017329176515340806, + "weighted_rkd_loss": 0.014752291329205035 + }, + { + "epoch": 8.594704684317719, + "flow_matching_loss": 0.018652574997395276, + "grad_norm": 0.19873076677322388, + "learning_rate": 2.695319512286206e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00578266098164022, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002453059982508421, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014018381759524346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8578334510326385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29540, + "weighted_flow_matching_loss": 0.018652574997395276, + "weighted_rkd_loss": 0.012025405559688807 + }, + { + "epoch": 8.597614198428863, + "flow_matching_loss": 0.019661367312073706, + "grad_norm": 0.18203586339950562, + "learning_rate": 2.695069735893089e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059848634991794825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027803506469354033, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014750077482312917, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8577420115470886, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29550, + "weighted_flow_matching_loss": 0.019661367312073706, + "weighted_rkd_loss": 0.012651743646711112 + }, + { + "epoch": 8.600523712540006, + "flow_matching_loss": 0.02091136369854212, + "grad_norm": 0.19680598378181458, + "learning_rate": 2.6948198687423133e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005928267911076546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026140826870687307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014470618683844805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8576505482196808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29560, + "weighted_flow_matching_loss": 0.02091136369854212, + "weighted_rkd_loss": 0.012410718202590942 + }, + { + "epoch": 8.603433226651148, + "flow_matching_loss": 0.020951419416815043, + "grad_norm": 0.19735853374004364, + "learning_rate": 2.6945699108528525e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005810674792155623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002378537773620337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013999887369573116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8575590550899506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29570, + "weighted_flow_matching_loss": 0.020951419416815043, + "weighted_rkd_loss": 0.012005740497261285 + }, + { + "epoch": 8.606342740762292, + "flow_matching_loss": 0.018948242999613285, + "grad_norm": 0.21544484794139862, + "learning_rate": 2.694319862243691e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059400375466793776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027613579411990942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014641433116048574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8574675440788269, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29580, + "weighted_flow_matching_loss": 0.018948242999613285, + "weighted_rkd_loss": 0.01255459589883685 + }, + { + "epoch": 8.609252254873436, + "flow_matching_loss": 0.01796740312129259, + "grad_norm": 0.2331053614616394, + "learning_rate": 2.6940697229338178e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060310598462820055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002660504647064954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01472262404859066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8573760151863098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29590, + "weighted_flow_matching_loss": 0.01796740312129259, + "weighted_rkd_loss": 0.012622808665037155 + }, + { + "epoch": 8.61216176898458, + "flow_matching_loss": 0.017660581227391958, + "grad_norm": 0.1727108359336853, + "learning_rate": 2.6938194929422296e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061565279494971035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002648070699069649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014961126446723937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8572844505310059, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29600, + "weighted_flow_matching_loss": 0.017660581227391958, + "weighted_rkd_loss": 0.012825996521860361 + }, + { + "epoch": 8.615071283095723, + "flow_matching_loss": 0.019129461608827114, + "grad_norm": 0.2344702035188675, + "learning_rate": 2.693569172287929e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007004707679152488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030136405606754125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017023056093603373, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.857192862033844, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29610, + "weighted_flow_matching_loss": 0.019129461608827114, + "weighted_rkd_loss": 0.014592059049755335 + }, + { + "epoch": 8.617980797206867, + "flow_matching_loss": 0.01856115832924843, + "grad_norm": 0.16466085612773895, + "learning_rate": 2.693318760989927e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006335778743959963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027970116934739053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015468569193035363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8571012556552887, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29620, + "weighted_flow_matching_loss": 0.01856115832924843, + "weighted_rkd_loss": 0.01325809359550476 + }, + { + "epoch": 8.62089031131801, + "flow_matching_loss": 0.01911896960809827, + "grad_norm": 0.22754330933094025, + "learning_rate": 2.69306825906724e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006361644435673952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002832164452411234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015555453207343816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.857009619474411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29630, + "weighted_flow_matching_loss": 0.01911896960809827, + "weighted_rkd_loss": 0.013331141136586667 + }, + { + "epoch": 8.623799825429153, + "flow_matching_loss": 0.019680337235331535, + "grad_norm": 0.219078928232193, + "learning_rate": 2.692817666538892e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006016886699944735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026560466852970422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014689819980412721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8569179594516754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29640, + "weighted_flow_matching_loss": 0.019680337235331535, + "weighted_rkd_loss": 0.012587984465062619 + }, + { + "epoch": 8.626709339540296, + "flow_matching_loss": 0.019816828519105913, + "grad_norm": 0.16302210092544556, + "learning_rate": 2.6925669834239145e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057235220447182655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024036453920416533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013850689306855202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8568262755870819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29650, + "weighted_flow_matching_loss": 0.019816828519105913, + "weighted_rkd_loss": 0.011867657024413347 + }, + { + "epoch": 8.62961885365144, + "flow_matching_loss": 0.017699056677520274, + "grad_norm": 0.17381463944911957, + "learning_rate": 2.692316209741345e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006284967483952642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026542034931480883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015224138367921114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8567345619201661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29660, + "weighted_flow_matching_loss": 0.017699056677520274, + "weighted_rkd_loss": 0.01304305950179696 + }, + { + "epoch": 8.632528367762584, + "flow_matching_loss": 0.01848694821819663, + "grad_norm": 0.21337850391864777, + "learning_rate": 2.6920653455102277e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006330786226317287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002696640440262854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015358212776482106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8566428363323212, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29670, + "weighted_flow_matching_loss": 0.01848694821819663, + "weighted_rkd_loss": 0.013156489096581936 + }, + { + "epoch": 8.635437881873727, + "flow_matching_loss": 0.021073550917208195, + "grad_norm": 0.2029689997434616, + "learning_rate": 2.6918143907496148e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00621492275968194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002671066694892943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015100912377238273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.856551069021225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29680, + "weighted_flow_matching_loss": 0.021073550917208195, + "weighted_rkd_loss": 0.012934685312211513 + }, + { + "epoch": 8.638347395984871, + "flow_matching_loss": 0.018070355337113143, + "grad_norm": 0.16520661115646362, + "learning_rate": 2.6915633454785637e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005921210162341595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002539543365128338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014381963666528464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8564593195915222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29690, + "weighted_flow_matching_loss": 0.018070355337113143, + "weighted_rkd_loss": 0.012317572627216578 + }, + { + "epoch": 8.641256910096015, + "flow_matching_loss": 0.018588382937014102, + "grad_norm": 0.1768079698085785, + "learning_rate": 2.6913122097161403e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005851040175184607, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025527307647280396, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014254811126738787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8563674747943878, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29700, + "weighted_flow_matching_loss": 0.018588382937014102, + "weighted_rkd_loss": 0.012207369599491358 + }, + { + "epoch": 8.644166424207157, + "flow_matching_loss": 0.02174352388828993, + "grad_norm": 0.1953853815793991, + "learning_rate": 2.691060983481417e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005764927063137293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002605494251474738, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01413534851744771, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8562756597995758, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29710, + "weighted_flow_matching_loss": 0.02174352388828993, + "weighted_rkd_loss": 0.012103775050491095 + }, + { + "epoch": 8.6470759383183, + "flow_matching_loss": 0.01943153003230691, + "grad_norm": 0.19510582089424133, + "learning_rate": 2.690809666793472e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006360920844599605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002700416976585984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01542225880548358, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8561838030815124, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29720, + "weighted_flow_matching_loss": 0.01943153003230691, + "weighted_rkd_loss": 0.013204289227724075 + }, + { + "epoch": 8.649985452429444, + "flow_matching_loss": 0.018088573683053256, + "grad_norm": 0.1435093879699707, + "learning_rate": 2.690558259671391e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005386078264564276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002445520670153201, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013217677362263203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8560919284820556, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29730, + "weighted_flow_matching_loss": 0.018088573683053256, + "weighted_rkd_loss": 0.011315584415569901 + }, + { + "epoch": 8.652894966540588, + "flow_matching_loss": 0.019281291309744118, + "grad_norm": 0.1963760405778885, + "learning_rate": 2.6903067621342675e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006500877207145095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00333816206548363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01633991617709398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8560000240802765, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29740, + "weighted_flow_matching_loss": 0.019281291309744118, + "weighted_rkd_loss": 0.013986976724117995 + }, + { + "epoch": 8.655804480651732, + "flow_matching_loss": 0.02042024992406368, + "grad_norm": 0.2166920006275177, + "learning_rate": 2.6900551742012006e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00621039611287415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027665789937600495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015187371335923672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8559080958366394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29750, + "weighted_flow_matching_loss": 0.02042024992406368, + "weighted_rkd_loss": 0.012998976651579142 + }, + { + "epoch": 8.658713994762875, + "flow_matching_loss": 0.019190945755690338, + "grad_norm": 0.19991159439086914, + "learning_rate": 2.6898034958912974e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064101500902324915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029644447611644866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01578474501147866, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8558161437511445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29760, + "weighted_flow_matching_loss": 0.019190945755690338, + "weighted_rkd_loss": 0.01350883962586522 + }, + { + "epoch": 8.661623508874017, + "flow_matching_loss": 0.019203115440905093, + "grad_norm": 0.20243795216083527, + "learning_rate": 2.6895517272236704e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006087064603343606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027822246076539157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014956353791058063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8557241678237915, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29770, + "weighted_flow_matching_loss": 0.019203115440905093, + "weighted_rkd_loss": 0.01279850397258997 + }, + { + "epoch": 8.664533022985161, + "flow_matching_loss": 0.019431588239967823, + "grad_norm": 0.2055610716342926, + "learning_rate": 2.6892998682174402e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006606796849519014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003015322203282267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01622891575098038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8556321620941162, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29780, + "weighted_flow_matching_loss": 0.019431588239967823, + "weighted_rkd_loss": 0.013885942753404379 + }, + { + "epoch": 8.667442537096305, + "flow_matching_loss": 0.01854599667713046, + "grad_norm": 0.20350410044193268, + "learning_rate": 2.6890479188917335e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006244193715974689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027696920908056198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015258079580962657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8555401384830474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29790, + "weighted_flow_matching_loss": 0.01854599667713046, + "weighted_rkd_loss": 0.013053886406123639 + }, + { + "epoch": 8.670352051207448, + "flow_matching_loss": 0.0172878110781312, + "grad_norm": 0.21529071033000946, + "learning_rate": 2.6887958792656846e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006387237505987286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002619769237935543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015394244343042374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8554480850696564, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29800, + "weighted_flow_matching_loss": 0.0172878110781312, + "weighted_rkd_loss": 0.01316897850483656 + }, + { + "epoch": 8.673261565318592, + "flow_matching_loss": 0.01927055586129427, + "grad_norm": 0.14543849229812622, + "learning_rate": 2.6885437493584336e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058300458826124665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025701817590743302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014230273384600877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8553560078144073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29810, + "weighted_flow_matching_loss": 0.01927055586129427, + "weighted_rkd_loss": 0.012171963043510915 + }, + { + "epoch": 8.676171079429736, + "flow_matching_loss": 0.018818453513085844, + "grad_norm": 0.20357568562030792, + "learning_rate": 2.688291529189129e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006597969355061651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002957141399383545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016153080202639104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8552639186382294, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29820, + "weighted_flow_matching_loss": 0.018818453513085844, + "weighted_rkd_loss": 0.01381518254056573 + }, + { + "epoch": 8.67908059354088, + "flow_matching_loss": 0.018160285893827677, + "grad_norm": 0.23379436135292053, + "learning_rate": 2.6880392187769253e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006784264370799065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003125942451879382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016694471426308154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8551717877388001, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29830, + "weighted_flow_matching_loss": 0.018160285893827677, + "weighted_rkd_loss": 0.014276603236794472 + }, + { + "epoch": 8.681990107652021, + "flow_matching_loss": 0.018552735447883606, + "grad_norm": 0.19576287269592285, + "learning_rate": 2.687786818140982e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060120952315628525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025771564804017546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014601346850395203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8550796449184418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29840, + "weighted_flow_matching_loss": 0.018552735447883606, + "weighted_rkd_loss": 0.012485332693904639 + }, + { + "epoch": 8.684899621763165, + "flow_matching_loss": 0.017372911982238292, + "grad_norm": 0.22177128493785858, + "learning_rate": 2.6875343273004693e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006203744560480118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002696183416992426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015103672537952662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8549874722957611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29850, + "weighted_flow_matching_loss": 0.017372911982238292, + "weighted_rkd_loss": 0.01291342917829752 + }, + { + "epoch": 8.687809135874309, + "flow_matching_loss": 0.020576360635459423, + "grad_norm": 0.17014050483703613, + "learning_rate": 2.6872817462745616e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006265300465747714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002776866254862398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015307467430830002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8548952758312225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29860, + "weighted_flow_matching_loss": 0.020576360635459423, + "weighted_rkd_loss": 0.013086287770420313 + }, + { + "epoch": 8.690718649985453, + "flow_matching_loss": 0.017318927962332965, + "grad_norm": 0.18939407169818878, + "learning_rate": 2.687029075082441e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005722784507088363, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002439180575311184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013884749449789523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.854803055524826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29870, + "weighted_flow_matching_loss": 0.017318927962332965, + "weighted_rkd_loss": 0.011868730559945107 + }, + { + "epoch": 8.693628164096596, + "flow_matching_loss": 0.019158017728477715, + "grad_norm": 0.2163781374692917, + "learning_rate": 2.686776313743295e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065205380320549015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027587309014052153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015799807105213403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8547108113765717, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29880, + "weighted_flow_matching_loss": 0.019158017728477715, + "weighted_rkd_loss": 0.013504283782094718 + }, + { + "epoch": 8.69653767820774, + "flow_matching_loss": 0.020013303495943546, + "grad_norm": 0.16880446672439575, + "learning_rate": 2.6865234622763205e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00616104134824127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026465153321623804, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01496859798207879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8546185433864594, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29890, + "weighted_flow_matching_loss": 0.020013303495943546, + "weighted_rkd_loss": 0.012792515009641648 + }, + { + "epoch": 8.699447192318882, + "flow_matching_loss": 0.01939530223608017, + "grad_norm": 0.18056049942970276, + "learning_rate": 2.686270520700719e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062166930641978976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028749234043061733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015308309532701968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8545262455940247, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29900, + "weighted_flow_matching_loss": 0.01939530223608017, + "weighted_rkd_loss": 0.01308135949075222 + }, + { + "epoch": 8.702356706430026, + "flow_matching_loss": 0.01874580020084977, + "grad_norm": 0.19293314218521118, + "learning_rate": 2.6860174890357e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005986184719949961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002611859026364982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014584228303283453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8544339299201965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29910, + "weighted_flow_matching_loss": 0.01874580020084977, + "weighted_rkd_loss": 0.012461265176534652 + }, + { + "epoch": 8.70526622054117, + "flow_matching_loss": 0.019018324557691813, + "grad_norm": 0.1457298845052719, + "learning_rate": 2.6857643673004807e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061998840421438215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002712165133561939, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015111932996660471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.854341596364975, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29920, + "weighted_flow_matching_loss": 0.019018324557691813, + "weighted_rkd_loss": 0.01291079567745328 + }, + { + "epoch": 8.708175734652313, + "flow_matching_loss": 0.018902915250509977, + "grad_norm": 0.1852768212556839, + "learning_rate": 2.685511155514283e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062016772106289865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002639595710206777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015042950306087733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8542492091655731, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29930, + "weighted_flow_matching_loss": 0.018902915250509977, + "weighted_rkd_loss": 0.012850444298237562 + }, + { + "epoch": 8.711085248763457, + "flow_matching_loss": 0.018898219987750053, + "grad_norm": 0.1743631809949875, + "learning_rate": 2.6852578536963358e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005881409579887986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00251196309691295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014274782221764325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8541568219661713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29940, + "weighted_flow_matching_loss": 0.018898219987750053, + "weighted_rkd_loss": 0.012192927487194537 + }, + { + "epoch": 8.7139947628746, + "flow_matching_loss": 0.0187858403660357, + "grad_norm": 0.17126412689685822, + "learning_rate": 2.6850044618658773e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005833990639075637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026240175240673123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014291999023407697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8540644228458405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29950, + "weighted_flow_matching_loss": 0.0187858403660357, + "weighted_rkd_loss": 0.012206288613379001 + }, + { + "epoch": 8.716904276985744, + "flow_matching_loss": 0.01906862948089838, + "grad_norm": 0.18657894432544708, + "learning_rate": 2.6847509800421498e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005998524837195873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025516588939353824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014548708777874708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8539719760417939, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29960, + "weighted_flow_matching_loss": 0.01906862948089838, + "weighted_rkd_loss": 0.012424176000058651 + }, + { + "epoch": 8.719813791096886, + "flow_matching_loss": 0.020413370244204997, + "grad_norm": 0.1789555847644806, + "learning_rate": 2.6844974082444048e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006215497292578221, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002698351989965886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015129346586763858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8538795173168182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29970, + "weighted_flow_matching_loss": 0.020413370244204997, + "weighted_rkd_loss": 0.012918658461421728 + }, + { + "epoch": 8.72272330520803, + "flow_matching_loss": 0.017829259857535362, + "grad_norm": 0.18651767075061798, + "learning_rate": 2.684243746491898e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005880370270460844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002686546219047159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014447286818176508, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8537870287895203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29980, + "weighted_flow_matching_loss": 0.017829259857535362, + "weighted_rkd_loss": 0.012334879860281945 + }, + { + "epoch": 8.725632819319173, + "flow_matching_loss": 0.018221269361674786, + "grad_norm": 0.16111408174037933, + "learning_rate": 2.683989994803894e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059560134541243315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002511506772134453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014423533901572227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8536945164203644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29990, + "weighted_flow_matching_loss": 0.018221269361674786, + "weighted_rkd_loss": 0.012313286121934653 + }, + { + "epoch": 8.728542333430317, + "flow_matching_loss": 0.017741544637829065, + "grad_norm": 0.18192058801651, + "learning_rate": 2.6837361531996635e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005797489592805505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026041280245408414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014199107512831688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8536019742488861, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 30000, + "weighted_flow_matching_loss": 0.017741544637829065, + "weighted_rkd_loss": 0.012120378855615854 + } + ], + "logging_steps": 10, + "max_steps": 120000, + "num_input_tokens_seen": 0, + "num_train_epochs": 35, + "save_steps": 30000, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 128, + "trial_name": null, + "trial_params": null +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..c7f88b4546d48d8c41997f93d82151190116619d --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5000065450392368, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/experiment_cfg/metadata.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..300eb070963337016e16465f50ba8351ea2eb25c --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.3139495849609375, + 0.4876587688922882, + 0.7196521759033203 + ], + "min": [ + -4.94293737411499, + -6.890198230743408, + 0.6996827125549316 + ], + "mean": [ + 2.749288365724937, + -2.0455216239909983, + 0.700532551020533 + ], + "std": [ + 1.6786295897046262, + 1.312897969434244, + 0.00133751758906258 + ], + "q01": [ + -1.4241907881148037, + -5.1716547935950885, + 0.6999670898768614 + ], + "q99": [ + 6.198111545831555, + -0.5873531334104489, + 0.7038833014242587 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.23084999587450247, + 0.6238431088581847 + ], + "std": [ + 0.0, + 0.0, + 0.6556404371644047, + 0.3572252105732042 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 1.7482354593273349e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.9014528393745422, + 0.8003877401351929, + 0.9829942584037781 + ], + "min": [ + -0.37471598386764526, + -0.8472502827644348, + -0.25070279836654663 + ], + "mean": [ + 0.28667332601863704, + -0.038315473170876704, + 0.45974658906777444 + ], + "std": [ + 0.17067058828305154, + 0.23569952037784195, + 0.21950702354181487 + ], + "q01": [ + 0.015172948963784228, + -0.44450502063220615, + 0.23314444103329338 + ], + "q99": [ + 0.5609069689572412, + 0.38454179128009547, + 0.7028102736473852 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999998807907104, + 0.9984455108642578, + 0.9434727430343628, + 0.9062229990959167 + ], + "min": [ + -0.9999930262565613, + -0.9988337755203247, + -0.9618149995803833, + 2.1872274658107926e-07 + ], + "mean": [ + -0.25672297401336, + 0.024425335806687216, + -0.08740977164483847, + 0.16608739644018397 + ], + "std": [ + 0.7932585208392383, + 0.3102260195580416, + 0.3772472576315148, + 0.17451959300158773 + ], + "q01": [ + -0.99379006987882, + -0.5478102050038828, + -0.6101777952971636, + 0.002274085014134748 + ], + "q99": [ + 0.8804856038896288, + 0.5910611092873037, + 0.5216058042993562, + 0.5231965974012255 + ] + }, + "gripper_qpos": { + "max": [ + 0.055869169533252716, + 0.010916369967162609 + ], + "min": [ + -0.011436971835792065, + -0.05664053186774254 + ], + "mean": [ + 0.031651551516052555, + -0.03161481387920421 + ], + "std": [ + 0.013119769222217526, + 0.01306078225192402 + ], + "q01": [ + 0.0060999246446777336, + -0.04062601653964198 + ], + "q99": [ + 0.04054891621517283, + -0.0059163567113456345 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.008144896753718373, + -0.00018893135146078263, + -0.0008739062727221845, + 0.0 + ], + "std": [ + 0.11030045424364411, + 0.10148082570313594, + 0.08975698373467506, + 0.0 + ], + "q01": [ + -0.07287373067581518, + -0.09446994199569948, + -0.07702818259249572, + 0.0 + ], + "q99": [ + 0.04485485414407898, + 0.09626941605965177, + 0.07610665906122742, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9221974313425939 + ], + "std": [ + 0.38670194342612674 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.384693946883099 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.01976129923017491, + -0.020870631344490034, + -0.05993476517349181 + ], + "std": [ + 0.4347024990804053, + 0.4221662976569997, + 0.3845506843044066 + ], + "q01": [ + -0.8835792939767807, + -0.9280919251713778, + -0.783361071470956 + ], + "q99": [ + 0.7622084438449307, + 0.8625125252609077, + 0.7470974753250706 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008089673068544335, + -0.026498254627927348, + 0.0029083551743059005 + ], + "std": [ + 0.11216734503733773, + 0.13206002613798629, + 0.12615870412692864 + ], + "q01": [ + -0.2814627972846672, + -0.4342386516115439, + -0.34489755628340574 + ], + "q99": [ + 0.31879353266939386, + 0.28411797683220563, + 0.3597718671669894 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3824228874769045 + ], + "std": [ + 0.9240015096469786 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.5597272305813592 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-60000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..2d6b2fbe869c27961185a131cdaad84c01fbc3b2 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.14644892343511284, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/experiment_cfg/metadata.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..300eb070963337016e16465f50ba8351ea2eb25c --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.3139495849609375, + 0.4876587688922882, + 0.7196521759033203 + ], + "min": [ + -4.94293737411499, + -6.890198230743408, + 0.6996827125549316 + ], + "mean": [ + 2.749288365724937, + -2.0455216239909983, + 0.700532551020533 + ], + "std": [ + 1.6786295897046262, + 1.312897969434244, + 0.00133751758906258 + ], + "q01": [ + -1.4241907881148037, + -5.1716547935950885, + 0.6999670898768614 + ], + "q99": [ + 6.198111545831555, + -0.5873531334104489, + 0.7038833014242587 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.23084999587450247, + 0.6238431088581847 + ], + "std": [ + 0.0, + 0.0, + 0.6556404371644047, + 0.3572252105732042 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 1.7482354593273349e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.9014528393745422, + 0.8003877401351929, + 0.9829942584037781 + ], + "min": [ + -0.37471598386764526, + -0.8472502827644348, + -0.25070279836654663 + ], + "mean": [ + 0.28667332601863704, + -0.038315473170876704, + 0.45974658906777444 + ], + "std": [ + 0.17067058828305154, + 0.23569952037784195, + 0.21950702354181487 + ], + "q01": [ + 0.015172948963784228, + -0.44450502063220615, + 0.23314444103329338 + ], + "q99": [ + 0.5609069689572412, + 0.38454179128009547, + 0.7028102736473852 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999998807907104, + 0.9984455108642578, + 0.9434727430343628, + 0.9062229990959167 + ], + "min": [ + -0.9999930262565613, + -0.9988337755203247, + -0.9618149995803833, + 2.1872274658107926e-07 + ], + "mean": [ + -0.25672297401336, + 0.024425335806687216, + -0.08740977164483847, + 0.16608739644018397 + ], + "std": [ + 0.7932585208392383, + 0.3102260195580416, + 0.3772472576315148, + 0.17451959300158773 + ], + "q01": [ + -0.99379006987882, + -0.5478102050038828, + -0.6101777952971636, + 0.002274085014134748 + ], + "q99": [ + 0.8804856038896288, + 0.5910611092873037, + 0.5216058042993562, + 0.5231965974012255 + ] + }, + "gripper_qpos": { + "max": [ + 0.055869169533252716, + 0.010916369967162609 + ], + "min": [ + -0.011436971835792065, + -0.05664053186774254 + ], + "mean": [ + 0.031651551516052555, + -0.03161481387920421 + ], + "std": [ + 0.013119769222217526, + 0.01306078225192402 + ], + "q01": [ + 0.0060999246446777336, + -0.04062601653964198 + ], + "q99": [ + 0.04054891621517283, + -0.0059163567113456345 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.008144896753718373, + -0.00018893135146078263, + -0.0008739062727221845, + 0.0 + ], + "std": [ + 0.11030045424364411, + 0.10148082570313594, + 0.08975698373467506, + 0.0 + ], + "q01": [ + -0.07287373067581518, + -0.09446994199569948, + -0.07702818259249572, + 0.0 + ], + "q99": [ + 0.04485485414407898, + 0.09626941605965177, + 0.07610665906122742, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9221974313425939 + ], + "std": [ + 0.38670194342612674 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.384693946883099 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.01976129923017491, + -0.020870631344490034, + -0.05993476517349181 + ], + "std": [ + 0.4347024990804053, + 0.4221662976569997, + 0.3845506843044066 + ], + "q01": [ + -0.8835792939767807, + -0.9280919251713778, + -0.783361071470956 + ], + "q99": [ + 0.7622084438449307, + 0.8625125252609077, + 0.7470974753250706 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008089673068544335, + -0.026498254627927348, + 0.0029083551743059005 + ], + "std": [ + 0.11216734503733773, + 0.13206002613798629, + 0.12615870412692864 + ], + "q01": [ + -0.2814627972846672, + -0.4342386516115439, + -0.34489755628340574 + ], + "q99": [ + 0.31879353266939386, + 0.28411797683220563, + 0.3597718671669894 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3824228874769045 + ], + "std": [ + 0.9240015096469786 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.5597272305813592 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/checkpoint-90000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/config.json new file mode 100644 index 0000000000000000000000000000000000000000..064785cc03d1e2223bf2981a1b22674a7e445646 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.0, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/experiment_cfg/metadata.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..300eb070963337016e16465f50ba8351ea2eb25c --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.3139495849609375, + 0.4876587688922882, + 0.7196521759033203 + ], + "min": [ + -4.94293737411499, + -6.890198230743408, + 0.6996827125549316 + ], + "mean": [ + 2.749288365724937, + -2.0455216239909983, + 0.700532551020533 + ], + "std": [ + 1.6786295897046262, + 1.312897969434244, + 0.00133751758906258 + ], + "q01": [ + -1.4241907881148037, + -5.1716547935950885, + 0.6999670898768614 + ], + "q99": [ + 6.198111545831555, + -0.5873531334104489, + 0.7038833014242587 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.23084999587450247, + 0.6238431088581847 + ], + "std": [ + 0.0, + 0.0, + 0.6556404371644047, + 0.3572252105732042 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 1.7482354593273349e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.9014528393745422, + 0.8003877401351929, + 0.9829942584037781 + ], + "min": [ + -0.37471598386764526, + -0.8472502827644348, + -0.25070279836654663 + ], + "mean": [ + 0.28667332601863704, + -0.038315473170876704, + 0.45974658906777444 + ], + "std": [ + 0.17067058828305154, + 0.23569952037784195, + 0.21950702354181487 + ], + "q01": [ + 0.015172948963784228, + -0.44450502063220615, + 0.23314444103329338 + ], + "q99": [ + 0.5609069689572412, + 0.38454179128009547, + 0.7028102736473852 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999998807907104, + 0.9984455108642578, + 0.9434727430343628, + 0.9062229990959167 + ], + "min": [ + -0.9999930262565613, + -0.9988337755203247, + -0.9618149995803833, + 2.1872274658107926e-07 + ], + "mean": [ + -0.25672297401336, + 0.024425335806687216, + -0.08740977164483847, + 0.16608739644018397 + ], + "std": [ + 0.7932585208392383, + 0.3102260195580416, + 0.3772472576315148, + 0.17451959300158773 + ], + "q01": [ + -0.99379006987882, + -0.5478102050038828, + -0.6101777952971636, + 0.002274085014134748 + ], + "q99": [ + 0.8804856038896288, + 0.5910611092873037, + 0.5216058042993562, + 0.5231965974012255 + ] + }, + "gripper_qpos": { + "max": [ + 0.055869169533252716, + 0.010916369967162609 + ], + "min": [ + -0.011436971835792065, + -0.05664053186774254 + ], + "mean": [ + 0.031651551516052555, + -0.03161481387920421 + ], + "std": [ + 0.013119769222217526, + 0.01306078225192402 + ], + "q01": [ + 0.0060999246446777336, + -0.04062601653964198 + ], + "q99": [ + 0.04054891621517283, + -0.0059163567113456345 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.008144896753718373, + -0.00018893135146078263, + -0.0008739062727221845, + 0.0 + ], + "std": [ + 0.11030045424364411, + 0.10148082570313594, + 0.08975698373467506, + 0.0 + ], + "q01": [ + -0.07287373067581518, + -0.09446994199569948, + -0.07702818259249572, + 0.0 + ], + "q99": [ + 0.04485485414407898, + 0.09626941605965177, + 0.07610665906122742, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9221974313425939 + ], + "std": [ + 0.38670194342612674 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.384693946883099 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.01976129923017491, + -0.020870631344490034, + -0.05993476517349181 + ], + "std": [ + 0.4347024990804053, + 0.4221662976569997, + 0.3845506843044066 + ], + "q01": [ + -0.8835792939767807, + -0.9280919251713778, + -0.783361071470956 + ], + "q99": [ + 0.7622084438449307, + 0.8625125252609077, + 0.7470974753250706 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008089673068544335, + -0.026498254627927348, + 0.0029083551743059005 + ], + "std": [ + 0.11216734503733773, + 0.13206002613798629, + 0.12615870412692864 + ], + "q01": [ + -0.2814627972846672, + -0.4342386516115439, + -0.34489755628340574 + ], + "q99": [ + 0.31879353266939386, + 0.28411797683220563, + 0.3597718671669894 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3824228874769045 + ], + "std": [ + 0.9240015096469786 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.5597272305813592 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/resolved_config.yaml b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/resolved_config.yaml new file mode 100644 index 0000000000000000000000000000000000000000..20a52b8401487722ecc796c33336d7d64eabbbda --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/resolved_config.yaml @@ -0,0 +1,41 @@ +experiment_name: RKDv3 TimewarpVAE DA, Student:APHead, single-phase FM+RKD cosine + decay from GR00T base to 120k +policy_type: groot_RKDv3 +base_model_path: nvidia/GR00T-N1.5-3B +output_root: ${HOME}/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k +dataset: + dataset_soup: my_atomic26_human +training: + num_gpus: 1 + batch_size: 128 + seed: 42 + report_to: wandb +model: + rkd_teacher_source: timewarp_action + rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46 + rkd_relation_mode: token_pair_mean + rkd_loss_type: distance_angle + rkd_student_reconstruction_head_pooling: attention + rkd_student_reconstruction_head_enabled: true + rkd_student_reconstruction_head_dim: 512 + rkd_student_reconstruction_head_hidden_dim: 512 + rkd_distance_loss_weight: 1.0 + rkd_angle_loss_weight: 2.0 + rkd_exclude_diagonal: true +phases: +- name: fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k + max_steps: 120000 + save_steps: 30000 + trainable: + tune_llm: false + tune_visual: false + tune_projector: true + tune_diffusion_model: true + losses: + rkd_enabled: true + rkd_fm_loss_weight: 1.0 + rkd_loss_weight: 1.0 + rkd_loss_weight_schedule: cosine_decay + rkd_loss_weight_start: 1.0 + rkd_loss_weight_end: 0.0 +resolved_sweep: {} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k.yaml b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k.yaml index 4184208254de419ae9435b6c180c49b4487d99eb..43b98993bd5b6030471873c7d8d18d4728804fda 100644 --- a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k.yaml +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k.yaml @@ -9,6 +9,7 @@ training: num_gpus: 1 batch_size: 128 seed: 42 + report_to: wandb model: rkd_teacher_source: timewarp_action rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46 diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..9a13e61b733d3ca3770b1b9a4290f6dfedb745ec --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.5000130901875584, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/experiment_cfg/metadata.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..300eb070963337016e16465f50ba8351ea2eb25c --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.3139495849609375, + 0.4876587688922882, + 0.7196521759033203 + ], + "min": [ + -4.94293737411499, + -6.890198230743408, + 0.6996827125549316 + ], + "mean": [ + 2.749288365724937, + -2.0455216239909983, + 0.700532551020533 + ], + "std": [ + 1.6786295897046262, + 1.312897969434244, + 0.00133751758906258 + ], + "q01": [ + -1.4241907881148037, + -5.1716547935950885, + 0.6999670898768614 + ], + "q99": [ + 6.198111545831555, + -0.5873531334104489, + 0.7038833014242587 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.23084999587450247, + 0.6238431088581847 + ], + "std": [ + 0.0, + 0.0, + 0.6556404371644047, + 0.3572252105732042 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 1.7482354593273349e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.9014528393745422, + 0.8003877401351929, + 0.9829942584037781 + ], + "min": [ + -0.37471598386764526, + -0.8472502827644348, + -0.25070279836654663 + ], + "mean": [ + 0.28667332601863704, + -0.038315473170876704, + 0.45974658906777444 + ], + "std": [ + 0.17067058828305154, + 0.23569952037784195, + 0.21950702354181487 + ], + "q01": [ + 0.015172948963784228, + -0.44450502063220615, + 0.23314444103329338 + ], + "q99": [ + 0.5609069689572412, + 0.38454179128009547, + 0.7028102736473852 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999998807907104, + 0.9984455108642578, + 0.9434727430343628, + 0.9062229990959167 + ], + "min": [ + -0.9999930262565613, + -0.9988337755203247, + -0.9618149995803833, + 2.1872274658107926e-07 + ], + "mean": [ + -0.25672297401336, + 0.024425335806687216, + -0.08740977164483847, + 0.16608739644018397 + ], + "std": [ + 0.7932585208392383, + 0.3102260195580416, + 0.3772472576315148, + 0.17451959300158773 + ], + "q01": [ + -0.99379006987882, + -0.5478102050038828, + -0.6101777952971636, + 0.002274085014134748 + ], + "q99": [ + 0.8804856038896288, + 0.5910611092873037, + 0.5216058042993562, + 0.5231965974012255 + ] + }, + "gripper_qpos": { + "max": [ + 0.055869169533252716, + 0.010916369967162609 + ], + "min": [ + -0.011436971835792065, + -0.05664053186774254 + ], + "mean": [ + 0.031651551516052555, + -0.03161481387920421 + ], + "std": [ + 0.013119769222217526, + 0.01306078225192402 + ], + "q01": [ + 0.0060999246446777336, + -0.04062601653964198 + ], + "q99": [ + 0.04054891621517283, + -0.0059163567113456345 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.008144896753718373, + -0.00018893135146078263, + -0.0008739062727221845, + 0.0 + ], + "std": [ + 0.11030045424364411, + 0.10148082570313594, + 0.08975698373467506, + 0.0 + ], + "q01": [ + -0.07287373067581518, + -0.09446994199569948, + -0.07702818259249572, + 0.0 + ], + "q99": [ + 0.04485485414407898, + 0.09626941605965177, + 0.07610665906122742, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9221974313425939 + ], + "std": [ + 0.38670194342612674 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.384693946883099 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.01976129923017491, + -0.020870631344490034, + -0.05993476517349181 + ], + "std": [ + 0.4347024990804053, + 0.4221662976569997, + 0.3845506843044066 + ], + "q01": [ + -0.8835792939767807, + -0.9280919251713778, + -0.783361071470956 + ], + "q99": [ + 0.7622084438449307, + 0.8625125252609077, + 0.7470974753250706 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008089673068544335, + -0.026498254627927348, + 0.0029083551743059005 + ], + "std": [ + 0.11216734503733773, + 0.13206002613798629, + 0.12615870412692864 + ], + "q01": [ + -0.2814627972846672, + -0.4342386516115439, + -0.34489755628340574 + ], + "q99": [ + 0.31879353266939386, + 0.28411797683220563, + 0.3597718671669894 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3824228874769045 + ], + "std": [ + 0.9240015096469786 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.5597272305813592 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/trainer_state.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/trainer_state.json new file mode 100644 index 0000000000000000000000000000000000000000..1f6911ef222d0ae2dcd25e406b888a1e755e2d35 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-30000/trainer_state.json @@ -0,0 +1,138034 @@ +{ + "best_global_step": null, + "best_metric": null, + "best_model_checkpoint": null, + "epoch": 8.728542333430317, + "eval_steps": 500, + "global_step": 30000, + "is_hyper_param_search": false, + "is_local_process_zero": true, + "is_world_process_zero": true, + "log_history": [ + { + "epoch": 0.002909514111143439, + "flow_matching_loss": 0.5736978411674499, + "grad_norm": 6.92530632019043, + "learning_rate": 9e-08, + "loss": 0.9088, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11319504007697105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10873725935816765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33512733578681947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999999821186065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10, + "weighted_flow_matching_loss": 0.5736978411674499, + "weighted_rkd_loss": 0.33512732684612273 + }, + { + "epoch": 0.005819028222286878, + "flow_matching_loss": 0.5558205008506775, + "grad_norm": 7.404500961303711, + "learning_rate": 1.9e-07, + "loss": 0.902, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11475425884127617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11665472611784936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3461632400751114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999998509883881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20, + "weighted_flow_matching_loss": 0.5558205008506775, + "weighted_rkd_loss": 0.3461631894111633 + }, + { + "epoch": 0.008728542333430317, + "flow_matching_loss": 0.552828025817871, + "grad_norm": 5.956160068511963, + "learning_rate": 2.9000000000000003e-07, + "loss": 0.9099, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1179118774831295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.12127644792199135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.35710020661354064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999995768070221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 30, + "weighted_flow_matching_loss": 0.552828025817871, + "weighted_rkd_loss": 0.35710005462169647 + }, + { + "epoch": 0.011638056444573755, + "flow_matching_loss": 0.5737781345844268, + "grad_norm": 6.132974147796631, + "learning_rate": 3.8999999999999997e-07, + "loss": 0.919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11557439342141151, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11409032940864564, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.34523911476135255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999999177455902, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 40, + "weighted_flow_matching_loss": 0.5737781345844268, + "weighted_rkd_loss": 0.34523882865905764 + }, + { + "epoch": 0.014547570555717195, + "flow_matching_loss": 0.5201578915119172, + "grad_norm": 6.637472152709961, + "learning_rate": 4.9e-07, + "loss": 0.8624, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11497188732028008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.11233537048101425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3422791510820389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999986529350281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 50, + "weighted_flow_matching_loss": 0.5201578915119172, + "weighted_rkd_loss": 0.34227868914604187 + }, + { + "epoch": 0.017457084666860634, + "flow_matching_loss": 0.5009894847869873, + "grad_norm": 5.5547099113464355, + "learning_rate": 5.9e-07, + "loss": 0.8297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11133255362510681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10600293278694153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.32866804003715516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999979555606842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 60, + "weighted_flow_matching_loss": 0.5009894847869873, + "weighted_rkd_loss": 0.32866736352443693 + }, + { + "epoch": 0.020366598778004074, + "flow_matching_loss": 0.44645200073719027, + "grad_norm": 5.240603446960449, + "learning_rate": 6.9e-07, + "loss": 0.7822, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.11359371468424798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10855495631694793, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33574238717556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999971389770508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 70, + "weighted_flow_matching_loss": 0.44645200073719027, + "weighted_rkd_loss": 0.33574143052101135 + }, + { + "epoch": 0.02327611288914751, + "flow_matching_loss": 0.41712434887886046, + "grad_norm": 4.440354347229004, + "learning_rate": 7.900000000000001e-07, + "loss": 0.7492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1123950146138668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10726893544197083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.33205896317958833, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999961912631988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 80, + "weighted_flow_matching_loss": 0.41712434887886046, + "weighted_rkd_loss": 0.33205769658088685 + }, + { + "epoch": 0.02618562700029095, + "flow_matching_loss": 0.3057377427816391, + "grad_norm": 3.3242709636688232, + "learning_rate": 8.900000000000001e-07, + "loss": 0.6327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10991297960281372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.10712310522794724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3269490659236908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999995106458664, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 90, + "weighted_flow_matching_loss": 0.3057377427816391, + "weighted_rkd_loss": 0.32694747149944303 + }, + { + "epoch": 0.02909514111143439, + "flow_matching_loss": 0.26383223086595536, + "grad_norm": 3.384511947631836, + "learning_rate": 9.9e-07, + "loss": 0.5856, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10963985919952393, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.1025143176317215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3217940330505371, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999938786029816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 100, + "weighted_flow_matching_loss": 0.26383223086595536, + "weighted_rkd_loss": 0.32179206013679507 + }, + { + "epoch": 0.03200465522257783, + "flow_matching_loss": 0.23409549593925477, + "grad_norm": 2.756331443786621, + "learning_rate": 1.0900000000000002e-06, + "loss": 0.5478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10780841335654259, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09805921390652657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3136760383844376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999925076961518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 110, + "weighted_flow_matching_loss": 0.23409549593925477, + "weighted_rkd_loss": 0.31367368400096896 + }, + { + "epoch": 0.03491416933372127, + "flow_matching_loss": 0.2177390307188034, + "grad_norm": 2.2592453956604004, + "learning_rate": 1.19e-06, + "loss": 0.5192, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.1049491100013256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09153911247849464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.3014373332262039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999910116195678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 120, + "weighted_flow_matching_loss": 0.2177390307188034, + "weighted_rkd_loss": 0.30143462121486664 + }, + { + "epoch": 0.037823683444864704, + "flow_matching_loss": 0.18370274752378463, + "grad_norm": 2.6151328086853027, + "learning_rate": 1.29e-06, + "loss": 0.4814, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10359546542167664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.09049091562628746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2976818472146988, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999893724918365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 130, + "weighted_flow_matching_loss": 0.18370274752378463, + "weighted_rkd_loss": 0.2976786881685257 + }, + { + "epoch": 0.04073319755600815, + "flow_matching_loss": 0.18502854704856872, + "grad_norm": 1.996734380722046, + "learning_rate": 1.39e-06, + "loss": 0.4769, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.10282047092914581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08620013520121575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2918410778045654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999876022338867, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 140, + "weighted_flow_matching_loss": 0.18502854704856872, + "weighted_rkd_loss": 0.2918374627828598 + }, + { + "epoch": 0.043642711667151585, + "flow_matching_loss": 0.16940540224313735, + "grad_norm": 1.5936223268508911, + "learning_rate": 1.49e-06, + "loss": 0.4509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09917353466153145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08313381299376488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2814808815717697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999856889247895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 150, + "weighted_flow_matching_loss": 0.16940540224313735, + "weighted_rkd_loss": 0.2814768522977829 + }, + { + "epoch": 0.04655222577829502, + "flow_matching_loss": 0.15659096240997314, + "grad_norm": 1.8602830171585083, + "learning_rate": 1.59e-06, + "loss": 0.4364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09940331727266312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.08103863447904587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.279845267534256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999836325645447, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 160, + "weighted_flow_matching_loss": 0.15659096240997314, + "weighted_rkd_loss": 0.279840686917305 + }, + { + "epoch": 0.049461739889438465, + "flow_matching_loss": 0.16221066564321518, + "grad_norm": 1.4860185384750366, + "learning_rate": 1.69e-06, + "loss": 0.4352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09731964021921158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0784039281308651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.27304320633411405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999814450740814, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 170, + "weighted_flow_matching_loss": 0.16221066564321518, + "weighted_rkd_loss": 0.2730381369590759 + }, + { + "epoch": 0.0523712540005819, + "flow_matching_loss": 0.15390500724315642, + "grad_norm": 1.128212571144104, + "learning_rate": 1.79e-06, + "loss": 0.4275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09779494479298592, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07804471179842949, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2736346036195755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999791383743286, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 180, + "weighted_flow_matching_loss": 0.15390500724315642, + "weighted_rkd_loss": 0.27362889647483823 + }, + { + "epoch": 0.055280768111725345, + "flow_matching_loss": 0.15091891437768937, + "grad_norm": 1.9683724641799927, + "learning_rate": 1.8900000000000001e-06, + "loss": 0.4223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09686053097248078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0776382178068161, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2713592782616615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999976658821106, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 190, + "weighted_flow_matching_loss": 0.15091891437768937, + "weighted_rkd_loss": 0.27135293930768967 + }, + { + "epoch": 0.05819028222286878, + "flow_matching_loss": 0.13371271193027495, + "grad_norm": 1.3183945417404175, + "learning_rate": 1.99e-06, + "loss": 0.4026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09664153903722764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07557290941476821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.26885598301887514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999740660190582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 200, + "weighted_flow_matching_loss": 0.13371271193027495, + "weighted_rkd_loss": 0.2688490092754364 + }, + { + "epoch": 0.06109979633401222, + "flow_matching_loss": 0.13600234985351561, + "grad_norm": 1.9926862716674805, + "learning_rate": 2.09e-06, + "loss": 0.4021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0960406668484211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07407169491052627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2661530256271362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999713242053986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 210, + "weighted_flow_matching_loss": 0.13600234985351561, + "weighted_rkd_loss": 0.26614539325237274 + }, + { + "epoch": 0.06400931044515566, + "flow_matching_loss": 0.1353871203958988, + "grad_norm": 1.3094382286071777, + "learning_rate": 2.1899999999999998e-06, + "loss": 0.3968, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09495279490947724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.07155792638659478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.26146351397037504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999684512615203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 220, + "weighted_flow_matching_loss": 0.1353871203958988, + "weighted_rkd_loss": 0.2614552691578865 + }, + { + "epoch": 0.06691882455629909, + "flow_matching_loss": 0.133908162266016, + "grad_norm": 0.8854209184646606, + "learning_rate": 2.29e-06, + "loss": 0.3866, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09234696701169014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06797327287495136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.25266720950603483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999654412269592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 230, + "weighted_flow_matching_loss": 0.133908162266016, + "weighted_rkd_loss": 0.2526584818959236 + }, + { + "epoch": 0.06982833866744254, + "flow_matching_loss": 0.13140231147408485, + "grad_norm": 1.3118855953216553, + "learning_rate": 2.39e-06, + "loss": 0.3869, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0928355760872364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06986774131655693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2555388867855072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999962306022644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 240, + "weighted_flow_matching_loss": 0.13140231147408485, + "weighted_rkd_loss": 0.2555292546749115 + }, + { + "epoch": 0.07273785277858598, + "flow_matching_loss": 0.13012973442673684, + "grad_norm": 1.0131131410598755, + "learning_rate": 2.4900000000000003e-06, + "loss": 0.3804, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09150774553418159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06723712868988514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.25025261640548707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999590218067169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 250, + "weighted_flow_matching_loss": 0.13012973442673684, + "weighted_rkd_loss": 0.2502423614263535 + }, + { + "epoch": 0.07564736688972941, + "flow_matching_loss": 0.12595354840159417, + "grad_norm": 1.1396840810775757, + "learning_rate": 2.59e-06, + "loss": 0.3827, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09341457933187484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0699301790446043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.25675933957099917, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999556005001068, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 260, + "weighted_flow_matching_loss": 0.12595354840159417, + "weighted_rkd_loss": 0.25674794763326647 + }, + { + "epoch": 0.07855688100087285, + "flow_matching_loss": 0.12678516581654548, + "grad_norm": 1.1156725883483887, + "learning_rate": 2.69e-06, + "loss": 0.3803, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09278700053691864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06799876764416694, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2535727694630623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999520480632782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 270, + "weighted_flow_matching_loss": 0.12678516581654548, + "weighted_rkd_loss": 0.25356061160564425 + }, + { + "epoch": 0.0814663951120163, + "flow_matching_loss": 0.12456624582409859, + "grad_norm": 0.8684354424476624, + "learning_rate": 2.79e-06, + "loss": 0.3735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0908495008945465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0672845683991909, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24898357391357423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999483466148377, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 280, + "weighted_flow_matching_loss": 0.12456624582409859, + "weighted_rkd_loss": 0.24897071868181228 + }, + { + "epoch": 0.08437590922315973, + "flow_matching_loss": 0.1210975967347622, + "grad_norm": 1.2391998767852783, + "learning_rate": 2.8900000000000003e-06, + "loss": 0.3719, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09170695170760154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06744340993463993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2508573144674301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999944519996643, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 290, + "weighted_flow_matching_loss": 0.1210975967347622, + "weighted_rkd_loss": 0.2508433908224106 + }, + { + "epoch": 0.08728542333430317, + "flow_matching_loss": 0.1151894636452198, + "grad_norm": 1.2003968954086304, + "learning_rate": 2.99e-06, + "loss": 0.3596, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08976181522011757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06487972475588322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24440335482358932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999940550327301, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 300, + "weighted_flow_matching_loss": 0.1151894636452198, + "weighted_rkd_loss": 0.2443888232111931 + }, + { + "epoch": 0.09019493744544661, + "flow_matching_loss": 0.10914413556456566, + "grad_norm": 1.0982060432434082, + "learning_rate": 3.09e-06, + "loss": 0.3579, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09130771681666375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06616605557501316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24878149181604386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999936431646347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 310, + "weighted_flow_matching_loss": 0.10914413556456566, + "weighted_rkd_loss": 0.2487656742334366 + }, + { + "epoch": 0.09310445155659004, + "flow_matching_loss": 0.10848486796021461, + "grad_norm": 1.1352776288986206, + "learning_rate": 3.19e-06, + "loss": 0.3547, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08994232267141342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06637923046946526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24626387506723404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999932199716568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 320, + "weighted_flow_matching_loss": 0.10848486796021461, + "weighted_rkd_loss": 0.24624718278646468 + }, + { + "epoch": 0.09601396566773349, + "flow_matching_loss": 0.11101428270339966, + "grad_norm": 1.0791059732437134, + "learning_rate": 3.29e-06, + "loss": 0.3546, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08947807550430298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06469087973237038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2436470329761505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999278128147125, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 330, + "weighted_flow_matching_loss": 0.11101428270339966, + "weighted_rkd_loss": 0.24362944513559343 + }, + { + "epoch": 0.09892347977887693, + "flow_matching_loss": 0.10945233255624771, + "grad_norm": 0.9268128871917725, + "learning_rate": 3.39e-06, + "loss": 0.3548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08984813913702965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06565033234655857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24534661024808885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999923312664032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 340, + "weighted_flow_matching_loss": 0.10945233255624771, + "weighted_rkd_loss": 0.24532780051231384 + }, + { + "epoch": 0.10183299389002037, + "flow_matching_loss": 0.11079161763191223, + "grad_norm": 1.244470477104187, + "learning_rate": 3.49e-06, + "loss": 0.3545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.09001897126436234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06369428969919681, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24373223185539244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999186396598816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 350, + "weighted_flow_matching_loss": 0.11079161763191223, + "weighted_rkd_loss": 0.2437124013900757 + }, + { + "epoch": 0.1047425080011638, + "flow_matching_loss": 0.10633470788598061, + "grad_norm": 0.8832881450653076, + "learning_rate": 3.5900000000000004e-06, + "loss": 0.3484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08914105221629143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06376236863434315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24204447269439697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999138534069061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 360, + "weighted_flow_matching_loss": 0.10633470788598061, + "weighted_rkd_loss": 0.24202362895011903 + }, + { + "epoch": 0.10765202211230725, + "flow_matching_loss": 0.10147337839007378, + "grad_norm": 0.8079697489738464, + "learning_rate": 3.6900000000000002e-06, + "loss": 0.3384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08727494478225709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06240457594394684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23695446252822877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999089300632477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 370, + "weighted_flow_matching_loss": 0.10147337839007378, + "weighted_rkd_loss": 0.2369328811764717 + }, + { + "epoch": 0.11056153622345069, + "flow_matching_loss": 0.10478209406137466, + "grad_norm": 0.7411929368972778, + "learning_rate": 3.7899999999999997e-06, + "loss": 0.346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0887658268213272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06372485123574734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.24125650376081467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9999038696289062, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 380, + "weighted_flow_matching_loss": 0.10478209406137466, + "weighted_rkd_loss": 0.24123331010341645 + }, + { + "epoch": 0.11347105033459412, + "flow_matching_loss": 0.10469116270542145, + "grad_norm": 0.9257120490074158, + "learning_rate": 3.890000000000001e-06, + "loss": 0.3355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08564353585243226, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.059553131088614465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23084020614624023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998986721038818, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 390, + "weighted_flow_matching_loss": 0.10469116270542145, + "weighted_rkd_loss": 0.23081682473421097 + }, + { + "epoch": 0.11638056444573756, + "flow_matching_loss": 0.10447645783424378, + "grad_norm": 0.918569803237915, + "learning_rate": 3.99e-06, + "loss": 0.3362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08605302348732949, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05959480963647366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2317008599638939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99989333152771, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 400, + "weighted_flow_matching_loss": 0.10447645783424378, + "weighted_rkd_loss": 0.23167615085840226 + }, + { + "epoch": 0.11929007855688101, + "flow_matching_loss": 0.10216280817985535, + "grad_norm": 0.9419168829917908, + "learning_rate": 4.09e-06, + "loss": 0.3277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08408195823431015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05740730278193951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22557121962308885, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998878479003906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 410, + "weighted_flow_matching_loss": 0.10216280817985535, + "weighted_rkd_loss": 0.22554592788219452 + }, + { + "epoch": 0.12219959266802444, + "flow_matching_loss": 0.1014733761548996, + "grad_norm": 0.7292332053184509, + "learning_rate": 4.19e-06, + "loss": 0.3325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08593395352363586, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05923335999250412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23110126703977585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998822391033173, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 420, + "weighted_flow_matching_loss": 0.1014733761548996, + "weighted_rkd_loss": 0.23107404559850692 + }, + { + "epoch": 0.12510910677916787, + "flow_matching_loss": 0.10002013072371482, + "grad_norm": 0.9631119966506958, + "learning_rate": 4.29e-06, + "loss": 0.3289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08510671854019165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0587234940379858, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22893693149089814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998764932155609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 430, + "weighted_flow_matching_loss": 0.10002013072371482, + "weighted_rkd_loss": 0.22890865057706833 + }, + { + "epoch": 0.12801862089031132, + "flow_matching_loss": 0.09638321027159691, + "grad_norm": 0.8285052180290222, + "learning_rate": 4.39e-06, + "loss": 0.3328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0877155415713787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.06101129986345768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23644238263368605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998705983161926, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 440, + "weighted_flow_matching_loss": 0.09638321027159691, + "weighted_rkd_loss": 0.2364117830991745 + }, + { + "epoch": 0.13092813500145475, + "flow_matching_loss": 0.09461152106523514, + "grad_norm": 0.6630043983459473, + "learning_rate": 4.49e-06, + "loss": 0.3251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08544972836971283, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05962145552039146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.23052091002464295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998645782470703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 450, + "weighted_flow_matching_loss": 0.09461152106523514, + "weighted_rkd_loss": 0.23048969507217407 + }, + { + "epoch": 0.13383764911259818, + "flow_matching_loss": 0.09653763771057129, + "grad_norm": 0.8007174730300903, + "learning_rate": 4.59e-06, + "loss": 0.3215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08358816355466843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05785164795815945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2250279739499092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998584032058716, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 460, + "weighted_flow_matching_loss": 0.09653763771057129, + "weighted_rkd_loss": 0.22499610781669616 + }, + { + "epoch": 0.13674716322374164, + "flow_matching_loss": 0.0938807561993599, + "grad_norm": 0.7467442154884338, + "learning_rate": 4.69e-06, + "loss": 0.3154, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0825396291911602, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0564431220293045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22152238041162492, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998521208763123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 470, + "weighted_flow_matching_loss": 0.0938807561993599, + "weighted_rkd_loss": 0.22148962020874025 + }, + { + "epoch": 0.13965667733488507, + "flow_matching_loss": 0.09231334254145622, + "grad_norm": 0.70647132396698, + "learning_rate": 4.790000000000001e-06, + "loss": 0.318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08377899006009101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05814089365303517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22569887489080429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999845689535141, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 480, + "weighted_flow_matching_loss": 0.09231334254145622, + "weighted_rkd_loss": 0.22566404342651367 + }, + { + "epoch": 0.1425661914460285, + "flow_matching_loss": 0.08810740038752556, + "grad_norm": 0.9095771908760071, + "learning_rate": 4.890000000000001e-06, + "loss": 0.3097, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08263843432068825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05638643130660057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.221663299202919, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998391091823577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 490, + "weighted_flow_matching_loss": 0.08810740038752556, + "weighted_rkd_loss": 0.22162764072418212 + }, + { + "epoch": 0.14547570555717196, + "flow_matching_loss": 0.0929937094449997, + "grad_norm": 0.7522123456001282, + "learning_rate": 4.9900000000000005e-06, + "loss": 0.312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08182251900434494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05543409399688244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2190791293978691, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998324036598205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 500, + "weighted_flow_matching_loss": 0.0929937094449997, + "weighted_rkd_loss": 0.21904240995645524 + }, + { + "epoch": 0.1483852196683154, + "flow_matching_loss": 0.08868019953370095, + "grad_norm": 1.006281852722168, + "learning_rate": 5.0899999999999995e-06, + "loss": 0.3129, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08315176442265511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05791780725121498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.224221333861351, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998255550861359, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 510, + "weighted_flow_matching_loss": 0.08868019953370095, + "weighted_rkd_loss": 0.2241822138428688 + }, + { + "epoch": 0.15129473377945882, + "flow_matching_loss": 0.08708657696843147, + "grad_norm": 0.9609445929527283, + "learning_rate": 5.1899999999999994e-06, + "loss": 0.3101, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08303640633821488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.056968745589256284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.22304155975580214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998185694217682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 520, + "weighted_flow_matching_loss": 0.08708657696843147, + "weighted_rkd_loss": 0.22300110310316085 + }, + { + "epoch": 0.15420424789060228, + "flow_matching_loss": 0.08783690482378007, + "grad_norm": 0.7173924446105957, + "learning_rate": 5.29e-06, + "loss": 0.307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08212394788861274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05491791032254696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2191658064723015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998114585876465, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 530, + "weighted_flow_matching_loss": 0.08783690482378007, + "weighted_rkd_loss": 0.21912448853254318 + }, + { + "epoch": 0.1571137620017457, + "flow_matching_loss": 0.08517100214958191, + "grad_norm": 1.2518157958984375, + "learning_rate": 5.39e-06, + "loss": 0.3028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08125649392604828, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05520350895822048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21771649718284608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9998041927814484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 540, + "weighted_flow_matching_loss": 0.08517100214958191, + "weighted_rkd_loss": 0.217673859000206 + }, + { + "epoch": 0.16002327611288913, + "flow_matching_loss": 0.08946003690361977, + "grad_norm": 0.7362799048423767, + "learning_rate": 5.49e-06, + "loss": 0.3038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08011512383818627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.054182767122983935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21441301256418227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997968018054962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 550, + "weighted_flow_matching_loss": 0.08946003690361977, + "weighted_rkd_loss": 0.21436944752931594 + }, + { + "epoch": 0.1629327902240326, + "flow_matching_loss": 0.08586404323577881, + "grad_norm": 0.637808084487915, + "learning_rate": 5.59e-06, + "loss": 0.3012, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08055554777383804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.054310758411884305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2154218539595604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997892677783966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 560, + "weighted_flow_matching_loss": 0.08586404323577881, + "weighted_rkd_loss": 0.21537645608186723 + }, + { + "epoch": 0.16584230433517602, + "flow_matching_loss": 0.08943888023495675, + "grad_norm": 0.7390864491462708, + "learning_rate": 5.690000000000001e-06, + "loss": 0.3041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08072824627161027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05324965752661228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21470614969730378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997815787792206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 570, + "weighted_flow_matching_loss": 0.08943888023495675, + "weighted_rkd_loss": 0.21465925872325897 + }, + { + "epoch": 0.16875181844631945, + "flow_matching_loss": 0.08474606052041053, + "grad_norm": 0.8167809844017029, + "learning_rate": 5.7900000000000005e-06, + "loss": 0.3, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08029596209526062, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.054733626544475555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2153255507349968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997737944126129, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 580, + "weighted_flow_matching_loss": 0.08474606052041053, + "weighted_rkd_loss": 0.2152768239378929 + }, + { + "epoch": 0.1716613325574629, + "flow_matching_loss": 0.08445861786603928, + "grad_norm": 0.7014927864074707, + "learning_rate": 5.89e-06, + "loss": 0.2989, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.08020877912640571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05402976050972939, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21444731801748276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997658431529999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 590, + "weighted_flow_matching_loss": 0.08445861786603928, + "weighted_rkd_loss": 0.21439709812402724 + }, + { + "epoch": 0.17457084666860634, + "flow_matching_loss": 0.08426291421055794, + "grad_norm": 0.7042760252952576, + "learning_rate": 5.99e-06, + "loss": 0.2949, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07918534129858017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05231046862900257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21068114936351776, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997577667236328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 600, + "weighted_flow_matching_loss": 0.08426291421055794, + "weighted_rkd_loss": 0.21063010543584823 + }, + { + "epoch": 0.17748036077974977, + "flow_matching_loss": 0.08502091318368912, + "grad_norm": 0.6444291472434998, + "learning_rate": 6.090000000000001e-06, + "loss": 0.295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07889029681682587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.052243133634328844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21002372801303865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997495412826538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 610, + "weighted_flow_matching_loss": 0.08502091318368912, + "weighted_rkd_loss": 0.2099711239337921 + }, + { + "epoch": 0.18038987489089323, + "flow_matching_loss": 0.08617300391197205, + "grad_norm": 0.8833863735198975, + "learning_rate": 6.19e-06, + "loss": 0.2944, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07845713496208191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05140906199812889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20832333117723464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997411906719208, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 620, + "weighted_flow_matching_loss": 0.08617300391197205, + "weighted_rkd_loss": 0.20826940685510636 + }, + { + "epoch": 0.18329938900203666, + "flow_matching_loss": 0.0797645129263401, + "grad_norm": 0.7839348912239075, + "learning_rate": 6.29e-06, + "loss": 0.2871, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07826979607343673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.050818595290184024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2073581874370575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997327029705048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 630, + "weighted_flow_matching_loss": 0.0797645129263401, + "weighted_rkd_loss": 0.20730275958776473 + }, + { + "epoch": 0.18620890311318009, + "flow_matching_loss": 0.08220899254083633, + "grad_norm": 0.8525615334510803, + "learning_rate": 6.39e-06, + "loss": 0.2906, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07846110314130783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05152080617845058, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20844301134347915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997240781784058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 640, + "weighted_flow_matching_loss": 0.08220899254083633, + "weighted_rkd_loss": 0.20838549733161926 + }, + { + "epoch": 0.18911841722432354, + "flow_matching_loss": 0.08250393643975258, + "grad_norm": 0.7629416584968567, + "learning_rate": 6.49e-06, + "loss": 0.2919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07861892282962799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05224643982946873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.2094842880964279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997153162956238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 650, + "weighted_flow_matching_loss": 0.08250393643975258, + "weighted_rkd_loss": 0.20942465215921402 + }, + { + "epoch": 0.19202793133546697, + "flow_matching_loss": 0.07984060272574425, + "grad_norm": 0.7958306074142456, + "learning_rate": 6.5900000000000004e-06, + "loss": 0.2903, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07914812788367272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.052181977406144144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.21047823280096054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9997064173221588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 660, + "weighted_flow_matching_loss": 0.07984060272574425, + "weighted_rkd_loss": 0.21041644215583802 + }, + { + "epoch": 0.1949374454466104, + "flow_matching_loss": 0.0773671254515648, + "grad_norm": 1.1440011262893677, + "learning_rate": 6.69e-06, + "loss": 0.2871, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07814187780022622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.053473472595214844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20975722670555114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996973752975464, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 670, + "weighted_flow_matching_loss": 0.0773671254515648, + "weighted_rkd_loss": 0.20969374030828475 + }, + { + "epoch": 0.19784695955775386, + "flow_matching_loss": 0.0805439293384552, + "grad_norm": 0.7460523247718811, + "learning_rate": 6.79e-06, + "loss": 0.2855, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07690391466021537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05119000338017941, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20499783307313918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996882021427155, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 680, + "weighted_flow_matching_loss": 0.0805439293384552, + "weighted_rkd_loss": 0.20493389815092086 + }, + { + "epoch": 0.2007564736688973, + "flow_matching_loss": 0.07923628911375999, + "grad_norm": 0.6386614441871643, + "learning_rate": 6.89e-06, + "loss": 0.2827, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07684445083141327, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04981967136263847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20350857228040695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999678885936737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 690, + "weighted_flow_matching_loss": 0.07923628911375999, + "weighted_rkd_loss": 0.20344322323799133 + }, + { + "epoch": 0.20366598778004075, + "flow_matching_loss": 0.08030270040035248, + "grad_norm": 0.776763379573822, + "learning_rate": 6.990000000000001e-06, + "loss": 0.2792, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07539087384939194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0481680516153574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19894980043172836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996694326400757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 700, + "weighted_flow_matching_loss": 0.08030270040035248, + "weighted_rkd_loss": 0.1988840341567993 + }, + { + "epoch": 0.20657550189118418, + "flow_matching_loss": 0.08024234548211098, + "grad_norm": 0.705723226070404, + "learning_rate": 7.090000000000001e-06, + "loss": 0.2864, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07731053382158279, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.05158485695719719, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20620592683553696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996598482131958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 710, + "weighted_flow_matching_loss": 0.08024234548211098, + "weighted_rkd_loss": 0.2061357691884041 + }, + { + "epoch": 0.2094850160023276, + "flow_matching_loss": 0.07555915042757988, + "grad_norm": 0.6222692131996155, + "learning_rate": 7.19e-06, + "loss": 0.2781, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07649101391434669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04959938079118729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.20258140861988067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996501326560974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 720, + "weighted_flow_matching_loss": 0.07555915042757988, + "weighted_rkd_loss": 0.20251053422689438 + }, + { + "epoch": 0.21239453011347106, + "flow_matching_loss": 0.07591287195682525, + "grad_norm": 0.742981493473053, + "learning_rate": 7.29e-06, + "loss": 0.2725, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07470131143927575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04729030355811119, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19669292867183685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996402621269226, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 730, + "weighted_flow_matching_loss": 0.07591287195682525, + "weighted_rkd_loss": 0.19662217497825624 + }, + { + "epoch": 0.2153040442246145, + "flow_matching_loss": 0.07665505446493626, + "grad_norm": 0.7453432679176331, + "learning_rate": 7.3899999999999995e-06, + "loss": 0.2759, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07547487840056419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04835131131112576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19930106848478318, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996302723884583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 740, + "weighted_flow_matching_loss": 0.07665505446493626, + "weighted_rkd_loss": 0.19922737926244735 + }, + { + "epoch": 0.21821355833575792, + "flow_matching_loss": 0.08112266734242439, + "grad_norm": 0.6849518418312073, + "learning_rate": 7.49e-06, + "loss": 0.2762, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07381652593612671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04756320416927338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1951962575316429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996201395988464, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 750, + "weighted_flow_matching_loss": 0.08112266734242439, + "weighted_rkd_loss": 0.1951221004128456 + }, + { + "epoch": 0.22112307244690138, + "flow_matching_loss": 0.07835286781191826, + "grad_norm": 0.5896384119987488, + "learning_rate": 7.59e-06, + "loss": 0.2744, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07408303618431092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04795410484075546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1961201786994934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9996098577976227, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 760, + "weighted_flow_matching_loss": 0.07835286781191826, + "weighted_rkd_loss": 0.19604366272687912 + }, + { + "epoch": 0.2240325865580448, + "flow_matching_loss": 0.07748444080352783, + "grad_norm": 0.732711136341095, + "learning_rate": 7.690000000000001e-06, + "loss": 0.2698, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07292542904615402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04654646031558514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19239732176065444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995994508266449, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 770, + "weighted_flow_matching_loss": 0.07748444080352783, + "weighted_rkd_loss": 0.192320254445076 + }, + { + "epoch": 0.22694210066918824, + "flow_matching_loss": 0.07352827377617359, + "grad_norm": 0.6509157419204712, + "learning_rate": 7.79e-06, + "loss": 0.2647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0727703869342804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04570450931787491, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19124528169631957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995889067649841, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 780, + "weighted_flow_matching_loss": 0.07352827377617359, + "weighted_rkd_loss": 0.19116665422916412 + }, + { + "epoch": 0.2298516147803317, + "flow_matching_loss": 0.07524842880666256, + "grad_norm": 0.6646036505699158, + "learning_rate": 7.89e-06, + "loss": 0.2638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07182851731777191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04499739445745945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18865442872047425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995782315731049, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 790, + "weighted_flow_matching_loss": 0.07524842880666256, + "weighted_rkd_loss": 0.18857485800981522 + }, + { + "epoch": 0.23276112889147513, + "flow_matching_loss": 0.07490801960229873, + "grad_norm": 0.6892244815826416, + "learning_rate": 7.99e-06, + "loss": 0.2748, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07561262398958206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.048727110773324964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19995235800743102, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995674014091491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 800, + "weighted_flow_matching_loss": 0.07490801960229873, + "weighted_rkd_loss": 0.19986587166786193 + }, + { + "epoch": 0.23567064300261856, + "flow_matching_loss": 0.07829235792160034, + "grad_norm": 0.6612560153007507, + "learning_rate": 8.09e-06, + "loss": 0.2777, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0754216268658638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04864251017570496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19948576390743256, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995564520359039, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 810, + "weighted_flow_matching_loss": 0.07829235792160034, + "weighted_rkd_loss": 0.19939727783203126 + }, + { + "epoch": 0.23858015711376201, + "flow_matching_loss": 0.07518614903092384, + "grad_norm": 0.6589637398719788, + "learning_rate": 8.190000000000001e-06, + "loss": 0.2685, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07304719164967537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.047298996895551684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1933933824300766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995453536510468, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 820, + "weighted_flow_matching_loss": 0.07518614903092384, + "weighted_rkd_loss": 0.1933054581284523 + }, + { + "epoch": 0.24148967122490544, + "flow_matching_loss": 0.07303978130221367, + "grad_norm": 0.6259734630584717, + "learning_rate": 8.29e-06, + "loss": 0.2662, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07298777997493744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04723733067512512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19321288764476777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995341181755066, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 830, + "weighted_flow_matching_loss": 0.07303978130221367, + "weighted_rkd_loss": 0.19312286823987962 + }, + { + "epoch": 0.24439918533604887, + "flow_matching_loss": 0.07419711574912072, + "grad_norm": 0.6265513896942139, + "learning_rate": 8.390000000000001e-06, + "loss": 0.2662, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07300727441906929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04609453268349171, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.19210907965898513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995227515697479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 840, + "weighted_flow_matching_loss": 0.07419711574912072, + "weighted_rkd_loss": 0.19201738834381105 + }, + { + "epoch": 0.24730869944719233, + "flow_matching_loss": 0.0714146763086319, + "grad_norm": 0.5895506739616394, + "learning_rate": 8.49e-06, + "loss": 0.2584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07113439589738846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04475980326533317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18702859431505203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9995112538337707, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 850, + "weighted_flow_matching_loss": 0.0714146763086319, + "weighted_rkd_loss": 0.1869371697306633 + }, + { + "epoch": 0.25021821355833573, + "flow_matching_loss": 0.07680913433432579, + "grad_norm": 0.6060665249824524, + "learning_rate": 8.59e-06, + "loss": 0.2642, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07149481922388076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.044539109244942664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1875287488102913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994996070861817, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 860, + "weighted_flow_matching_loss": 0.07680913433432579, + "weighted_rkd_loss": 0.18743490427732468 + }, + { + "epoch": 0.2531277276694792, + "flow_matching_loss": 0.0710462260991335, + "grad_norm": 0.6766586899757385, + "learning_rate": 8.690000000000002e-06, + "loss": 0.2628, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07277815416455269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04633987545967102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1918961837887764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999487841129303, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 870, + "weighted_flow_matching_loss": 0.0710462260991335, + "weighted_rkd_loss": 0.19179787784814833 + }, + { + "epoch": 0.25603724178062265, + "flow_matching_loss": 0.07112515345215797, + "grad_norm": 0.7284952402114868, + "learning_rate": 8.79e-06, + "loss": 0.2588, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07131092175841332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04512246660888195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18774431049823762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994759202003479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 880, + "weighted_flow_matching_loss": 0.07112515345215797, + "weighted_rkd_loss": 0.18764590322971345 + }, + { + "epoch": 0.2589467558917661, + "flow_matching_loss": 0.07490266412496567, + "grad_norm": 0.7346463799476624, + "learning_rate": 8.89e-06, + "loss": 0.2566, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06942812949419022, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.042927513271570204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18178377151489258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994638621807098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 890, + "weighted_flow_matching_loss": 0.07490266412496567, + "weighted_rkd_loss": 0.18168631196022034 + }, + { + "epoch": 0.2618562700029095, + "flow_matching_loss": 0.07150647789239883, + "grad_norm": 0.7060105800628662, + "learning_rate": 8.99e-06, + "loss": 0.2576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07095074355602264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.044249072670936584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18615055978298187, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994516789913177, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 900, + "weighted_flow_matching_loss": 0.07150647789239883, + "weighted_rkd_loss": 0.18604848235845567 + }, + { + "epoch": 0.26476578411405294, + "flow_matching_loss": 0.07330931350588799, + "grad_norm": 0.6461315155029297, + "learning_rate": 9.09e-06, + "loss": 0.2576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07047767788171769, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04346267133951187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1844180256128311, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994393527507782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 910, + "weighted_flow_matching_loss": 0.07330931350588799, + "weighted_rkd_loss": 0.18431463092565536 + }, + { + "epoch": 0.26767529822519637, + "flow_matching_loss": 0.07129888497292995, + "grad_norm": 0.6290547251701355, + "learning_rate": 9.19e-06, + "loss": 0.2567, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07091343253850937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04371358416974545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18554044961929322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994268715381622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 920, + "weighted_flow_matching_loss": 0.07129888497292995, + "weighted_rkd_loss": 0.18543411195278167 + }, + { + "epoch": 0.27058481233633985, + "flow_matching_loss": 0.071526437997818, + "grad_norm": 0.744698166847229, + "learning_rate": 9.289999999999999e-06, + "loss": 0.257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07019438669085502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04514442943036556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18553320169448853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994142949581146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 930, + "weighted_flow_matching_loss": 0.071526437997818, + "weighted_rkd_loss": 0.1854245290160179 + }, + { + "epoch": 0.2734943264474833, + "flow_matching_loss": 0.0720653485506773, + "grad_norm": 0.5706762671470642, + "learning_rate": 9.39e-06, + "loss": 0.254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06989560574293137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04229824207723141, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18208945393562317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9994015514850616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 940, + "weighted_flow_matching_loss": 0.0720653485506773, + "weighted_rkd_loss": 0.1819804683327675 + }, + { + "epoch": 0.2764038405586267, + "flow_matching_loss": 0.0712877407670021, + "grad_norm": 0.6607420444488525, + "learning_rate": 9.49e-06, + "loss": 0.2554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07006370425224304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04406103007495403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18418843895196915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993886768817901, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 950, + "weighted_flow_matching_loss": 0.0712877407670021, + "weighted_rkd_loss": 0.18407583832740784 + }, + { + "epoch": 0.27931335466977014, + "flow_matching_loss": 0.07200824208557606, + "grad_norm": 0.677792489528656, + "learning_rate": 9.59e-06, + "loss": 0.2569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.07053182423114776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04399193413555622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18505558222532273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993756711483002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 960, + "weighted_flow_matching_loss": 0.07200824208557606, + "weighted_rkd_loss": 0.18494004607200623 + }, + { + "epoch": 0.28222286878091357, + "flow_matching_loss": 0.06952105313539506, + "grad_norm": 0.9199199080467224, + "learning_rate": 9.69e-06, + "loss": 0.252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0696253940463066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04332479760050774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.182575586438179, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993625104427337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 970, + "weighted_flow_matching_loss": 0.06952105313539506, + "weighted_rkd_loss": 0.1824591875076294 + }, + { + "epoch": 0.285132382892057, + "flow_matching_loss": 0.0701353169977665, + "grad_norm": 0.7141896486282349, + "learning_rate": 9.79e-06, + "loss": 0.2506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06879517361521721, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04296623468399048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18055658042430878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993492364883423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 980, + "weighted_flow_matching_loss": 0.0701353169977665, + "weighted_rkd_loss": 0.1804390847682953 + }, + { + "epoch": 0.2880418970032005, + "flow_matching_loss": 0.06777559295296669, + "grad_norm": 0.6364426016807556, + "learning_rate": 9.89e-06, + "loss": 0.2462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06864332109689712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04124658517539501, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1785332292318344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993358135223389, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 990, + "weighted_flow_matching_loss": 0.06777559295296669, + "weighted_rkd_loss": 0.1784146562218666 + }, + { + "epoch": 0.2909514111143439, + "flow_matching_loss": 0.0682857658714056, + "grad_norm": 0.7142160534858704, + "learning_rate": 9.990000000000001e-06, + "loss": 0.2502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06932044178247451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.043411750346422195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18205263316631318, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993222475051879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1000, + "weighted_flow_matching_loss": 0.0682857658714056, + "weighted_rkd_loss": 0.18192919641733168 + }, + { + "epoch": 0.29386092522548735, + "flow_matching_loss": 0.07061850540339946, + "grad_norm": 0.6974548697471619, + "learning_rate": 1.009e-05, + "loss": 0.2503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06868342310190201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04242396540939808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17979080975055695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9993085503578186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1010, + "weighted_flow_matching_loss": 0.07061850540339946, + "weighted_rkd_loss": 0.1796665072441101 + }, + { + "epoch": 0.2967704393366308, + "flow_matching_loss": 0.06650051437318324, + "grad_norm": 0.6738964319229126, + "learning_rate": 1.019e-05, + "loss": 0.2496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06996619552373887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0432788223028183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.18321121335029603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992947280406952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1020, + "weighted_flow_matching_loss": 0.06650051437318324, + "weighted_rkd_loss": 0.1830819934606552 + }, + { + "epoch": 0.2996799534477742, + "flow_matching_loss": 0.06653415709733963, + "grad_norm": 0.5752034783363342, + "learning_rate": 1.0290000000000001e-05, + "loss": 0.245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06847032979130745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0416898101568222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17863047271966934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992807626724243, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1030, + "weighted_flow_matching_loss": 0.06653415709733963, + "weighted_rkd_loss": 0.1785019963979721 + }, + { + "epoch": 0.30258946755891764, + "flow_matching_loss": 0.06748146675527096, + "grad_norm": 0.6195955276489258, + "learning_rate": 1.039e-05, + "loss": 0.2412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06681029647588729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040251510590314864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1738721027970314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992666482925415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1040, + "weighted_flow_matching_loss": 0.06748146675527096, + "weighted_rkd_loss": 0.17374458611011506 + }, + { + "epoch": 0.3054989816700611, + "flow_matching_loss": 0.06478513590991497, + "grad_norm": 0.6037260293960571, + "learning_rate": 1.0490000000000001e-05, + "loss": 0.2421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06781971529126167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04178802520036697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1774274542927742, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992524206638336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1050, + "weighted_flow_matching_loss": 0.06478513590991497, + "weighted_rkd_loss": 0.1772948056459427 + }, + { + "epoch": 0.30840849578120455, + "flow_matching_loss": 0.06623898446559906, + "grad_norm": 0.6009768843650818, + "learning_rate": 1.059e-05, + "loss": 0.2421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06774492561817169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0405364952981472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17602634727954863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992380261421203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1060, + "weighted_flow_matching_loss": 0.06623898446559906, + "weighted_rkd_loss": 0.1758922293782234 + }, + { + "epoch": 0.311318009892348, + "flow_matching_loss": 0.06769019104540348, + "grad_norm": 0.615286648273468, + "learning_rate": 1.0690000000000001e-05, + "loss": 0.2446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06766204237937927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04176550805568695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1770895928144455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992235124111175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1070, + "weighted_flow_matching_loss": 0.06769019104540348, + "weighted_rkd_loss": 0.1769520804286003 + }, + { + "epoch": 0.3142275240034914, + "flow_matching_loss": 0.06501130312681198, + "grad_norm": 0.6789621710777283, + "learning_rate": 1.0790000000000002e-05, + "loss": 0.2359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06597024984657765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03912312164902687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1710636243224144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9992088615894318, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1080, + "weighted_flow_matching_loss": 0.06501130312681198, + "weighted_rkd_loss": 0.17092826217412949 + }, + { + "epoch": 0.31713703811463484, + "flow_matching_loss": 0.06427712552249432, + "grad_norm": 0.6440759301185608, + "learning_rate": 1.089e-05, + "loss": 0.2351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06572520472109318, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03947395943105221, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17092436850070952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991940557956696, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1090, + "weighted_flow_matching_loss": 0.06427712552249432, + "weighted_rkd_loss": 0.17078662663698196 + }, + { + "epoch": 0.32004655222577827, + "flow_matching_loss": 0.06708350665867328, + "grad_norm": 0.6923602223396301, + "learning_rate": 1.099e-05, + "loss": 0.2381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06554259695112705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.040113503113389015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17119869589805603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991791367530822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1100, + "weighted_flow_matching_loss": 0.06708350665867328, + "weighted_rkd_loss": 0.17105815857648848 + }, + { + "epoch": 0.32295606633692175, + "flow_matching_loss": 0.06311287768185139, + "grad_norm": 0.64277184009552, + "learning_rate": 1.1089999999999999e-05, + "loss": 0.2347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0659221924841404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.039887198433279994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1717315837740898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999164080619812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1110, + "weighted_flow_matching_loss": 0.06311287768185139, + "weighted_rkd_loss": 0.1715880364179611 + }, + { + "epoch": 0.3258655804480652, + "flow_matching_loss": 0.06459731608629227, + "grad_norm": 0.5594890117645264, + "learning_rate": 1.119e-05, + "loss": 0.2334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06501440890133381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03895126543939113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16898008286952973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991488814353943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1120, + "weighted_flow_matching_loss": 0.06459731608629227, + "weighted_rkd_loss": 0.16883628368377684 + }, + { + "epoch": 0.3287750945592086, + "flow_matching_loss": 0.06804474592208862, + "grad_norm": 0.7491141557693481, + "learning_rate": 1.129e-05, + "loss": 0.2391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06549725458025932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04019718728959561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1711916968226433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991335451602936, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1130, + "weighted_flow_matching_loss": 0.06804474592208862, + "weighted_rkd_loss": 0.17104337066411973 + }, + { + "epoch": 0.33168460867035204, + "flow_matching_loss": 0.06503581143915653, + "grad_norm": 0.6382285952568054, + "learning_rate": 1.139e-05, + "loss": 0.2379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06617366522550583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.04064494967460632, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.172992280125618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991180717945098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1140, + "weighted_flow_matching_loss": 0.06503581143915653, + "weighted_rkd_loss": 0.17283970564603807 + }, + { + "epoch": 0.3345941227814955, + "flow_matching_loss": 0.06813149787485599, + "grad_norm": 0.5795291066169739, + "learning_rate": 1.149e-05, + "loss": 0.2403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06636480912566185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03962232023477554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.17235193848609925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9991024553775787, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1150, + "weighted_flow_matching_loss": 0.06813149787485599, + "weighted_rkd_loss": 0.17219724506139755 + }, + { + "epoch": 0.3375036368926389, + "flow_matching_loss": 0.06660553440451622, + "grad_norm": 0.6970534324645996, + "learning_rate": 1.1589999999999999e-05, + "loss": 0.2359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06518320329487323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.039132575318217275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16949898153543472, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990867078304291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1160, + "weighted_flow_matching_loss": 0.06660553440451622, + "weighted_rkd_loss": 0.16934420615434648 + }, + { + "epoch": 0.3404131510037824, + "flow_matching_loss": 0.06454597413539886, + "grad_norm": 0.530734121799469, + "learning_rate": 1.169e-05, + "loss": 0.2334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06481802277266979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03939153477549553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16902758032083512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999070817232132, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1170, + "weighted_flow_matching_loss": 0.06454597413539886, + "weighted_rkd_loss": 0.16887052953243256 + }, + { + "epoch": 0.3433226651149258, + "flow_matching_loss": 0.06852093748748303, + "grad_norm": 0.5841692686080933, + "learning_rate": 1.179e-05, + "loss": 0.2375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06480909250676632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03952840156853199, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16914658844470978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990548193454742, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1180, + "weighted_flow_matching_loss": 0.06852093748748303, + "weighted_rkd_loss": 0.1689867228269577 + }, + { + "epoch": 0.34623217922606925, + "flow_matching_loss": 0.06458857245743274, + "grad_norm": 0.5904214978218079, + "learning_rate": 1.189e-05, + "loss": 0.2324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06418657787144184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03962148129940033, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16799463480710983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990386366844177, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1190, + "weighted_flow_matching_loss": 0.06458857245743274, + "weighted_rkd_loss": 0.16783316433429718 + }, + { + "epoch": 0.3491416933372127, + "flow_matching_loss": 0.06399866938591003, + "grad_norm": 0.5365495085716248, + "learning_rate": 1.199e-05, + "loss": 0.2303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06399146839976311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0385175596922636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16650049686431884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9990223526954651, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1200, + "weighted_flow_matching_loss": 0.06399866938591003, + "weighted_rkd_loss": 0.1663377359509468 + }, + { + "epoch": 0.3520512074483561, + "flow_matching_loss": 0.06430470831692219, + "grad_norm": 0.652228593826294, + "learning_rate": 1.2090000000000001e-05, + "loss": 0.2292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06311481036245822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03885001540184021, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16507963538169862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.999005913734436, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1210, + "weighted_flow_matching_loss": 0.06430470831692219, + "weighted_rkd_loss": 0.16491552740335463 + }, + { + "epoch": 0.35496072155949954, + "flow_matching_loss": 0.0647670492529869, + "grad_norm": 0.5240232944488525, + "learning_rate": 1.219e-05, + "loss": 0.229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06342056579887867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.037603475525975226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16444460898637772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998989337682724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1220, + "weighted_flow_matching_loss": 0.0647670492529869, + "weighted_rkd_loss": 0.16427839994430543 + }, + { + "epoch": 0.357870235670643, + "flow_matching_loss": 0.06376069262623787, + "grad_norm": 0.539127767086029, + "learning_rate": 1.2290000000000001e-05, + "loss": 0.2273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06279191859066487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03815358821302652, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16373742520809173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989726364612579, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1230, + "weighted_flow_matching_loss": 0.06376069262623787, + "weighted_rkd_loss": 0.16356921941041946 + }, + { + "epoch": 0.36077974978178645, + "flow_matching_loss": 0.06312300376594067, + "grad_norm": 0.5767908096313477, + "learning_rate": 1.239e-05, + "loss": 0.2245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06217693649232388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.037171924114227296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16152579784393312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989557921886444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1240, + "weighted_flow_matching_loss": 0.06312300376594067, + "weighted_rkd_loss": 0.16135714948177338 + }, + { + "epoch": 0.3636892638929299, + "flow_matching_loss": 0.06185674108564854, + "grad_norm": 0.6265656352043152, + "learning_rate": 1.249e-05, + "loss": 0.2265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06341938301920891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03795466236770153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1647934287786484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989388108253479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1250, + "weighted_flow_matching_loss": 0.06185674108564854, + "weighted_rkd_loss": 0.16461855471134185 + }, + { + "epoch": 0.3665987780040733, + "flow_matching_loss": 0.062290961667895314, + "grad_norm": 0.5840176939964294, + "learning_rate": 1.2590000000000001e-05, + "loss": 0.2225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.061511201038956645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03740823566913605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1604306384921074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989216983318329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1260, + "weighted_flow_matching_loss": 0.062290961667895314, + "weighted_rkd_loss": 0.1602576419711113 + }, + { + "epoch": 0.36950829211521674, + "flow_matching_loss": 0.06340278871357441, + "grad_norm": 0.6387100219726562, + "learning_rate": 1.269e-05, + "loss": 0.2248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.062165068835020064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03722011856734753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16155025660991668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9989044368267059, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1270, + "weighted_flow_matching_loss": 0.06340278871357441, + "weighted_rkd_loss": 0.16137327253818512 + }, + { + "epoch": 0.37241780622636017, + "flow_matching_loss": 0.06579688563942909, + "grad_norm": 0.5477023720741272, + "learning_rate": 1.2790000000000001e-05, + "loss": 0.2276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.062176978960633275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03761634267866611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1619703009724617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988870561122895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1280, + "weighted_flow_matching_loss": 0.06579688563942909, + "weighted_rkd_loss": 0.16179000735282897 + }, + { + "epoch": 0.37532732033750366, + "flow_matching_loss": 0.06416797079145908, + "grad_norm": 0.5831474661827087, + "learning_rate": 1.289e-05, + "loss": 0.2253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06225238107144833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03680981509387493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16131457686424255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988695383071899, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1290, + "weighted_flow_matching_loss": 0.06416797079145908, + "weighted_rkd_loss": 0.16113223135471344 + }, + { + "epoch": 0.3782368344486471, + "flow_matching_loss": 0.06138844192028046, + "grad_norm": 1.363525390625, + "learning_rate": 1.2990000000000001e-05, + "loss": 0.2198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06135286539793015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.035845667123794556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1585514008998871, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988518714904785, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1300, + "weighted_flow_matching_loss": 0.06138844192028046, + "weighted_rkd_loss": 0.15836936682462693 + }, + { + "epoch": 0.3811463485597905, + "flow_matching_loss": 0.06152864061295986, + "grad_norm": 0.6831033229827881, + "learning_rate": 1.309e-05, + "loss": 0.2223, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0616653710603714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03767133168876171, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1610020712018013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988340735435486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1310, + "weighted_flow_matching_loss": 0.06152864061295986, + "weighted_rkd_loss": 0.1608143299818039 + }, + { + "epoch": 0.38405586267093395, + "flow_matching_loss": 0.061520756036043164, + "grad_norm": 0.6293010711669922, + "learning_rate": 1.3189999999999999e-05, + "loss": 0.2233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06261815950274467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03674996681511402, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.16198628395795822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9988161325454712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1320, + "weighted_flow_matching_loss": 0.061520756036043164, + "weighted_rkd_loss": 0.16179450750350952 + }, + { + "epoch": 0.3869653767820774, + "flow_matching_loss": 0.06257367245852948, + "grad_norm": 0.6122461557388306, + "learning_rate": 1.329e-05, + "loss": 0.2172, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05970876030623913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03536820039153099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15478572398424148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987980604171753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1330, + "weighted_flow_matching_loss": 0.06257367245852948, + "weighted_rkd_loss": 0.15459967404603958 + }, + { + "epoch": 0.3898748908932208, + "flow_matching_loss": 0.058758814632892606, + "grad_norm": 0.5892841219902039, + "learning_rate": 1.339e-05, + "loss": 0.2165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06086628399789333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03616726882755757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15789983719587325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987798511981965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1340, + "weighted_flow_matching_loss": 0.058758814632892606, + "weighted_rkd_loss": 0.15770718306303025 + }, + { + "epoch": 0.3927844050043643, + "flow_matching_loss": 0.06284408830106258, + "grad_norm": 0.6827016472816467, + "learning_rate": 1.349e-05, + "loss": 0.2257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06238686293363571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.038279734551906586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.163053460419178, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987614989280701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1350, + "weighted_flow_matching_loss": 0.06284408830106258, + "weighted_rkd_loss": 0.16285153329372407 + }, + { + "epoch": 0.3956939191155077, + "flow_matching_loss": 0.06169194802641868, + "grad_norm": 0.5867838263511658, + "learning_rate": 1.359e-05, + "loss": 0.2202, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06092356517910957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03683057799935341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15867770910263063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987430155277253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1360, + "weighted_flow_matching_loss": 0.06169194802641868, + "weighted_rkd_loss": 0.1584782660007477 + }, + { + "epoch": 0.39860343322665115, + "flow_matching_loss": 0.0639577254652977, + "grad_norm": 0.605930507183075, + "learning_rate": 1.369e-05, + "loss": 0.2186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05947614163160324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.035842386446893214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15479467064142227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987244009971619, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1370, + "weighted_flow_matching_loss": 0.0639577254652977, + "weighted_rkd_loss": 0.1545971989631653 + }, + { + "epoch": 0.4015129473377946, + "flow_matching_loss": 0.05999830365180969, + "grad_norm": 0.5016064643859863, + "learning_rate": 1.379e-05, + "loss": 0.2156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06023227423429489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0353527270257473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15581727623939515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9987056434154511, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1380, + "weighted_flow_matching_loss": 0.05999830365180969, + "weighted_rkd_loss": 0.15561558306217194 + }, + { + "epoch": 0.404422461448938, + "flow_matching_loss": 0.06105146668851376, + "grad_norm": 0.7059028744697571, + "learning_rate": 1.389e-05, + "loss": 0.22, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06092785522341728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.037295582517981526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.159151291847229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986867487430573, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1390, + "weighted_flow_matching_loss": 0.06105146668851376, + "weighted_rkd_loss": 0.15894228368997573 + }, + { + "epoch": 0.4073319755600815, + "flow_matching_loss": 0.06416475549340248, + "grad_norm": 0.6147602200508118, + "learning_rate": 1.399e-05, + "loss": 0.2213, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06067289598286152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.035957798920571805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15730358958244323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986677169799805, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1400, + "weighted_flow_matching_loss": 0.06416475549340248, + "weighted_rkd_loss": 0.15709396600723266 + }, + { + "epoch": 0.4102414896712249, + "flow_matching_loss": 0.06264398954808711, + "grad_norm": 0.5866729617118835, + "learning_rate": 1.409e-05, + "loss": 0.2191, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.06029133461415768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03606866169720888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15665132999420167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986485421657563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1410, + "weighted_flow_matching_loss": 0.06264398954808711, + "weighted_rkd_loss": 0.1564395993947983 + }, + { + "epoch": 0.41315100378236835, + "flow_matching_loss": 0.0590840719640255, + "grad_norm": 0.5870229005813599, + "learning_rate": 1.419e-05, + "loss": 0.2129, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05936398543417454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.035317326337099074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15404529571533204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998629242181778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1420, + "weighted_flow_matching_loss": 0.0590840719640255, + "weighted_rkd_loss": 0.15383413583040237 + }, + { + "epoch": 0.4160605178935118, + "flow_matching_loss": 0.06227588728070259, + "grad_norm": 0.6109619736671448, + "learning_rate": 1.429e-05, + "loss": 0.2165, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.059579220786690715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.035279137454926966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1544375792145729, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9986097991466523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1430, + "weighted_flow_matching_loss": 0.06227588728070259, + "weighted_rkd_loss": 0.15422288477420806 + }, + { + "epoch": 0.4189700320046552, + "flow_matching_loss": 0.06098487079143524, + "grad_norm": 0.5176849961280823, + "learning_rate": 1.4390000000000001e-05, + "loss": 0.2155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.059667091071605685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.035416897386312485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15475107878446578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985902011394501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1440, + "weighted_flow_matching_loss": 0.06098487079143524, + "weighted_rkd_loss": 0.15453291982412337 + }, + { + "epoch": 0.42187954611579864, + "flow_matching_loss": 0.06077432669699192, + "grad_norm": 0.5647070407867432, + "learning_rate": 1.449e-05, + "loss": 0.2089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.057412998378276826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03347273096442223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14829872846603392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985705077648163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1450, + "weighted_flow_matching_loss": 0.06077432669699192, + "weighted_rkd_loss": 0.14808675050735473 + }, + { + "epoch": 0.42478906022694213, + "flow_matching_loss": 0.06021335609257221, + "grad_norm": 0.6295836567878723, + "learning_rate": 1.4590000000000001e-05, + "loss": 0.2166, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05976539067924023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03705167099833488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15658245384693145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985506594181061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1460, + "weighted_flow_matching_loss": 0.06021335609257221, + "weighted_rkd_loss": 0.15635550767183304 + }, + { + "epoch": 0.42769857433808556, + "flow_matching_loss": 0.05982303060591221, + "grad_norm": 0.5745300054550171, + "learning_rate": 1.469e-05, + "loss": 0.2085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05743393860757351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03400519043207169, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14887306839227676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985306680202484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1470, + "weighted_flow_matching_loss": 0.05982303060591221, + "weighted_rkd_loss": 0.1486543297767639 + }, + { + "epoch": 0.430608088449229, + "flow_matching_loss": 0.05818596594035626, + "grad_norm": 0.5870999693870544, + "learning_rate": 1.479e-05, + "loss": 0.2115, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.059471774473786355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034586765244603156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1535303145647049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9985105395317078, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1480, + "weighted_flow_matching_loss": 0.05818596594035626, + "weighted_rkd_loss": 0.15330163836479188 + }, + { + "epoch": 0.4335176025603724, + "flow_matching_loss": 0.059531744569540024, + "grad_norm": 0.5458042621612549, + "learning_rate": 1.4890000000000001e-05, + "loss": 0.2095, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05790624618530273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03439147565513849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1502039685845375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998490285873413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1490, + "weighted_flow_matching_loss": 0.059531744569540024, + "weighted_rkd_loss": 0.14997722059488297 + }, + { + "epoch": 0.43642711667151585, + "flow_matching_loss": 0.06029549166560173, + "grad_norm": 0.5635703802108765, + "learning_rate": 1.499e-05, + "loss": 0.2146, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0591334767639637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.036245496571064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1545124515891075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998469889163971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1500, + "weighted_flow_matching_loss": 0.06029549166560173, + "weighted_rkd_loss": 0.1542760357260704 + }, + { + "epoch": 0.4393366307826593, + "flow_matching_loss": 0.06139848157763481, + "grad_norm": 0.5401477217674255, + "learning_rate": 1.5090000000000001e-05, + "loss": 0.2107, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05761522650718689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03430751468986273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14953796714544296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984493494033814, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1510, + "weighted_flow_matching_loss": 0.06139848157763481, + "weighted_rkd_loss": 0.14930611997842788 + }, + { + "epoch": 0.44224614489380276, + "flow_matching_loss": 0.06402742005884647, + "grad_norm": 0.5807371735572815, + "learning_rate": 1.519e-05, + "loss": 0.216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05886232368648052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.034456427209079266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15218107402324677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984286844730377, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1520, + "weighted_flow_matching_loss": 0.06402742005884647, + "weighted_rkd_loss": 0.15194197446107865 + }, + { + "epoch": 0.4451556590049462, + "flow_matching_loss": 0.06260242275893688, + "grad_norm": 0.5608530044555664, + "learning_rate": 1.529e-05, + "loss": 0.2128, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05746493488550186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03551623746752739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1504461094737053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9984078705310822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1530, + "weighted_flow_matching_loss": 0.06260242275893688, + "weighted_rkd_loss": 0.15020656883716582 + }, + { + "epoch": 0.4480651731160896, + "flow_matching_loss": 0.05956002362072468, + "grad_norm": 0.5210747718811035, + "learning_rate": 1.539e-05, + "loss": 0.2074, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05731587037444115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03347500581294298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14810674637556076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983869314193725, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1540, + "weighted_flow_matching_loss": 0.05956002362072468, + "weighted_rkd_loss": 0.1478678598999977 + }, + { + "epoch": 0.45097468722723305, + "flow_matching_loss": 0.05970576629042625, + "grad_norm": 0.5607830882072449, + "learning_rate": 1.549e-05, + "loss": 0.2056, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.056220512464642526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03368508275598288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1461261048913002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983658492565155, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1550, + "weighted_flow_matching_loss": 0.05970576629042625, + "weighted_rkd_loss": 0.14588733464479448 + }, + { + "epoch": 0.4538842013383765, + "flow_matching_loss": 0.05877410732209683, + "grad_norm": 0.5882740020751953, + "learning_rate": 1.559e-05, + "loss": 0.2063, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05670394338667393, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03439527936279774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14780316650867462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998344624042511, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1560, + "weighted_flow_matching_loss": 0.05877410732209683, + "weighted_rkd_loss": 0.1475585252046585 + }, + { + "epoch": 0.4567937154495199, + "flow_matching_loss": 0.05829805471003056, + "grad_norm": 0.6169350743293762, + "learning_rate": 1.569e-05, + "loss": 0.2048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05673755928874016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03329098802059889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1467661067843437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983232617378235, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1570, + "weighted_flow_matching_loss": 0.05829805471003056, + "weighted_rkd_loss": 0.14652000963687897 + }, + { + "epoch": 0.4597032295606634, + "flow_matching_loss": 0.057548172771930695, + "grad_norm": 0.6377531290054321, + "learning_rate": 1.579e-05, + "loss": 0.2079, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.057999709621071815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03456991240382194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1505693331360817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9983017861843109, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1580, + "weighted_flow_matching_loss": 0.057548172771930695, + "weighted_rkd_loss": 0.1503136411309242 + }, + { + "epoch": 0.4626127436718068, + "flow_matching_loss": 0.05894651263952255, + "grad_norm": 0.5890589356422424, + "learning_rate": 1.589e-05, + "loss": 0.2055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05650113932788372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03381718546152115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14681946188211442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982801616191864, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1590, + "weighted_flow_matching_loss": 0.05894651263952255, + "weighted_rkd_loss": 0.1465669721364975 + }, + { + "epoch": 0.46552225778295026, + "flow_matching_loss": 0.059783725813031194, + "grad_norm": 0.5366976857185364, + "learning_rate": 1.599e-05, + "loss": 0.2062, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05623951517045498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0341960184276104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14667504727840425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982583880424499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1600, + "weighted_flow_matching_loss": 0.059783725813031194, + "weighted_rkd_loss": 0.14641959071159363 + }, + { + "epoch": 0.4684317718940937, + "flow_matching_loss": 0.0616166990250349, + "grad_norm": 0.6528120040893555, + "learning_rate": 1.609e-05, + "loss": 0.2115, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05759902521967888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03495076559484005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.15014881491661072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9982364892959594, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1610, + "weighted_flow_matching_loss": 0.0616166990250349, + "weighted_rkd_loss": 0.14988401681184768 + }, + { + "epoch": 0.4713412860052371, + "flow_matching_loss": 0.05623636059463024, + "grad_norm": 0.5126147866249084, + "learning_rate": 1.619e-05, + "loss": 0.2026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.056711257249116895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03319262079894543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14661513715982438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998214453458786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1620, + "weighted_flow_matching_loss": 0.05623636059463024, + "weighted_rkd_loss": 0.1463533326983452 + }, + { + "epoch": 0.47425080011638054, + "flow_matching_loss": 0.05807607248425484, + "grad_norm": 0.6414763927459717, + "learning_rate": 1.629e-05, + "loss": 0.1981, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05418750531971454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.031947195157408716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1403222054243088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981922805309296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1630, + "weighted_flow_matching_loss": 0.05807607248425484, + "weighted_rkd_loss": 0.14006855338811874 + }, + { + "epoch": 0.47716031422752403, + "flow_matching_loss": 0.05775036178529262, + "grad_norm": 0.577686607837677, + "learning_rate": 1.639e-05, + "loss": 0.2, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.055179092660546304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03215635120868683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14251453727483748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981699764728547, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1640, + "weighted_flow_matching_loss": 0.05775036178529262, + "weighted_rkd_loss": 0.14225371181964874 + }, + { + "epoch": 0.48006982833866746, + "flow_matching_loss": 0.06113330461084843, + "grad_norm": 0.5099589824676514, + "learning_rate": 1.649e-05, + "loss": 0.201, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05409995019435883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.031917515024542806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14011741578578948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981475174427032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1650, + "weighted_flow_matching_loss": 0.06113330461084843, + "weighted_rkd_loss": 0.13985785990953445 + }, + { + "epoch": 0.4829793424498109, + "flow_matching_loss": 0.05370611809194088, + "grad_norm": 0.5872613191604614, + "learning_rate": 1.6590000000000002e-05, + "loss": 0.1919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05381052419543266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03087841458618641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1384994640946388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9981249570846558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1660, + "weighted_flow_matching_loss": 0.05370611809194088, + "weighted_rkd_loss": 0.13823977410793303 + }, + { + "epoch": 0.4858888565609543, + "flow_matching_loss": 0.05849910192191601, + "grad_norm": 0.5813825726509094, + "learning_rate": 1.669e-05, + "loss": 0.2023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05549469329416752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03308991063386202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1440792977809906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998102217912674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1670, + "weighted_flow_matching_loss": 0.05849910192191601, + "weighted_rkd_loss": 0.14380587190389632 + }, + { + "epoch": 0.48879837067209775, + "flow_matching_loss": 0.05550353266298771, + "grad_norm": 0.49163174629211426, + "learning_rate": 1.679e-05, + "loss": 0.1924, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05326992720365524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03064707648009062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13718693181872368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980793595314026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1680, + "weighted_flow_matching_loss": 0.05550353266298771, + "weighted_rkd_loss": 0.1369234211742878 + }, + { + "epoch": 0.4917078847832412, + "flow_matching_loss": 0.06208140514791012, + "grad_norm": 0.5948945879936218, + "learning_rate": 1.689e-05, + "loss": 0.2045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05524585582315922, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03219337910413742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14268508851528167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980563700199128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1690, + "weighted_flow_matching_loss": 0.06208140514791012, + "weighted_rkd_loss": 0.1424077570438385 + }, + { + "epoch": 0.49461739889438466, + "flow_matching_loss": 0.05968714840710163, + "grad_norm": 0.6160577535629272, + "learning_rate": 1.699e-05, + "loss": 0.2021, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.055005692318081854, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.032710271328687666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14272165298461914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9980332434177399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1700, + "weighted_flow_matching_loss": 0.05968714840710163, + "weighted_rkd_loss": 0.1424409568309784 + }, + { + "epoch": 0.4975269130055281, + "flow_matching_loss": 0.05596230775117874, + "grad_norm": 0.4984531104564667, + "learning_rate": 1.709e-05, + "loss": 0.188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05119093172252178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02991884555667639, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13230070993304252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.998009979724884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1710, + "weighted_flow_matching_loss": 0.05596230775117874, + "weighted_rkd_loss": 0.13203743994235992 + }, + { + "epoch": 0.5004364271166715, + "flow_matching_loss": 0.05697788745164871, + "grad_norm": 0.6086043119430542, + "learning_rate": 1.719e-05, + "loss": 0.196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05377676524221897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03179795295000076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13935148268938063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979865670204162, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1720, + "weighted_flow_matching_loss": 0.05697788745164871, + "weighted_rkd_loss": 0.13907092213630676 + }, + { + "epoch": 0.503345941227815, + "flow_matching_loss": 0.05891929604113102, + "grad_norm": 0.5567821264266968, + "learning_rate": 1.7290000000000002e-05, + "loss": 0.2003, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05447782017290592, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0326828945428133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.14163853526115416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979630291461945, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1730, + "weighted_flow_matching_loss": 0.05891929604113102, + "weighted_rkd_loss": 0.14135005250573157 + }, + { + "epoch": 0.5062554553389584, + "flow_matching_loss": 0.05610118098556995, + "grad_norm": 0.5245714783668518, + "learning_rate": 1.739e-05, + "loss": 0.1891, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05188671238720417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029541786387562753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13331520929932594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979393601417541, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1740, + "weighted_flow_matching_loss": 0.05610118098556995, + "weighted_rkd_loss": 0.13304048553109168 + }, + { + "epoch": 0.5091649694501018, + "flow_matching_loss": 0.05759929642081261, + "grad_norm": 0.5140729546546936, + "learning_rate": 1.749e-05, + "loss": 0.1916, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05199942737817764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030322078987956046, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13432093411684037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9979155480861663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1750, + "weighted_flow_matching_loss": 0.05759929642081261, + "weighted_rkd_loss": 0.1340409517288208 + }, + { + "epoch": 0.5120744835612453, + "flow_matching_loss": 0.0595508873462677, + "grad_norm": 0.6019054055213928, + "learning_rate": 1.7590000000000003e-05, + "loss": 0.1961, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.052801938727498055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.031264723464846614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13686860203742982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978915929794312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1760, + "weighted_flow_matching_loss": 0.0595508873462677, + "weighted_rkd_loss": 0.13658000975847245 + }, + { + "epoch": 0.5149839976723887, + "flow_matching_loss": 0.05804345794022083, + "grad_norm": 0.5405665636062622, + "learning_rate": 1.7690000000000002e-05, + "loss": 0.1948, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05308881923556328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030837083421647548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13701472356915473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978675127029419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1770, + "weighted_flow_matching_loss": 0.05804345794022083, + "weighted_rkd_loss": 0.13672254458069802 + }, + { + "epoch": 0.5178935117835322, + "flow_matching_loss": 0.05405088253319264, + "grad_norm": 0.5711567997932434, + "learning_rate": 1.779e-05, + "loss": 0.1908, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05282965488731861, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03136758673936128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13702689707279206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978432893753052, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1780, + "weighted_flow_matching_loss": 0.05405088253319264, + "weighted_rkd_loss": 0.13673135489225388 + }, + { + "epoch": 0.5208030258946756, + "flow_matching_loss": 0.05652104578912258, + "grad_norm": 0.6035967469215393, + "learning_rate": 1.7890000000000003e-05, + "loss": 0.1942, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0533910010010004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.031200664304196834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13798266649246216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9978189170360565, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1790, + "weighted_flow_matching_loss": 0.05652104578912258, + "weighted_rkd_loss": 0.13768169656395912 + }, + { + "epoch": 0.523712540005819, + "flow_matching_loss": 0.05743761844933033, + "grad_norm": 0.6653772592544556, + "learning_rate": 1.7990000000000002e-05, + "loss": 0.1871, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050424085929989815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02910866718739271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12995683774352074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977944314479827, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1800, + "weighted_flow_matching_loss": 0.05743761844933033, + "weighted_rkd_loss": 0.1296702116727829 + }, + { + "epoch": 0.5266220541169625, + "flow_matching_loss": 0.0556731965392828, + "grad_norm": 0.5921669006347656, + "learning_rate": 1.809e-05, + "loss": 0.1932, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.052973647788167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03193303868174553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13788033351302148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977697968482971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1810, + "weighted_flow_matching_loss": 0.0556731965392828, + "weighted_rkd_loss": 0.13757282570004464 + }, + { + "epoch": 0.5295315682281059, + "flow_matching_loss": 0.05571222528815269, + "grad_norm": 0.5556854009628296, + "learning_rate": 1.819e-05, + "loss": 0.1883, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.051653002947568895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029612818360328676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13291882276535033, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977450370788574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1820, + "weighted_flow_matching_loss": 0.05571222528815269, + "weighted_rkd_loss": 0.1326190784573555 + }, + { + "epoch": 0.5324410823392494, + "flow_matching_loss": 0.0564237155020237, + "grad_norm": 0.5312942266464233, + "learning_rate": 1.8290000000000003e-05, + "loss": 0.1862, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05038435310125351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029265640117228032, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13003434613347054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9977201223373413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1830, + "weighted_flow_matching_loss": 0.0564237155020237, + "weighted_rkd_loss": 0.12973788157105445 + }, + { + "epoch": 0.5353505964503927, + "flow_matching_loss": 0.05667468085885048, + "grad_norm": 0.48456326127052307, + "learning_rate": 1.8390000000000002e-05, + "loss": 0.1896, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05129343941807747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030662547424435617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13324942737817763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976950824260712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1840, + "weighted_flow_matching_loss": 0.05667468085885048, + "weighted_rkd_loss": 0.13294230177998542 + }, + { + "epoch": 0.5382601105615362, + "flow_matching_loss": 0.054743841663002966, + "grad_norm": 0.5790840983390808, + "learning_rate": 1.8489999999999997e-05, + "loss": 0.1853, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05062412992119789, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029659966379404067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1309082254767418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997669917345047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1850, + "weighted_flow_matching_loss": 0.054743841663002966, + "weighted_rkd_loss": 0.13060321137309075 + }, + { + "epoch": 0.5411696246726797, + "flow_matching_loss": 0.05756110474467278, + "grad_norm": 0.6424657106399536, + "learning_rate": 1.859e-05, + "loss": 0.188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.050635718554258344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029500465467572214, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13077190294861793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997644591331482, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1860, + "weighted_flow_matching_loss": 0.05756110474467278, + "weighted_rkd_loss": 0.13046384751796722 + }, + { + "epoch": 0.5440791387838231, + "flow_matching_loss": 0.05290517881512642, + "grad_norm": 0.5803369283676147, + "learning_rate": 1.869e-05, + "loss": 0.1871, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0519152894616127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03071385286748409, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13454443216323853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9976191461086273, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1870, + "weighted_flow_matching_loss": 0.05290517881512642, + "weighted_rkd_loss": 0.13422408625483512 + }, + { + "epoch": 0.5469886528949666, + "flow_matching_loss": 0.05721490792930126, + "grad_norm": 0.5934505462646484, + "learning_rate": 1.8789999999999998e-05, + "loss": 0.1909, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05143427550792694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.031160043366253377, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1340285949409008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975935518741608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1880, + "weighted_flow_matching_loss": 0.05721490792930126, + "weighted_rkd_loss": 0.13370605558156967 + }, + { + "epoch": 0.5498981670061099, + "flow_matching_loss": 0.05463487431406975, + "grad_norm": 0.6434962749481201, + "learning_rate": 1.889e-05, + "loss": 0.189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05195629298686981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.03079066276550293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13470325097441674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975678384304046, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1890, + "weighted_flow_matching_loss": 0.05463487431406975, + "weighted_rkd_loss": 0.13437563925981522 + }, + { + "epoch": 0.5528076811172534, + "flow_matching_loss": 0.055353787168860434, + "grad_norm": 0.588276743888855, + "learning_rate": 1.899e-05, + "loss": 0.1826, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.049138176068663594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029316872358322144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12759322449564933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9975419819355011, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1900, + "weighted_flow_matching_loss": 0.055353787168860434, + "weighted_rkd_loss": 0.127279569208622 + }, + { + "epoch": 0.5557171952283969, + "flow_matching_loss": 0.055916224420070645, + "grad_norm": 0.5400403141975403, + "learning_rate": 1.909e-05, + "loss": 0.1829, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04936334453523159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028585171699523924, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12731186002492906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.99751598238945, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1910, + "weighted_flow_matching_loss": 0.055916224420070645, + "weighted_rkd_loss": 0.1269955776631832 + }, + { + "epoch": 0.5586267093395403, + "flow_matching_loss": 0.05552534274756908, + "grad_norm": 0.5576940774917603, + "learning_rate": 1.919e-05, + "loss": 0.1842, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.049654354155063626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02967826370149851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12898697406053544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974898576736451, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1920, + "weighted_flow_matching_loss": 0.05552534274756908, + "weighted_rkd_loss": 0.12866323292255402 + }, + { + "epoch": 0.5615362234506838, + "flow_matching_loss": 0.05726088881492615, + "grad_norm": 0.6764211654663086, + "learning_rate": 1.929e-05, + "loss": 0.1875, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05021041370928288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.030131379142403603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1305522069334984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997463583946228, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1930, + "weighted_flow_matching_loss": 0.05726088881492615, + "weighted_rkd_loss": 0.13022103831171988 + }, + { + "epoch": 0.5644457375618271, + "flow_matching_loss": 0.05579238571226597, + "grad_norm": 0.6350774168968201, + "learning_rate": 1.939e-05, + "loss": 0.1824, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.049024952575564384, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028910464607179165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1269603691995144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974371731281281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1940, + "weighted_flow_matching_loss": 0.05579238571226597, + "weighted_rkd_loss": 0.12663500159978866 + }, + { + "epoch": 0.5673552516729706, + "flow_matching_loss": 0.05889507308602333, + "grad_norm": 0.5498497486114502, + "learning_rate": 1.9489999999999998e-05, + "loss": 0.1901, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05102414712309837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02945069074630737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13149898499250412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9974106311798095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1950, + "weighted_flow_matching_loss": 0.05889507308602333, + "weighted_rkd_loss": 0.1311585135757923 + }, + { + "epoch": 0.570264765784114, + "flow_matching_loss": 0.06319047585129738, + "grad_norm": 0.5621337890625, + "learning_rate": 1.959e-05, + "loss": 0.1859, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.047767091542482376, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027540363557636737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12307454571127892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973839581012726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1960, + "weighted_flow_matching_loss": 0.06319047585129738, + "weighted_rkd_loss": 0.12275255098938942 + }, + { + "epoch": 0.5731742798952575, + "flow_matching_loss": 0.058029620349407195, + "grad_norm": 0.513742208480835, + "learning_rate": 1.969e-05, + "loss": 0.1802, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04781434983015061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026825775392353535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1224544733762741, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973571419715881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1970, + "weighted_flow_matching_loss": 0.058029620349407195, + "weighted_rkd_loss": 0.12213084250688552 + }, + { + "epoch": 0.576083794006401, + "flow_matching_loss": 0.054222879186272624, + "grad_norm": 0.5922276377677917, + "learning_rate": 1.979e-05, + "loss": 0.1849, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.05094967633485794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.029152772203087807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.13105212524533272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973301887512207, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1980, + "weighted_flow_matching_loss": 0.054222879186272624, + "weighted_rkd_loss": 0.13070226684212685 + }, + { + "epoch": 0.5789933081175443, + "flow_matching_loss": 0.051737356930971146, + "grad_norm": 0.608072817325592, + "learning_rate": 1.989e-05, + "loss": 0.1761, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.048162514716386794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028345989622175695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12467101886868477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9973030984401703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 1990, + "weighted_flow_matching_loss": 0.051737356930971146, + "weighted_rkd_loss": 0.1243347778916359 + }, + { + "epoch": 0.5819028222286878, + "flow_matching_loss": 0.05285007320344448, + "grad_norm": 0.6203151345252991, + "learning_rate": 1.999e-05, + "loss": 0.1766, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04803949557244778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02796255759894848, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12404154837131501, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9972758829593659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2000, + "weighted_flow_matching_loss": 0.05285007320344448, + "weighted_rkd_loss": 0.12370362654328346 + }, + { + "epoch": 0.5848123363398312, + "flow_matching_loss": 0.054153700545430186, + "grad_norm": 0.5443220734596252, + "learning_rate": 2.009e-05, + "loss": 0.178, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.048370113223791124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027449827268719675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12419005408883095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9972485303878784, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2010, + "weighted_flow_matching_loss": 0.054153700545430186, + "weighted_rkd_loss": 0.1238483414053917 + }, + { + "epoch": 0.5877218504509747, + "flow_matching_loss": 0.05442479029297829, + "grad_norm": 0.5388229489326477, + "learning_rate": 2.019e-05, + "loss": 0.1783, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.048275042325258255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027636226452887058, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12418631240725517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997221028804779, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2020, + "weighted_flow_matching_loss": 0.05442479029297829, + "weighted_rkd_loss": 0.12384120970964432 + }, + { + "epoch": 0.5906313645621182, + "flow_matching_loss": 0.05339267551898956, + "grad_norm": 0.5593013167381287, + "learning_rate": 2.029e-05, + "loss": 0.1743, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04708329550921917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02710596937686205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12127256020903587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971933960914612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2030, + "weighted_flow_matching_loss": 0.05339267551898956, + "weighted_rkd_loss": 0.12093222513794899 + }, + { + "epoch": 0.5935408786732616, + "flow_matching_loss": 0.05637866742908955, + "grad_norm": 0.5842090249061584, + "learning_rate": 2.039e-05, + "loss": 0.1828, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04882003478705883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0291368680074811, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12677693665027617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971656262874603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2040, + "weighted_flow_matching_loss": 0.05637866742908955, + "weighted_rkd_loss": 0.12641756981611252 + }, + { + "epoch": 0.596450392784405, + "flow_matching_loss": 0.05538248121738434, + "grad_norm": 0.5600670576095581, + "learning_rate": 2.0490000000000002e-05, + "loss": 0.182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.049132202938199046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028711120039224623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12697552591562272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997137725353241, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2050, + "weighted_flow_matching_loss": 0.05538248121738434, + "weighted_rkd_loss": 0.12661205977201462 + }, + { + "epoch": 0.5993599068955484, + "flow_matching_loss": 0.052161000296473505, + "grad_norm": 0.49407991766929626, + "learning_rate": 2.059e-05, + "loss": 0.1779, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0489718671888113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028119000792503356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1260627366602421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9971096754074097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2060, + "weighted_flow_matching_loss": 0.052161000296473505, + "weighted_rkd_loss": 0.12569838166236877 + }, + { + "epoch": 0.6022694210066919, + "flow_matching_loss": 0.052539733424782756, + "grad_norm": 0.500371515750885, + "learning_rate": 2.069e-05, + "loss": 0.177, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04850462526082992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02784420847892761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1248534582555294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970815062522889, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2070, + "weighted_flow_matching_loss": 0.052539733424782756, + "weighted_rkd_loss": 0.12448909282684326 + }, + { + "epoch": 0.6051789351178353, + "flow_matching_loss": 0.05107822082936764, + "grad_norm": 0.4900391697883606, + "learning_rate": 2.079e-05, + "loss": 0.169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04626177698373794, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025736024789512157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11825958043336868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.997053188085556, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2080, + "weighted_flow_matching_loss": 0.05107822082936764, + "weighted_rkd_loss": 0.1179110899567604 + }, + { + "epoch": 0.6080884492289788, + "flow_matching_loss": 0.05440058596432209, + "grad_norm": 0.5605370402336121, + "learning_rate": 2.089e-05, + "loss": 0.1793, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0485592644661665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028162283450365068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12528081163764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9970247447490692, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2090, + "weighted_flow_matching_loss": 0.05440058596432209, + "weighted_rkd_loss": 0.12490806058049202 + }, + { + "epoch": 0.6109979633401222, + "flow_matching_loss": 0.05378255136311054, + "grad_norm": 0.5620014667510986, + "learning_rate": 2.099e-05, + "loss": 0.1768, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.047698841989040376, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.027970213629305362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1233678974211216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969961524009705, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2100, + "weighted_flow_matching_loss": 0.05378255136311054, + "weighted_rkd_loss": 0.12299735471606255 + }, + { + "epoch": 0.6139074774512656, + "flow_matching_loss": 0.053931647911667824, + "grad_norm": 0.6311306953430176, + "learning_rate": 2.109e-05, + "loss": 0.1779, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04780441075563431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.028726800717413424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12433562129735946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969674289226532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2110, + "weighted_flow_matching_loss": 0.053931647911667824, + "weighted_rkd_loss": 0.12395856380462647 + }, + { + "epoch": 0.6168169915624091, + "flow_matching_loss": 0.05394850410521031, + "grad_norm": 0.5054706335067749, + "learning_rate": 2.1190000000000002e-05, + "loss": 0.1743, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046975770592689516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02673444040119648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1206859827041626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969385743141175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2120, + "weighted_flow_matching_loss": 0.05394850410521031, + "weighted_rkd_loss": 0.12031650245189666 + }, + { + "epoch": 0.6197265056735525, + "flow_matching_loss": 0.051417041569948196, + "grad_norm": 0.5505133271217346, + "learning_rate": 2.129e-05, + "loss": 0.173, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04733505211770535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02725957278162241, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12192967608571052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9969095766544342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2130, + "weighted_flow_matching_loss": 0.051417041569948196, + "weighted_rkd_loss": 0.12155286148190499 + }, + { + "epoch": 0.622636019784696, + "flow_matching_loss": 0.05311718992888927, + "grad_norm": 0.5402168035507202, + "learning_rate": 2.139e-05, + "loss": 0.1736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04719863459467888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026470870524644852, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1208681397140026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996880441904068, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2140, + "weighted_flow_matching_loss": 0.05311718992888927, + "weighted_rkd_loss": 0.12049108445644378 + }, + { + "epoch": 0.6255455338958394, + "flow_matching_loss": 0.055306820571422576, + "grad_norm": 0.563251256942749, + "learning_rate": 2.1490000000000003e-05, + "loss": 0.1773, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04735132083296776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02771898601204157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12242162898182869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9968511760234833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2150, + "weighted_flow_matching_loss": 0.055306820571422576, + "weighted_rkd_loss": 0.12203614562749862 + }, + { + "epoch": 0.6284550480069828, + "flow_matching_loss": 0.055497553944587705, + "grad_norm": 0.5424150824546814, + "learning_rate": 2.159e-05, + "loss": 0.1692, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04454019814729691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024967667274177076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11404806226491929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9968217790126801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2160, + "weighted_flow_matching_loss": 0.055497553944587705, + "weighted_rkd_loss": 0.11368558630347252 + }, + { + "epoch": 0.6313645621181263, + "flow_matching_loss": 0.05638206601142883, + "grad_norm": 0.5782237648963928, + "learning_rate": 2.169e-05, + "loss": 0.1747, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04598008468747139, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026731061190366744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11869123131036759, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967922329902649, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2170, + "weighted_flow_matching_loss": 0.05638206601142883, + "weighted_rkd_loss": 0.11831049844622613 + }, + { + "epoch": 0.6342740762292697, + "flow_matching_loss": 0.055029870569705965, + "grad_norm": 0.5372181534767151, + "learning_rate": 2.1790000000000003e-05, + "loss": 0.1757, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04698068872094154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02714649885892868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12110787630081177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967625617980957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2180, + "weighted_flow_matching_loss": 0.055029870569705965, + "weighted_rkd_loss": 0.1207157626748085 + }, + { + "epoch": 0.6371835903404132, + "flow_matching_loss": 0.0515475895255804, + "grad_norm": 0.5861160755157471, + "learning_rate": 2.1890000000000002e-05, + "loss": 0.1685, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.045482371002435684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026344573125243187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11730931550264359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996732747554779, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2190, + "weighted_flow_matching_loss": 0.0515475895255804, + "weighted_rkd_loss": 0.11692603901028634 + }, + { + "epoch": 0.6400931044515565, + "flow_matching_loss": 0.0521878570318222, + "grad_norm": 0.5456554889678955, + "learning_rate": 2.199e-05, + "loss": 0.1686, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.045478053763508795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025881086476147175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11683719456195832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9967027962207794, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2200, + "weighted_flow_matching_loss": 0.0521878570318222, + "weighted_rkd_loss": 0.11645193994045258 + }, + { + "epoch": 0.6430026185627, + "flow_matching_loss": 0.05374699309468269, + "grad_norm": 0.6357887387275696, + "learning_rate": 2.209e-05, + "loss": 0.1738, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046811658516526224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026805162616074085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12042848020792007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966727137565613, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2210, + "weighted_flow_matching_loss": 0.05374699309468269, + "weighted_rkd_loss": 0.12002776116132736 + }, + { + "epoch": 0.6459121326738435, + "flow_matching_loss": 0.05321662425994873, + "grad_norm": 0.5458799600601196, + "learning_rate": 2.2190000000000003e-05, + "loss": 0.1692, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.045352769643068315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025686719082295896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11639225855469704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966424882411957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2220, + "weighted_flow_matching_loss": 0.05321662425994873, + "weighted_rkd_loss": 0.11600152030587196 + }, + { + "epoch": 0.6488216467849869, + "flow_matching_loss": 0.04898485243320465, + "grad_norm": 0.5108712315559387, + "learning_rate": 2.2290000000000002e-05, + "loss": 0.1667, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04572970718145371, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026608534343540668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11806794926524163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9966121435165405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2230, + "weighted_flow_matching_loss": 0.04898485243320465, + "weighted_rkd_loss": 0.11766794323921204 + }, + { + "epoch": 0.6517311608961304, + "flow_matching_loss": 0.05149602927267551, + "grad_norm": 0.5611505508422852, + "learning_rate": 2.239e-05, + "loss": 0.1722, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04718761518597603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02675973903387785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.12113496959209442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.996581643819809, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2240, + "weighted_flow_matching_loss": 0.05149602927267551, + "weighted_rkd_loss": 0.12072088867425919 + }, + { + "epoch": 0.6546406750072737, + "flow_matching_loss": 0.05378027521073818, + "grad_norm": 0.5492702126502991, + "learning_rate": 2.2490000000000003e-05, + "loss": 0.1706, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04595927521586418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025313912704586983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11723246276378632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9965510070323944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2250, + "weighted_flow_matching_loss": 0.05378027521073818, + "weighted_rkd_loss": 0.1168281614780426 + }, + { + "epoch": 0.6575501891184172, + "flow_matching_loss": 0.0544489499181509, + "grad_norm": 0.4795908033847809, + "learning_rate": 2.2590000000000002e-05, + "loss": 0.1724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.046101932600140574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026183218322694303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11838708445429802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9965202391147614, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2260, + "weighted_flow_matching_loss": 0.0544489499181509, + "weighted_rkd_loss": 0.11797515749931335 + }, + { + "epoch": 0.6604597032295607, + "flow_matching_loss": 0.05402884148061275, + "grad_norm": 0.5716680288314819, + "learning_rate": 2.269e-05, + "loss": 0.174, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0468517679721117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026737209223210812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1204407460987568, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964893460273743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2270, + "weighted_flow_matching_loss": 0.05402884148061275, + "weighted_rkd_loss": 0.12001789882779121 + }, + { + "epoch": 0.6633692173407041, + "flow_matching_loss": 0.05335509963333607, + "grad_norm": 0.5917899012565613, + "learning_rate": 2.279e-05, + "loss": 0.1704, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04605582132935524, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0253536369651556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11746527999639511, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964583098888398, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2280, + "weighted_flow_matching_loss": 0.05335509963333607, + "weighted_rkd_loss": 0.11704924926161767 + }, + { + "epoch": 0.6662787314518476, + "flow_matching_loss": 0.051287881284952166, + "grad_norm": 0.4779157042503357, + "learning_rate": 2.289e-05, + "loss": 0.1655, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0445807509124279, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025462519377470016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11462402120232582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9964271366596222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2290, + "weighted_flow_matching_loss": 0.051287881284952166, + "weighted_rkd_loss": 0.11421449705958367 + }, + { + "epoch": 0.669188245562991, + "flow_matching_loss": 0.05509746186435223, + "grad_norm": 0.5703138113021851, + "learning_rate": 2.299e-05, + "loss": 0.1734, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04602058231830597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026723263040184974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11876442804932594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963958263397217, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2300, + "weighted_flow_matching_loss": 0.05509746186435223, + "weighted_rkd_loss": 0.11833635494112968 + }, + { + "epoch": 0.6720977596741344, + "flow_matching_loss": 0.054733379557728765, + "grad_norm": 0.5187135338783264, + "learning_rate": 2.309e-05, + "loss": 0.1708, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0452336635440588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02606401015073061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11653133556246757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963643729686738, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2310, + "weighted_flow_matching_loss": 0.054733379557728765, + "weighted_rkd_loss": 0.11610773950815201 + }, + { + "epoch": 0.6750072737852778, + "flow_matching_loss": 0.049514494463801385, + "grad_norm": 0.5620117783546448, + "learning_rate": 2.319e-05, + "loss": 0.1662, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0453199926763773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026511294580996037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11715127974748611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963327884674072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2320, + "weighted_flow_matching_loss": 0.049514494463801385, + "weighted_rkd_loss": 0.11672168299555778 + }, + { + "epoch": 0.6779167878964213, + "flow_matching_loss": 0.054025353863835335, + "grad_norm": 0.5476577877998352, + "learning_rate": 2.329e-05, + "loss": 0.1712, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04563356563448906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026345320977270603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11761245056986809, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9963010728359223, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2330, + "weighted_flow_matching_loss": 0.054025353863835335, + "weighted_rkd_loss": 0.11717741265892982 + }, + { + "epoch": 0.6808263020075648, + "flow_matching_loss": 0.05141115300357342, + "grad_norm": 0.46907171607017517, + "learning_rate": 2.3389999999999998e-05, + "loss": 0.1629, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04353310875594616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02484888583421707, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11191510334610939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962692201137543, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2340, + "weighted_flow_matching_loss": 0.05141115300357342, + "weighted_rkd_loss": 0.11149755492806435 + }, + { + "epoch": 0.6837358161187082, + "flow_matching_loss": 0.05575459599494934, + "grad_norm": 0.4986708164215088, + "learning_rate": 2.349e-05, + "loss": 0.1742, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04618692062795162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026503456756472588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11887729912996292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962372303009033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2350, + "weighted_flow_matching_loss": 0.05575459599494934, + "weighted_rkd_loss": 0.1184299848973751 + }, + { + "epoch": 0.6866453302298516, + "flow_matching_loss": 0.053281518444418904, + "grad_norm": 0.5370136499404907, + "learning_rate": 2.359e-05, + "loss": 0.1703, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04569406546652317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.026059943065047265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11744807288050652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9962050914764404, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2360, + "weighted_flow_matching_loss": 0.053281518444418904, + "weighted_rkd_loss": 0.11700233742594719 + }, + { + "epoch": 0.689554844340995, + "flow_matching_loss": 0.05346147269010544, + "grad_norm": 0.5890445113182068, + "learning_rate": 2.369e-05, + "loss": 0.1671, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04432973489165306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02542418222874403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11408365070819855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961728453636169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2370, + "weighted_flow_matching_loss": 0.05346147269010544, + "weighted_rkd_loss": 0.11364702880382538 + }, + { + "epoch": 0.6924643584521385, + "flow_matching_loss": 0.05415492951869964, + "grad_norm": 0.5303980112075806, + "learning_rate": 2.379e-05, + "loss": 0.1671, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.044439565390348434, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024546646513044834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11342577785253524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961404442787171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2380, + "weighted_flow_matching_loss": 0.05415492951869964, + "weighted_rkd_loss": 0.11298803240060806 + }, + { + "epoch": 0.695373872563282, + "flow_matching_loss": 0.051307781413197515, + "grad_norm": 0.5838856101036072, + "learning_rate": 2.389e-05, + "loss": 0.1637, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04406416565179825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024677804484963418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11280613392591476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9961079120635986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2390, + "weighted_flow_matching_loss": 0.051307781413197515, + "weighted_rkd_loss": 0.11236703842878341 + }, + { + "epoch": 0.6982833866744254, + "flow_matching_loss": 0.05088139176368713, + "grad_norm": 0.49891626834869385, + "learning_rate": 2.399e-05, + "loss": 0.1638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04417569749057293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0250244228169322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11337581798434257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960752367973328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2400, + "weighted_flow_matching_loss": 0.05088139176368713, + "weighted_rkd_loss": 0.11293085813522338 + }, + { + "epoch": 0.7011929007855688, + "flow_matching_loss": 0.05334013104438782, + "grad_norm": 0.5421711206436157, + "learning_rate": 2.409e-05, + "loss": 0.1645, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04366948865354061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02423339430242777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11157237067818641, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960424363613128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2410, + "weighted_flow_matching_loss": 0.05334013104438782, + "weighted_rkd_loss": 0.11113079488277436 + }, + { + "epoch": 0.7041024148967122, + "flow_matching_loss": 0.05199198499321937, + "grad_norm": 0.5416830778121948, + "learning_rate": 2.419e-05, + "loss": 0.1644, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04390571042895317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025095304660499095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11290672570466995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9960094869136811, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2420, + "weighted_flow_matching_loss": 0.05199198499321937, + "weighted_rkd_loss": 0.11245618015527725 + }, + { + "epoch": 0.7070119290078557, + "flow_matching_loss": 0.054762056842446324, + "grad_norm": 0.47949132323265076, + "learning_rate": 2.429e-05, + "loss": 0.1623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04245818741619587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023007651418447496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10792402550578117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959764122962952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2430, + "weighted_flow_matching_loss": 0.054762056842446324, + "weighted_rkd_loss": 0.10748978704214096 + }, + { + "epoch": 0.7099214431189991, + "flow_matching_loss": 0.05295904874801636, + "grad_norm": 0.5494281649589539, + "learning_rate": 2.439e-05, + "loss": 0.1653, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04397435635328293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024860300309956073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11280901208519936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959432005882263, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2440, + "weighted_flow_matching_loss": 0.05295904874801636, + "weighted_rkd_loss": 0.11235137283802032 + }, + { + "epoch": 0.7128309572301426, + "flow_matching_loss": 0.05059077776968479, + "grad_norm": 0.5533782243728638, + "learning_rate": 2.449e-05, + "loss": 0.1639, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04452265240252018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02473781891167164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1137831225991249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9959098517894744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2450, + "weighted_flow_matching_loss": 0.05059077776968479, + "weighted_rkd_loss": 0.11331772282719613 + }, + { + "epoch": 0.715740471341286, + "flow_matching_loss": 0.055373409762978554, + "grad_norm": 0.5230960845947266, + "learning_rate": 2.459e-05, + "loss": 0.1691, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.044484008476138114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025275311432778835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11424332857131958, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958763599395752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2460, + "weighted_flow_matching_loss": 0.055373409762978554, + "weighted_rkd_loss": 0.11377221792936325 + }, + { + "epoch": 0.7186499854524294, + "flow_matching_loss": 0.051661786437034604, + "grad_norm": 0.6260476112365723, + "learning_rate": 2.469e-05, + "loss": 0.164, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04404332153499126, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02467823475599289, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11276487782597541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958427429199219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2470, + "weighted_flow_matching_loss": 0.051661786437034604, + "weighted_rkd_loss": 0.1122960940003395 + }, + { + "epoch": 0.7215594995635729, + "flow_matching_loss": 0.05253085605800152, + "grad_norm": 0.5415815114974976, + "learning_rate": 2.479e-05, + "loss": 0.1617, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.042876213788986206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02388490382581949, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10963733196258545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9958089888095856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2480, + "weighted_flow_matching_loss": 0.05253085605800152, + "weighted_rkd_loss": 0.10917786359786988 + }, + { + "epoch": 0.7244690136747163, + "flow_matching_loss": 0.049555976316332816, + "grad_norm": 0.5590134859085083, + "learning_rate": 2.489e-05, + "loss": 0.1582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04265338368713856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023813246563076974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10912001430988312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957750737667084, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2490, + "weighted_flow_matching_loss": 0.049555976316332816, + "weighted_rkd_loss": 0.10865896567702293 + }, + { + "epoch": 0.7273785277858598, + "flow_matching_loss": 0.04960497803986073, + "grad_norm": 0.5180374383926392, + "learning_rate": 2.499e-05, + "loss": 0.1563, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0418891329318285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02334959525614977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10712786093354225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957410573959351, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2500, + "weighted_flow_matching_loss": 0.04960497803986073, + "weighted_rkd_loss": 0.10667162165045738 + }, + { + "epoch": 0.7302880418970032, + "flow_matching_loss": 0.05205758661031723, + "grad_norm": 0.573921263217926, + "learning_rate": 2.5090000000000002e-05, + "loss": 0.1584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04153285399079323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023699966631829737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10676567479968072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9957068979740142, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2510, + "weighted_flow_matching_loss": 0.05205758661031723, + "weighted_rkd_loss": 0.10630731955170632 + }, + { + "epoch": 0.7331975560081466, + "flow_matching_loss": 0.050658106058835986, + "grad_norm": 0.5549129247665405, + "learning_rate": 2.519e-05, + "loss": 0.1625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04356885217130184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025228959694504737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1123666636645794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956725835800171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2520, + "weighted_flow_matching_loss": 0.050658106058835986, + "weighted_rkd_loss": 0.11188037544488907 + }, + { + "epoch": 0.7361070701192901, + "flow_matching_loss": 0.048825059086084366, + "grad_norm": 0.5698189735412598, + "learning_rate": 2.529e-05, + "loss": 0.1602, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043567783758044244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024720311723649503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11185588017106056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956381559371948, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2530, + "weighted_flow_matching_loss": 0.048825059086084366, + "weighted_rkd_loss": 0.11136800572276115 + }, + { + "epoch": 0.7390165842304335, + "flow_matching_loss": 0.052494828775525096, + "grad_norm": 0.5588210821151733, + "learning_rate": 2.5390000000000003e-05, + "loss": 0.1589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04153067171573639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0237941175699234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10685546025633812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9956035852432251, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2540, + "weighted_flow_matching_loss": 0.052494828775525096, + "weighted_rkd_loss": 0.10638565868139267 + }, + { + "epoch": 0.741926098341577, + "flow_matching_loss": 0.04999199472367764, + "grad_norm": 0.4756923317909241, + "learning_rate": 2.549e-05, + "loss": 0.157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04184359535574913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023768777213990688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10745596736669541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9955688655376435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2550, + "weighted_flow_matching_loss": 0.04999199472367764, + "weighted_rkd_loss": 0.10697983801364899 + }, + { + "epoch": 0.7448356124527203, + "flow_matching_loss": 0.05147750154137611, + "grad_norm": 0.6412276029586792, + "learning_rate": 2.559e-05, + "loss": 0.1645, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043799896538257596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.025890850462019444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11349064409732819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9955340266227722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2560, + "weighted_flow_matching_loss": 0.05147750154137611, + "weighted_rkd_loss": 0.11298379451036453 + }, + { + "epoch": 0.7477451265638638, + "flow_matching_loss": 0.05100189931690693, + "grad_norm": 0.5729312896728516, + "learning_rate": 2.569e-05, + "loss": 0.1624, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04362533427774906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024696500785648823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11194717139005661, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954990386962891, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2570, + "weighted_flow_matching_loss": 0.05100189931690693, + "weighted_rkd_loss": 0.11144326776266097 + }, + { + "epoch": 0.7506546406750073, + "flow_matching_loss": 0.05034074112772942, + "grad_norm": 0.5441398620605469, + "learning_rate": 2.5790000000000002e-05, + "loss": 0.1555, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0412424273788929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023164234682917596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10564908906817436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954639375209808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2580, + "weighted_flow_matching_loss": 0.05034074112772942, + "weighted_rkd_loss": 0.10516986027359962 + }, + { + "epoch": 0.7535641547861507, + "flow_matching_loss": 0.05011311024427414, + "grad_norm": 0.5083782076835632, + "learning_rate": 2.589e-05, + "loss": 0.1554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041397878900170326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02301386520266533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10580962300300598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9954286694526673, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2590, + "weighted_flow_matching_loss": 0.05011311024427414, + "weighted_rkd_loss": 0.1053258940577507 + }, + { + "epoch": 0.7564736688972942, + "flow_matching_loss": 0.05242548398673534, + "grad_norm": 0.5428669452667236, + "learning_rate": 2.599e-05, + "loss": 0.1577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04117696098983288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02338318172842264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10573710352182389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9953932762145996, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2600, + "weighted_flow_matching_loss": 0.05242548398673534, + "weighted_rkd_loss": 0.10525000393390656 + }, + { + "epoch": 0.7593831830084375, + "flow_matching_loss": 0.04724690392613411, + "grad_norm": 0.5787995457649231, + "learning_rate": 2.6090000000000003e-05, + "loss": 0.152, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04106377214193344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02314300648868084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10527055040001869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9953577578067779, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2610, + "weighted_flow_matching_loss": 0.04724690392613411, + "weighted_rkd_loss": 0.1047818548977375 + }, + { + "epoch": 0.762292697119581, + "flow_matching_loss": 0.051124942302703855, + "grad_norm": 0.49515506625175476, + "learning_rate": 2.619e-05, + "loss": 0.1571, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041340882703661916, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023819271847605704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10650103688240051, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9953221023082733, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2620, + "weighted_flow_matching_loss": 0.051124942302703855, + "weighted_rkd_loss": 0.10600287541747093 + }, + { + "epoch": 0.7652022112307245, + "flow_matching_loss": 0.05053756088018417, + "grad_norm": 0.5329370498657227, + "learning_rate": 2.629e-05, + "loss": 0.1537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04048781543970108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022715229913592337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10369086191058159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9952863037586213, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2630, + "weighted_flow_matching_loss": 0.05053756088018417, + "weighted_rkd_loss": 0.10320208594202995 + }, + { + "epoch": 0.7681117253418679, + "flow_matching_loss": 0.04830462671816349, + "grad_norm": 0.5310961008071899, + "learning_rate": 2.6390000000000003e-05, + "loss": 0.1527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.040819539502263066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023260057903826237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10489913821220398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9952503621578217, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2640, + "weighted_flow_matching_loss": 0.04830462671816349, + "weighted_rkd_loss": 0.10440090000629425 + }, + { + "epoch": 0.7710212394530114, + "flow_matching_loss": 0.04921962283551693, + "grad_norm": 0.5405465960502625, + "learning_rate": 2.6490000000000002e-05, + "loss": 0.1521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04010241031646729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02312566079199314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10333048179745674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.995214295387268, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2650, + "weighted_flow_matching_loss": 0.04921962283551693, + "weighted_rkd_loss": 0.1028360053896904 + }, + { + "epoch": 0.7739307535641547, + "flow_matching_loss": 0.050336581841111186, + "grad_norm": 0.4889579117298126, + "learning_rate": 2.659e-05, + "loss": 0.1541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04097251445055008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02231267187744379, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10425770059227943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951781034469604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2660, + "weighted_flow_matching_loss": 0.050336581841111186, + "weighted_rkd_loss": 0.10375497862696648 + }, + { + "epoch": 0.7768402676752982, + "flow_matching_loss": 0.0500234629958868, + "grad_norm": 0.5247243046760559, + "learning_rate": 2.6690000000000004e-05, + "loss": 0.1531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04062359966337681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022366767376661302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10361396819353104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951417624950409, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2670, + "weighted_flow_matching_loss": 0.0500234629958868, + "weighted_rkd_loss": 0.10311058089137078 + }, + { + "epoch": 0.7797497817864416, + "flow_matching_loss": 0.05122898370027542, + "grad_norm": 0.5294696688652039, + "learning_rate": 2.6790000000000003e-05, + "loss": 0.1594, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04251380562782288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023713782243430615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10874139368534089, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9951052844524384, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2680, + "weighted_flow_matching_loss": 0.05122898370027542, + "weighted_rkd_loss": 0.10820918679237365 + }, + { + "epoch": 0.7826592958975851, + "flow_matching_loss": 0.04825787432491779, + "grad_norm": 0.4704005718231201, + "learning_rate": 2.6890000000000002e-05, + "loss": 0.1555, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041912109777331355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023997674882411956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10782189518213273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9950686693191528, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2690, + "weighted_flow_matching_loss": 0.04825787432491779, + "weighted_rkd_loss": 0.1072902038693428 + }, + { + "epoch": 0.7855688100087286, + "flow_matching_loss": 0.049618583917617795, + "grad_norm": 0.5227334499359131, + "learning_rate": 2.699e-05, + "loss": 0.1494, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03922275155782699, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02182105742394924, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10026655867695808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9950319230556488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2700, + "weighted_flow_matching_loss": 0.049618583917617795, + "weighted_rkd_loss": 0.09976842179894448 + }, + { + "epoch": 0.788478324119872, + "flow_matching_loss": 0.04810768254101276, + "grad_norm": 0.48230230808258057, + "learning_rate": 2.709e-05, + "loss": 0.1592, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.043391787260770795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.024898509494960308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.11168208420276642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9949950575828552, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2710, + "weighted_flow_matching_loss": 0.04810768254101276, + "weighted_rkd_loss": 0.11112310960888863 + }, + { + "epoch": 0.7913878382310154, + "flow_matching_loss": 0.05083151645958424, + "grad_norm": 0.5349370837211609, + "learning_rate": 2.719e-05, + "loss": 0.1552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041052252426743506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02282407581806183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10492857992649078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9949580252170562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2720, + "weighted_flow_matching_loss": 0.05083151645958424, + "weighted_rkd_loss": 0.10439951941370965 + }, + { + "epoch": 0.7942973523421588, + "flow_matching_loss": 0.0478250190615654, + "grad_norm": 0.4620766043663025, + "learning_rate": 2.7289999999999998e-05, + "loss": 0.1545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0416364636272192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023992829583585264, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10726575702428817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9949208736419678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2730, + "weighted_flow_matching_loss": 0.0478250190615654, + "weighted_rkd_loss": 0.10672098249197007 + }, + { + "epoch": 0.7972068664533023, + "flow_matching_loss": 0.05177161954343319, + "grad_norm": 0.5323125720024109, + "learning_rate": 2.739e-05, + "loss": 0.1574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.041388196498155595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023368835262954236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10614522770047188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948835849761963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2740, + "weighted_flow_matching_loss": 0.05177161954343319, + "weighted_rkd_loss": 0.10560211911797523 + }, + { + "epoch": 0.8001163805644458, + "flow_matching_loss": 0.050645352154970166, + "grad_norm": 0.4672781527042389, + "learning_rate": 2.749e-05, + "loss": 0.1491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03851804211735725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021968541853129863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09900462552905083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948461651802063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2750, + "weighted_flow_matching_loss": 0.050645352154970166, + "weighted_rkd_loss": 0.09849439188838005 + }, + { + "epoch": 0.8030258946755892, + "flow_matching_loss": 0.051600281894207, + "grad_norm": 0.5621912479400635, + "learning_rate": 2.759e-05, + "loss": 0.1548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.040663906186819074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02245228737592697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10378010049462319, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9948085963726043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2760, + "weighted_flow_matching_loss": 0.051600281894207, + "weighted_rkd_loss": 0.10324133858084679 + }, + { + "epoch": 0.8059354087867326, + "flow_matching_loss": 0.050525176152586934, + "grad_norm": 0.4575623571872711, + "learning_rate": 2.769e-05, + "loss": 0.1512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03970180749893189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02176643759012222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10117005184292793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9947708964347839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2770, + "weighted_flow_matching_loss": 0.050525176152586934, + "weighted_rkd_loss": 0.10064108222723007 + }, + { + "epoch": 0.808844922897876, + "flow_matching_loss": 0.05117805078625679, + "grad_norm": 0.548370361328125, + "learning_rate": 2.779e-05, + "loss": 0.1502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03876141756772995, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022018945775926113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09954178109765052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9947330713272095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2780, + "weighted_flow_matching_loss": 0.05117805078625679, + "weighted_rkd_loss": 0.09901749119162559 + }, + { + "epoch": 0.8117544370090195, + "flow_matching_loss": 0.04776485748589039, + "grad_norm": 0.5168671607971191, + "learning_rate": 2.789e-05, + "loss": 0.1509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04045403525233269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02277027703821659, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10367834866046906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946951031684875, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2790, + "weighted_flow_matching_loss": 0.04776485748589039, + "weighted_rkd_loss": 0.10312828570604324 + }, + { + "epoch": 0.814663951120163, + "flow_matching_loss": 0.049614714458584785, + "grad_norm": 0.5143334269523621, + "learning_rate": 2.799e-05, + "loss": 0.1559, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04149929620325565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023867253586649893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10686584487557411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946569800376892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2800, + "weighted_flow_matching_loss": 0.049614714458584785, + "weighted_rkd_loss": 0.1062948651611805 + }, + { + "epoch": 0.8175734652313064, + "flow_matching_loss": 0.04650114625692368, + "grad_norm": 0.4943876266479492, + "learning_rate": 2.809e-05, + "loss": 0.1446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03862036727368832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021409447118639947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09865018203854561, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9946187555789947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2810, + "weighted_flow_matching_loss": 0.04650114625692368, + "weighted_rkd_loss": 0.09811936616897583 + }, + { + "epoch": 0.8204829793424498, + "flow_matching_loss": 0.047924245893955233, + "grad_norm": 0.4329610764980316, + "learning_rate": 2.819e-05, + "loss": 0.1515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04076582193374634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022565412148833276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10409705713391304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9945803880691528, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2820, + "weighted_flow_matching_loss": 0.047924245893955233, + "weighted_rkd_loss": 0.10353291556239128 + }, + { + "epoch": 0.8233924934535932, + "flow_matching_loss": 0.05062179006636143, + "grad_norm": 0.4993090033531189, + "learning_rate": 2.829e-05, + "loss": 0.1514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03958411179482937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02211454901844263, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10128277316689491, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994541871547699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2830, + "weighted_flow_matching_loss": 0.05062179006636143, + "weighted_rkd_loss": 0.10072996690869332 + }, + { + "epoch": 0.8263020075647367, + "flow_matching_loss": 0.0495597705245018, + "grad_norm": 0.4905972480773926, + "learning_rate": 2.839e-05, + "loss": 0.1517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04038701765239239, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021927704103291035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10270173847675323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9945032298564911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2840, + "weighted_flow_matching_loss": 0.0495597705245018, + "weighted_rkd_loss": 0.10213727876543999 + }, + { + "epoch": 0.8292115216758801, + "flow_matching_loss": 0.050468889996409416, + "grad_norm": 0.4633789360523224, + "learning_rate": 2.849e-05, + "loss": 0.1488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03882929757237434, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021192493848502636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09885108917951584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9944644510746002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2850, + "weighted_flow_matching_loss": 0.050468889996409416, + "weighted_rkd_loss": 0.09830392152070999 + }, + { + "epoch": 0.8321210357870236, + "flow_matching_loss": 0.04902223981916905, + "grad_norm": 0.5236678719520569, + "learning_rate": 2.859e-05, + "loss": 0.1556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04171886965632439, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023733679950237275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.107171418517828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9944255232810975, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2860, + "weighted_flow_matching_loss": 0.04902223981916905, + "weighted_rkd_loss": 0.10657404735684395 + }, + { + "epoch": 0.835030549898167, + "flow_matching_loss": 0.051423327252268794, + "grad_norm": 0.4819902777671814, + "learning_rate": 2.869e-05, + "loss": 0.1535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03985278457403183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02291281186044216, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1026183806359768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994386476278305, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2870, + "weighted_flow_matching_loss": 0.051423327252268794, + "weighted_rkd_loss": 0.10204233080148697 + }, + { + "epoch": 0.8379400640093104, + "flow_matching_loss": 0.046987352520227434, + "grad_norm": 0.4745895564556122, + "learning_rate": 2.879e-05, + "loss": 0.1439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03823507502675057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020991276763379575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09746142625808715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9943472921848298, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2880, + "weighted_flow_matching_loss": 0.046987352520227434, + "weighted_rkd_loss": 0.09691053107380868 + }, + { + "epoch": 0.8408495781204539, + "flow_matching_loss": 0.04760455563664436, + "grad_norm": 0.47848474979400635, + "learning_rate": 2.889e-05, + "loss": 0.1435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03803386427462101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02038103938102722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0964487686753273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9943079710006714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2890, + "weighted_flow_matching_loss": 0.04760455563664436, + "weighted_rkd_loss": 0.09589979201555252 + }, + { + "epoch": 0.8437590922315973, + "flow_matching_loss": 0.047036370262503624, + "grad_norm": 0.5035362839698792, + "learning_rate": 2.8990000000000002e-05, + "loss": 0.1453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03852329291403293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02176240924745798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0988089956343174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9942685067653656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2900, + "weighted_flow_matching_loss": 0.047036370262503624, + "weighted_rkd_loss": 0.09824260547757149 + }, + { + "epoch": 0.8466686063427408, + "flow_matching_loss": 0.04631877355277538, + "grad_norm": 0.5864942669868469, + "learning_rate": 2.909e-05, + "loss": 0.1524, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04181923195719719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023061145283281802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10669961050152779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9942289173603058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2910, + "weighted_flow_matching_loss": 0.04631877355277538, + "weighted_rkd_loss": 0.10608382001519204 + }, + { + "epoch": 0.8495781204538843, + "flow_matching_loss": 0.04867085106670856, + "grad_norm": 0.6138477921485901, + "learning_rate": 2.919e-05, + "loss": 0.1477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03903111070394516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021521013975143433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09958323463797569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9941891849040985, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2920, + "weighted_flow_matching_loss": 0.04867085106670856, + "weighted_rkd_loss": 0.09900455102324486 + }, + { + "epoch": 0.8524876345650276, + "flow_matching_loss": 0.049237731099128726, + "grad_norm": 0.5372241735458374, + "learning_rate": 2.9290000000000002e-05, + "loss": 0.1502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.039525607600808144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02251260690391064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.1015638217329979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9941492974758148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2930, + "weighted_flow_matching_loss": 0.049237731099128726, + "weighted_rkd_loss": 0.10096962228417397 + }, + { + "epoch": 0.8553971486761711, + "flow_matching_loss": 0.05028913989663124, + "grad_norm": 0.5150318145751953, + "learning_rate": 2.939e-05, + "loss": 0.1536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.04041461683809757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.023069774731993675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10389900878071785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9941093146800994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2940, + "weighted_flow_matching_loss": 0.05028913989663124, + "weighted_rkd_loss": 0.10328696966171265 + }, + { + "epoch": 0.8583066627873145, + "flow_matching_loss": 0.04753070212900638, + "grad_norm": 0.4849422574043274, + "learning_rate": 2.949e-05, + "loss": 0.143, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03773023448884487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020591726526618004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09605219587683678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9940691828727722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2950, + "weighted_flow_matching_loss": 0.04753070212900638, + "weighted_rkd_loss": 0.09548252746462822 + }, + { + "epoch": 0.861216176898458, + "flow_matching_loss": 0.04755702167749405, + "grad_norm": 0.5490691661834717, + "learning_rate": 2.959e-05, + "loss": 0.1457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03876064196228981, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021256034821271898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09877731949090958, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.994028913974762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2960, + "weighted_flow_matching_loss": 0.04755702167749405, + "weighted_rkd_loss": 0.09818748533725738 + }, + { + "epoch": 0.8641256910096013, + "flow_matching_loss": 0.048003634065389635, + "grad_norm": 0.4515547752380371, + "learning_rate": 2.9690000000000002e-05, + "loss": 0.1453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.038513320684432986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020824857242405414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09785149842500687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9939885020256043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2970, + "weighted_flow_matching_loss": 0.048003634065389635, + "weighted_rkd_loss": 0.0972633071243763 + }, + { + "epoch": 0.8670352051207448, + "flow_matching_loss": 0.048235032334923746, + "grad_norm": 0.5774533748626709, + "learning_rate": 2.979e-05, + "loss": 0.1476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03929909244179726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021339144557714462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09993733018636704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993947958946228, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2980, + "weighted_flow_matching_loss": 0.048235032334923746, + "weighted_rkd_loss": 0.09933245107531548 + }, + { + "epoch": 0.8699447192318883, + "flow_matching_loss": 0.04642029106616974, + "grad_norm": 0.4307015836238861, + "learning_rate": 2.989e-05, + "loss": 0.1387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03619678840041161, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020404907409101725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09279848337173462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9939072787761688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 2990, + "weighted_flow_matching_loss": 0.04642029106616974, + "weighted_rkd_loss": 0.0922330990433693 + }, + { + "epoch": 0.8728542333430317, + "flow_matching_loss": 0.04821936748921871, + "grad_norm": 0.46883624792099, + "learning_rate": 2.9990000000000003e-05, + "loss": 0.1424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.037429331243038176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019925561361014844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09478422477841378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9938664674758911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3000, + "weighted_flow_matching_loss": 0.04821936748921871, + "weighted_rkd_loss": 0.09420285969972611 + }, + { + "epoch": 0.8757637474541752, + "flow_matching_loss": 0.05216869488358498, + "grad_norm": 0.4538305401802063, + "learning_rate": 2.9999998154575393e-05, + "loss": 0.1468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.037168441154062745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02092776857316494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09526465088129044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993825513124466, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3010, + "weighted_flow_matching_loss": 0.05216869488358498, + "weighted_rkd_loss": 0.09467647075653077 + }, + { + "epoch": 0.8786732615653186, + "flow_matching_loss": 0.04644184149801731, + "grad_norm": 0.5230268239974976, + "learning_rate": 2.999999177533042e-05, + "loss": 0.1448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03872239217162132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02149899210780859, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0989437758922577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9937844395637512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3020, + "weighted_flow_matching_loss": 0.04644184149801731, + "weighted_rkd_loss": 0.09832876622676849 + }, + { + "epoch": 0.881582775676462, + "flow_matching_loss": 0.04906134344637394, + "grad_norm": 0.5152180194854736, + "learning_rate": 2.9999980839483992e-05, + "loss": 0.1429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03706381618976593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020312903076410295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09444053545594215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9937432110309601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3030, + "weighted_flow_matching_loss": 0.04906134344637394, + "weighted_rkd_loss": 0.09384963810443878 + }, + { + "epoch": 0.8844922897876055, + "flow_matching_loss": 0.047951854020357135, + "grad_norm": 0.4524039030075073, + "learning_rate": 2.999996534703944e-05, + "loss": 0.1471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03892589658498764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021912765130400656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09976455792784691, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9937018573284149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3040, + "weighted_flow_matching_loss": 0.047951854020357135, + "weighted_rkd_loss": 0.09913622215390205 + }, + { + "epoch": 0.8874018038987489, + "flow_matching_loss": 0.04450833983719349, + "grad_norm": 0.5350182056427002, + "learning_rate": 2.9999945298001468e-05, + "loss": 0.1407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03779990151524544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021198047697544097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09679785072803497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9936603724956512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3050, + "weighted_flow_matching_loss": 0.04450833983719349, + "weighted_rkd_loss": 0.09618414491415024 + }, + { + "epoch": 0.8903113180098924, + "flow_matching_loss": 0.04646945968270302, + "grad_norm": 0.4527464509010315, + "learning_rate": 2.9999920692376165e-05, + "loss": 0.139, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03662700913846493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01991404891014099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09316806718707085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9936187267303467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3060, + "weighted_flow_matching_loss": 0.04646945968270302, + "weighted_rkd_loss": 0.09257355779409408 + }, + { + "epoch": 0.8932208321210358, + "flow_matching_loss": 0.04644052945077419, + "grad_norm": 0.6429787874221802, + "learning_rate": 2.9999891530171002e-05, + "loss": 0.1389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03672195561230183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01958857011049986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09303248152136803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993576979637146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3070, + "weighted_flow_matching_loss": 0.04644052945077419, + "weighted_rkd_loss": 0.09243491590023041 + }, + { + "epoch": 0.8961303462321792, + "flow_matching_loss": 0.047708649188280106, + "grad_norm": 0.5666985511779785, + "learning_rate": 2.9999857811394845e-05, + "loss": 0.1495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0400277990847826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.022352588176727296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.10240818560123444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9935350835323333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3080, + "weighted_flow_matching_loss": 0.047708649188280106, + "weighted_rkd_loss": 0.10174608156085015 + }, + { + "epoch": 0.8990398603433226, + "flow_matching_loss": 0.04652015827596188, + "grad_norm": 0.48497459292411804, + "learning_rate": 2.9999819536057935e-05, + "loss": 0.1419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.037771689146757124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020479545928537846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09602292403578758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9934930562973022, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3090, + "weighted_flow_matching_loss": 0.04652015827596188, + "weighted_rkd_loss": 0.0953981526196003 + }, + { + "epoch": 0.9019493744544661, + "flow_matching_loss": 0.04469442516565323, + "grad_norm": 0.49144110083580017, + "learning_rate": 2.9999776704171894e-05, + "loss": 0.1373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03647078089416027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020229834504425524, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09317139461636544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9934508860111236, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3100, + "weighted_flow_matching_loss": 0.04469442516565323, + "weighted_rkd_loss": 0.09256125465035439 + }, + { + "epoch": 0.9048588885656096, + "flow_matching_loss": 0.047894684970378874, + "grad_norm": 0.4889221489429474, + "learning_rate": 2.999972931574973e-05, + "loss": 0.145, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0381736159324646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.021365000493824483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09771223366260529, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9934085786342621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3110, + "weighted_flow_matching_loss": 0.047894684970378874, + "weighted_rkd_loss": 0.09706819579005241 + }, + { + "epoch": 0.907768402676753, + "flow_matching_loss": 0.045112861320376396, + "grad_norm": 0.4787847101688385, + "learning_rate": 2.9999677370805858e-05, + "loss": 0.138, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03675225377082825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019992126431316138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0934966340661049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993366140127182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3120, + "weighted_flow_matching_loss": 0.045112861320376396, + "weighted_rkd_loss": 0.0928764134645462 + }, + { + "epoch": 0.9106779167878964, + "flow_matching_loss": 0.0465823695063591, + "grad_norm": 0.4537099599838257, + "learning_rate": 2.9999620869356034e-05, + "loss": 0.1412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03732055127620697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02061483506113291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09525593817234039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.993323564529419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3130, + "weighted_flow_matching_loss": 0.0465823695063591, + "weighted_rkd_loss": 0.09461994767189026 + }, + { + "epoch": 0.9135874308990398, + "flow_matching_loss": 0.04697315767407417, + "grad_norm": 0.42567557096481323, + "learning_rate": 2.9999559811417436e-05, + "loss": 0.1368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03578305430710316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018872874043881892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09043898358941078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9932808578014374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3140, + "weighted_flow_matching_loss": 0.04697315767407417, + "weighted_rkd_loss": 0.08983130678534508 + }, + { + "epoch": 0.9164969450101833, + "flow_matching_loss": 0.047309132665395735, + "grad_norm": 0.46436846256256104, + "learning_rate": 2.999949419700861e-05, + "loss": 0.1393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03635299578309059, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019886588491499425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09259257987141609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9932380318641663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3150, + "weighted_flow_matching_loss": 0.047309132665395735, + "weighted_rkd_loss": 0.09196645841002464 + }, + { + "epoch": 0.9194064591213268, + "flow_matching_loss": 0.04680666588246822, + "grad_norm": 0.4517979323863983, + "learning_rate": 2.9999424026149484e-05, + "loss": 0.1387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03677400313317776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019009533710777758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09255753979086875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9931950390338897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3160, + "weighted_flow_matching_loss": 0.04680666588246822, + "weighted_rkd_loss": 0.09192775264382362 + }, + { + "epoch": 0.9223159732324702, + "flow_matching_loss": 0.04979852102696895, + "grad_norm": 0.584871232509613, + "learning_rate": 2.9999349298861376e-05, + "loss": 0.1469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03854408636689186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.02072336934506893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09781154245138168, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9931519269943238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3170, + "weighted_flow_matching_loss": 0.04979852102696895, + "weighted_rkd_loss": 0.09714166149497032 + }, + { + "epoch": 0.9252254873436137, + "flow_matching_loss": 0.04690953195095062, + "grad_norm": 0.4904327094554901, + "learning_rate": 2.9999270015166983e-05, + "loss": 0.1398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03669877648353577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020090522058308125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09348807483911514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9931086778640748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3180, + "weighted_flow_matching_loss": 0.04690953195095062, + "weighted_rkd_loss": 0.09284379929304123 + }, + { + "epoch": 0.928135001454757, + "flow_matching_loss": 0.04528206028044224, + "grad_norm": 0.41327375173568726, + "learning_rate": 2.999918617509039e-05, + "loss": 0.1356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03606251664459705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018860765919089317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09098579883575439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9930652976036072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3190, + "weighted_flow_matching_loss": 0.04528206028044224, + "weighted_rkd_loss": 0.09035487398505211 + }, + { + "epoch": 0.9310445155659005, + "flow_matching_loss": 0.046080075576901434, + "grad_norm": 0.44631877541542053, + "learning_rate": 2.9999097778657074e-05, + "loss": 0.1379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036420421302318574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01962029356509447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09246113672852516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9930217802524567, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3200, + "weighted_flow_matching_loss": 0.046080075576901434, + "weighted_rkd_loss": 0.09181590378284454 + }, + { + "epoch": 0.9339540296770439, + "flow_matching_loss": 0.04580134525895119, + "grad_norm": 0.424525648355484, + "learning_rate": 2.9999004825893878e-05, + "loss": 0.1335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034979114681482314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018394858669489624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08835308775305747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9929781138896943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3210, + "weighted_flow_matching_loss": 0.04580134525895119, + "weighted_rkd_loss": 0.08773270174860955 + }, + { + "epoch": 0.9368635437881874, + "flow_matching_loss": 0.04957567267119885, + "grad_norm": 0.4820626676082611, + "learning_rate": 2.9998907316829043e-05, + "loss": 0.1407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03582391068339348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020101370103657246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0917491927742958, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9929343283176422, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3220, + "weighted_flow_matching_loss": 0.04957567267119885, + "weighted_rkd_loss": 0.09110097289085388 + }, + { + "epoch": 0.9397730578993309, + "flow_matching_loss": 0.04605819657444954, + "grad_norm": 0.46978023648262024, + "learning_rate": 2.999880525149219e-05, + "loss": 0.1366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03606070876121521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01903413012623787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09115554764866829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9928904056549073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3230, + "weighted_flow_matching_loss": 0.04605819657444954, + "weighted_rkd_loss": 0.0905074842274189 + }, + { + "epoch": 0.9426825720104742, + "flow_matching_loss": 0.044680796563625336, + "grad_norm": 0.46820908784866333, + "learning_rate": 2.999869862991432e-05, + "loss": 0.1347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035767813213169575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01918234135955572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09071796834468841, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9928463578224183, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3240, + "weighted_flow_matching_loss": 0.044680796563625336, + "weighted_rkd_loss": 0.09006899520754814 + }, + { + "epoch": 0.9455920861216177, + "flow_matching_loss": 0.045825622603297235, + "grad_norm": 0.5093196630477905, + "learning_rate": 2.9998587452127822e-05, + "loss": 0.1288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0330185616388917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017586899548769, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08362402394413948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9928021490573883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3250, + "weighted_flow_matching_loss": 0.045825622603297235, + "weighted_rkd_loss": 0.08302212432026863 + }, + { + "epoch": 0.9485016002327611, + "flow_matching_loss": 0.04425225406885147, + "grad_norm": 0.4564504325389862, + "learning_rate": 2.999847171816647e-05, + "loss": 0.1285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03328915610909462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018325239419937134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08490355089306831, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9927578210830689, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3260, + "weighted_flow_matching_loss": 0.04425225406885147, + "weighted_rkd_loss": 0.0842886783182621 + }, + { + "epoch": 0.9514111143439046, + "flow_matching_loss": 0.046575745567679405, + "grad_norm": 0.4690440595149994, + "learning_rate": 2.999835142806543e-05, + "loss": 0.1322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03381998408585787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018632646650075912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08627261519432068, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.992713350057602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3270, + "weighted_flow_matching_loss": 0.046575745567679405, + "weighted_rkd_loss": 0.08564400747418403 + }, + { + "epoch": 0.9543206284550481, + "flow_matching_loss": 0.04824050217866897, + "grad_norm": 0.5173328518867493, + "learning_rate": 2.9998226581861227e-05, + "loss": 0.1351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03431972842663526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018843844160437585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08748330026865006, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9926687598228454, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3280, + "weighted_flow_matching_loss": 0.04824050217866897, + "weighted_rkd_loss": 0.08684198334813117 + }, + { + "epoch": 0.9572301425661914, + "flow_matching_loss": 0.047890130430459976, + "grad_norm": 0.4755450487136841, + "learning_rate": 2.9998097179591793e-05, + "loss": 0.14, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03637285865843296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020012057200074197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09275777563452721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9926240146160126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3290, + "weighted_flow_matching_loss": 0.047890130430459976, + "weighted_rkd_loss": 0.09207353293895722 + }, + { + "epoch": 0.9601396566773349, + "flow_matching_loss": 0.04625278078019619, + "grad_norm": 0.5078374147415161, + "learning_rate": 2.9997963221296443e-05, + "loss": 0.1376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036116741970181464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019817555509507655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0920510396361351, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9925791501998902, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3300, + "weighted_flow_matching_loss": 0.04625278078019619, + "weighted_rkd_loss": 0.09136797487735748 + }, + { + "epoch": 0.9630491707884783, + "flow_matching_loss": 0.04251082725822926, + "grad_norm": 0.539983868598938, + "learning_rate": 2.9997824707015858e-05, + "loss": 0.1317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035590702667832375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018695314042270185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08987671956419944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9925341427326202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3310, + "weighted_flow_matching_loss": 0.04251082725822926, + "weighted_rkd_loss": 0.08920570835471153 + }, + { + "epoch": 0.9659586848996218, + "flow_matching_loss": 0.044784382358193396, + "grad_norm": 0.4814179241657257, + "learning_rate": 2.9997681636792126e-05, + "loss": 0.1283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03326151929795742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017650485783815385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08417352363467216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9924890041351319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3320, + "weighted_flow_matching_loss": 0.044784382358193396, + "weighted_rkd_loss": 0.0835412785410881 + }, + { + "epoch": 0.9688681990107652, + "flow_matching_loss": 0.047401920706033704, + "grad_norm": 0.4869292676448822, + "learning_rate": 2.9997534010668707e-05, + "loss": 0.1403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.036854899302124974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019935167953372003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09364496469497681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9924437284469605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3330, + "weighted_flow_matching_loss": 0.047401920706033704, + "weighted_rkd_loss": 0.0929373823106289 + }, + { + "epoch": 0.9717777131219086, + "flow_matching_loss": 0.047272012755274775, + "grad_norm": 0.5229285955429077, + "learning_rate": 2.9997381828690435e-05, + "loss": 0.1377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03587535880506039, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019378357753157615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09112907499074936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9923983097076416, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3340, + "weighted_flow_matching_loss": 0.047272012755274775, + "weighted_rkd_loss": 0.09043632224202156 + }, + { + "epoch": 0.9746872272330521, + "flow_matching_loss": 0.04838186204433441, + "grad_norm": 0.4267224371433258, + "learning_rate": 2.9997225090903552e-05, + "loss": 0.1372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034785402566194536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019903060141950847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08947386518120766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9923527717590332, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3350, + "weighted_flow_matching_loss": 0.04838186204433441, + "weighted_rkd_loss": 0.08878966420888901 + }, + { + "epoch": 0.9775967413441955, + "flow_matching_loss": 0.04971563182771206, + "grad_norm": 0.48255977034568787, + "learning_rate": 2.9997063797355664e-05, + "loss": 0.1372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0345697158947587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019040234200656415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08817966654896736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9923070907592774, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3360, + "weighted_flow_matching_loss": 0.04971563182771206, + "weighted_rkd_loss": 0.08750130981206894 + }, + { + "epoch": 0.980506255455339, + "flow_matching_loss": 0.04694647416472435, + "grad_norm": 0.5253341197967529, + "learning_rate": 2.9996897948095768e-05, + "loss": 0.1364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0354805763810873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019152098521590232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09011325240135193, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9922612607479095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3370, + "weighted_flow_matching_loss": 0.04694647416472435, + "weighted_rkd_loss": 0.08941586166620255 + }, + { + "epoch": 0.9834157695664824, + "flow_matching_loss": 0.04691217876970768, + "grad_norm": 0.49413493275642395, + "learning_rate": 2.9996727543174244e-05, + "loss": 0.1334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03446252848953009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0182595974765718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08718465343117714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9922153174877166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3380, + "weighted_flow_matching_loss": 0.04691217876970768, + "weighted_rkd_loss": 0.08650595545768738 + }, + { + "epoch": 0.9863252836776258, + "flow_matching_loss": 0.04382318183779717, + "grad_norm": 0.49311864376068115, + "learning_rate": 2.9996552582642857e-05, + "loss": 0.1348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03612521253526211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019479446671903135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09172987267374992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9921692371368408, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3390, + "weighted_flow_matching_loss": 0.04382318183779717, + "weighted_rkd_loss": 0.09101158007979393 + }, + { + "epoch": 0.9892347977887693, + "flow_matching_loss": 0.047038722410798074, + "grad_norm": 0.48314693570137024, + "learning_rate": 2.9996373066554763e-05, + "loss": 0.1336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034419662319123745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018374863918870686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0872141882777214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9921230137348175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3400, + "weighted_flow_matching_loss": 0.047038722410798074, + "weighted_rkd_loss": 0.08652722015976906 + }, + { + "epoch": 0.9921443118999127, + "flow_matching_loss": 0.04746960289776325, + "grad_norm": 0.5305962562561035, + "learning_rate": 2.999618899496448e-05, + "loss": 0.1393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035956916585564616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.020632021874189378, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09254585579037666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9920766651630402, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3410, + "weighted_flow_matching_loss": 0.04746960289776325, + "weighted_rkd_loss": 0.09181254580616952 + }, + { + "epoch": 0.9950538260110562, + "flow_matching_loss": 0.048713619261980055, + "grad_norm": 0.4840381443500519, + "learning_rate": 2.9996000367927936e-05, + "loss": 0.1361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03465256243944168, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018791612423956394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08809673711657524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9920301735401154, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3420, + "weighted_flow_matching_loss": 0.048713619261980055, + "weighted_rkd_loss": 0.08739461824297905 + }, + { + "epoch": 0.9979633401221996, + "flow_matching_loss": 0.04613268487155438, + "grad_norm": 0.5344859957695007, + "learning_rate": 2.9995807185502425e-05, + "loss": 0.1354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035262960381805895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01947529073804617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09000121131539345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9919835448265075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3430, + "weighted_flow_matching_loss": 0.04613268487155438, + "weighted_rkd_loss": 0.08927972316741943 + }, + { + "epoch": 1.000872854233343, + "flow_matching_loss": 0.04403458759188652, + "grad_norm": 0.5217034220695496, + "learning_rate": 2.9995609447746636e-05, + "loss": 0.1338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035814004205167296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01890266630798578, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.090530676394701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9919367849826812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3440, + "weighted_flow_matching_loss": 0.04403458759188652, + "weighted_rkd_loss": 0.08980070725083351 + }, + { + "epoch": 1.0037823683444864, + "flow_matching_loss": 0.04638935998082161, + "grad_norm": 0.4400027394294739, + "learning_rate": 2.999540715472063e-05, + "loss": 0.1324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03420263938605785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018345018848776817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08675029650330543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.991889888048172, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3450, + "weighted_flow_matching_loss": 0.04638935998082161, + "weighted_rkd_loss": 0.08604671210050582 + }, + { + "epoch": 1.00669188245563, + "flow_matching_loss": 0.04353085458278656, + "grad_norm": 0.5085476636886597, + "learning_rate": 2.999520030648586e-05, + "loss": 0.1323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035047631710767746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.019386520236730577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08948178440332413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9918428361415863, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3460, + "weighted_flow_matching_loss": 0.04353085458278656, + "weighted_rkd_loss": 0.0887519046664238 + }, + { + "epoch": 1.0096013965667734, + "flow_matching_loss": 0.04404603913426399, + "grad_norm": 0.4596129059791565, + "learning_rate": 2.9994988903105163e-05, + "loss": 0.1341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.035602858662605284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0196189533919096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09082467034459114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9917956948280334, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3470, + "weighted_flow_matching_loss": 0.04404603913426399, + "weighted_rkd_loss": 0.09007952883839607 + }, + { + "epoch": 1.0125109106779169, + "flow_matching_loss": 0.04422372914850712, + "grad_norm": 0.472866952419281, + "learning_rate": 2.999477294464276e-05, + "loss": 0.1309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03464791774749756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018141245003789663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08743708059191704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9917483866214752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3480, + "weighted_flow_matching_loss": 0.04422372914850712, + "weighted_rkd_loss": 0.08671562150120735 + }, + { + "epoch": 1.0154204247890601, + "flow_matching_loss": 0.04234891794621944, + "grad_norm": 0.4491816759109497, + "learning_rate": 2.9994552431164247e-05, + "loss": 0.1266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.033509586565196516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01794901955872774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08496819287538529, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9917009532451629, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3490, + "weighted_flow_matching_loss": 0.04234891794621944, + "weighted_rkd_loss": 0.08426307290792465 + }, + { + "epoch": 1.0183299389002036, + "flow_matching_loss": 0.04560910239815712, + "grad_norm": 0.4233321249485016, + "learning_rate": 2.9994327362736617e-05, + "loss": 0.1335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03469482492655516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01922950064763427, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08861914947628975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9916533827781677, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3500, + "weighted_flow_matching_loss": 0.04560910239815712, + "weighted_rkd_loss": 0.08787955865263938 + }, + { + "epoch": 1.021239453011347, + "flow_matching_loss": 0.043707939237356185, + "grad_norm": 0.528974711894989, + "learning_rate": 2.9994097739428233e-05, + "loss": 0.1312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03508197329938412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01810248950496316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0882664367556572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9916056811809539, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3510, + "weighted_flow_matching_loss": 0.043707939237356185, + "weighted_rkd_loss": 0.08752544224262238 + }, + { + "epoch": 1.0241489671224906, + "flow_matching_loss": 0.0454102098941803, + "grad_norm": 0.46956634521484375, + "learning_rate": 2.9993863561308856e-05, + "loss": 0.1304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0337782945483923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018134241364896296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08569083139300346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9915578424930572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3520, + "weighted_flow_matching_loss": 0.0454102098941803, + "weighted_rkd_loss": 0.08496739268302918 + }, + { + "epoch": 1.027058481233634, + "flow_matching_loss": 0.04742983803153038, + "grad_norm": 0.5300124287605286, + "learning_rate": 2.9993624828449616e-05, + "loss": 0.1331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03384074401110411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0187005165964365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08638200387358666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9915098667144775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3530, + "weighted_flow_matching_loss": 0.04742983803153038, + "weighted_rkd_loss": 0.08564855828881264 + }, + { + "epoch": 1.0299679953447773, + "flow_matching_loss": 0.043258067592978476, + "grad_norm": 0.44006821513175964, + "learning_rate": 2.9993381540923037e-05, + "loss": 0.1315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034962558187544346, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01903127394616604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08895639106631278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9914617657661438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3540, + "weighted_flow_matching_loss": 0.043258067592978476, + "weighted_rkd_loss": 0.08819688335061074 + }, + { + "epoch": 1.0328775094559208, + "flow_matching_loss": 0.0452148325741291, + "grad_norm": 0.5117239952087402, + "learning_rate": 2.9993133698803025e-05, + "loss": 0.1277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03270390033721924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017768512200564145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08317631185054779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9914135038852692, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3550, + "weighted_flow_matching_loss": 0.0452148325741291, + "weighted_rkd_loss": 0.08246207982301712 + }, + { + "epoch": 1.0357870235670643, + "flow_matching_loss": 0.04537465050816536, + "grad_norm": 0.47335073351860046, + "learning_rate": 2.9992881302164862e-05, + "loss": 0.1348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03565401192754507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018918202444911002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.09022622630000114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9913651466369628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3560, + "weighted_flow_matching_loss": 0.04537465050816536, + "weighted_rkd_loss": 0.08944713175296784 + }, + { + "epoch": 1.0386965376782078, + "flow_matching_loss": 0.04309504255652428, + "grad_norm": 0.4875049889087677, + "learning_rate": 2.9992624351085226e-05, + "loss": 0.1278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.033692685142159465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018046553246676922, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08543192371726036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9913166344165802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3570, + "weighted_flow_matching_loss": 0.04309504255652428, + "weighted_rkd_loss": 0.08469011709094047 + }, + { + "epoch": 1.0416060517893513, + "flow_matching_loss": 0.0443675771355629, + "grad_norm": 0.46711382269859314, + "learning_rate": 2.9992362845642167e-05, + "loss": 0.1277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03304113820195198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018025855347514153, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0841081328690052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9912679851055145, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3580, + "weighted_flow_matching_loss": 0.0443675771355629, + "weighted_rkd_loss": 0.08337366431951523 + }, + { + "epoch": 1.0445155659004945, + "flow_matching_loss": 0.044041484221816064, + "grad_norm": 0.4149789810180664, + "learning_rate": 2.9992096785915132e-05, + "loss": 0.1265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03275566957890987, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01768642133101821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08319776207208633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9912192106246949, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3590, + "weighted_flow_matching_loss": 0.044041484221816064, + "weighted_rkd_loss": 0.08246726989746093 + }, + { + "epoch": 1.047425080011638, + "flow_matching_loss": 0.04543373622000217, + "grad_norm": 0.496957927942276, + "learning_rate": 2.999182617198493e-05, + "loss": 0.1288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03299370147287846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018091273959726095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08407867625355721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9911702871322632, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3600, + "weighted_flow_matching_loss": 0.04543373622000217, + "weighted_rkd_loss": 0.08333630114793777 + }, + { + "epoch": 1.0503345941227815, + "flow_matching_loss": 0.04501156136393547, + "grad_norm": 0.461067259311676, + "learning_rate": 2.999155100393378e-05, + "loss": 0.1321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03449998423457146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01886427402496338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08786424323916435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9911212384700775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3610, + "weighted_flow_matching_loss": 0.04501156136393547, + "weighted_rkd_loss": 0.08708405196666717 + }, + { + "epoch": 1.053244108233925, + "flow_matching_loss": 0.04456478096544743, + "grad_norm": 0.5001130104064941, + "learning_rate": 2.9991271281845256e-05, + "loss": 0.1278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03304407000541687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017883376125246285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08397151604294777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9910720527172089, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3620, + "weighted_flow_matching_loss": 0.04456478096544743, + "weighted_rkd_loss": 0.08322183936834335 + }, + { + "epoch": 1.0561536223450685, + "flow_matching_loss": 0.04595290645956993, + "grad_norm": 0.46374115347862244, + "learning_rate": 2.9990987005804347e-05, + "loss": 0.1296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03334242049604654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017744736559689046, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08442957773804664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9910227239131928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3630, + "weighted_flow_matching_loss": 0.04595290645956993, + "weighted_rkd_loss": 0.0836716540157795 + }, + { + "epoch": 1.0590631364562118, + "flow_matching_loss": 0.0440427977591753, + "grad_norm": 0.4926258623600006, + "learning_rate": 2.999069817589739e-05, + "loss": 0.1245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03229667190462351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016624969709664584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0812183141708374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9909732639789581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3640, + "weighted_flow_matching_loss": 0.0440427977591753, + "weighted_rkd_loss": 0.08048512414097786 + }, + { + "epoch": 1.0619726505673552, + "flow_matching_loss": 0.0484482865780592, + "grad_norm": 0.43041738867759705, + "learning_rate": 2.9990404792212142e-05, + "loss": 0.1297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03226968050003052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017470532562583685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08200989216566086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9909236788749695, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3650, + "weighted_flow_matching_loss": 0.0484482865780592, + "weighted_rkd_loss": 0.08126559630036354 + }, + { + "epoch": 1.0648821646784987, + "flow_matching_loss": 0.04461631327867508, + "grad_norm": 0.4851453900337219, + "learning_rate": 2.9990106854837715e-05, + "loss": 0.1256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03228714894503355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017118009459227323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08169230669736863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9908739566802979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3660, + "weighted_flow_matching_loss": 0.04461631327867508, + "weighted_rkd_loss": 0.08094682246446609 + }, + { + "epoch": 1.0677916787896422, + "flow_matching_loss": 0.04428999237716198, + "grad_norm": 0.5860196948051453, + "learning_rate": 2.9989804363864615e-05, + "loss": 0.1302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.034108230099081996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018472318537533283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08668877929449081, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9908241033554077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3670, + "weighted_flow_matching_loss": 0.04428999237716198, + "weighted_rkd_loss": 0.08589325249195098 + }, + { + "epoch": 1.0707011929007855, + "flow_matching_loss": 0.04573749266564846, + "grad_norm": 0.5319429636001587, + "learning_rate": 2.998949731938473e-05, + "loss": 0.1282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03291695937514305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01738279853016138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0832167163491249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9907740950584412, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3680, + "weighted_flow_matching_loss": 0.04573749266564846, + "weighted_rkd_loss": 0.08244893923401833 + }, + { + "epoch": 1.073610707011929, + "flow_matching_loss": 0.04488167352974415, + "grad_norm": 0.43998783826828003, + "learning_rate": 2.9989185721491338e-05, + "loss": 0.1246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031851994805037975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016731715109199286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08043570443987846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.990723979473114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3690, + "weighted_flow_matching_loss": 0.04488167352974415, + "weighted_rkd_loss": 0.07968959957361221 + }, + { + "epoch": 1.0765202211230724, + "flow_matching_loss": 0.04584039114415646, + "grad_norm": 0.554949939250946, + "learning_rate": 2.998886957027909e-05, + "loss": 0.1301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03345735259354114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.018090845737606286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08500555008649827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.990673714876175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3700, + "weighted_flow_matching_loss": 0.04584039114415646, + "weighted_rkd_loss": 0.08421272337436676 + }, + { + "epoch": 1.079429735234216, + "flow_matching_loss": 0.040666161477565764, + "grad_norm": 0.4420180916786194, + "learning_rate": 2.998854886584402e-05, + "loss": 0.1196, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031643917039036754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01637472501024604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07966255918145179, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9906233131885529, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3710, + "weighted_flow_matching_loss": 0.040666161477565764, + "weighted_rkd_loss": 0.07891559228301048 + }, + { + "epoch": 1.0823392493453594, + "flow_matching_loss": 0.04127015620470047, + "grad_norm": 0.4763873815536499, + "learning_rate": 2.9988223608283557e-05, + "loss": 0.1254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.033456813916563986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017980526387691497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08489415273070336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9905727803707123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3720, + "weighted_flow_matching_loss": 0.04127015620470047, + "weighted_rkd_loss": 0.08409383594989776 + }, + { + "epoch": 1.0852487634565027, + "flow_matching_loss": 0.043866214156150815, + "grad_norm": 0.4694620668888092, + "learning_rate": 2.9987893797696502e-05, + "loss": 0.1273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03330539278686047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01758556831628084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08419635444879532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9905221104621887, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3730, + "weighted_flow_matching_loss": 0.043866214156150815, + "weighted_rkd_loss": 0.08339838832616805 + }, + { + "epoch": 1.0881582775676462, + "flow_matching_loss": 0.041367531940341, + "grad_norm": 0.49686866998672485, + "learning_rate": 2.9987559434183046e-05, + "loss": 0.1225, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03215599190443754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01759177092462778, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08190375491976738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9904713213443757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3740, + "weighted_flow_matching_loss": 0.041367531940341, + "weighted_rkd_loss": 0.0811232477426529 + }, + { + "epoch": 1.0910677916787896, + "flow_matching_loss": 0.046709807217121126, + "grad_norm": 0.4249069094657898, + "learning_rate": 2.998722051784476e-05, + "loss": 0.125, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031087074242532254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016872006375342606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07904615327715873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9904203891754151, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3750, + "weighted_flow_matching_loss": 0.046709807217121126, + "weighted_rkd_loss": 0.07828899100422859 + }, + { + "epoch": 1.0939773057899331, + "flow_matching_loss": 0.04251766055822372, + "grad_norm": 0.43775463104248047, + "learning_rate": 2.9986877048784594e-05, + "loss": 0.1238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032274245843291284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01754233883693814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08209083080291749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.990369313955307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3760, + "weighted_flow_matching_loss": 0.04251766055822372, + "weighted_rkd_loss": 0.08130023404955863 + }, + { + "epoch": 1.0968868199010766, + "flow_matching_loss": 0.044484587386250496, + "grad_norm": 0.44643107056617737, + "learning_rate": 2.9986529027106885e-05, + "loss": 0.126, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03242856599390507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017491146735846997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08234827890992165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9903181195259094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3770, + "weighted_flow_matching_loss": 0.044484587386250496, + "weighted_rkd_loss": 0.08155102506279946 + }, + { + "epoch": 1.0997963340122199, + "flow_matching_loss": 0.04440926685929299, + "grad_norm": 0.4147959053516388, + "learning_rate": 2.998617645291735e-05, + "loss": 0.1246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03202536590397358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016945678275078535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08099640905857086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9902667760848999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3780, + "weighted_flow_matching_loss": 0.04440926685929299, + "weighted_rkd_loss": 0.0802080862224102 + }, + { + "epoch": 1.1027058481233634, + "flow_matching_loss": 0.044334438443183896, + "grad_norm": 0.4103468656539917, + "learning_rate": 2.9985819326323097e-05, + "loss": 0.1227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031399757601320744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01631895238533616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07911846861243248, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9902153015136719, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3790, + "weighted_flow_matching_loss": 0.044334438443183896, + "weighted_rkd_loss": 0.07834439128637313 + }, + { + "epoch": 1.1056153622345068, + "flow_matching_loss": 0.04570821300148964, + "grad_norm": 0.3951588571071625, + "learning_rate": 2.9985457647432613e-05, + "loss": 0.1256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03205643240362406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016615511476993562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08072837591171264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9901636958122253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3800, + "weighted_flow_matching_loss": 0.04570821300148964, + "weighted_rkd_loss": 0.07993433699011802 + }, + { + "epoch": 1.1085248763456503, + "flow_matching_loss": 0.0396812729537487, + "grad_norm": 0.39888796210289, + "learning_rate": 2.998509141635576e-05, + "loss": 0.119, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03187416587024927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01639127526432276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08013960793614387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9901119589805603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3810, + "weighted_flow_matching_loss": 0.0396812729537487, + "weighted_rkd_loss": 0.07934723980724812 + }, + { + "epoch": 1.1114343904567938, + "flow_matching_loss": 0.043942775949835776, + "grad_norm": 0.4209740459918976, + "learning_rate": 2.99847206332038e-05, + "loss": 0.1229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031715014949440956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01627635918557644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07970638871192932, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9900600790977478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3820, + "weighted_flow_matching_loss": 0.043942775949835776, + "weighted_rkd_loss": 0.07891410887241364 + }, + { + "epoch": 1.114343904567937, + "flow_matching_loss": 0.04229394197463989, + "grad_norm": 0.5164817571640015, + "learning_rate": 2.9984345298089356e-05, + "loss": 0.1209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031236148066818716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01688261888921261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07935491502285004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9900080680847168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3830, + "weighted_flow_matching_loss": 0.04229394197463989, + "weighted_rkd_loss": 0.07856205329298974 + }, + { + "epoch": 1.1172534186790806, + "flow_matching_loss": 0.04163449853658676, + "grad_norm": 0.449421226978302, + "learning_rate": 2.998396541112645e-05, + "loss": 0.1157, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02972401920706034, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015323937684297562, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0747719757258892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9899559199810029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3840, + "weighted_flow_matching_loss": 0.04163449853658676, + "weighted_rkd_loss": 0.0740209698677063 + }, + { + "epoch": 1.120162932790224, + "flow_matching_loss": 0.04432117342948914, + "grad_norm": 0.43072566390037537, + "learning_rate": 2.998358097243048e-05, + "loss": 0.1194, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030197258293628692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015494022145867347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0758885383605957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9899036467075348, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3850, + "weighted_flow_matching_loss": 0.04432117342948914, + "weighted_rkd_loss": 0.07512233108282089 + }, + { + "epoch": 1.1230724469013675, + "flow_matching_loss": 0.04250144585967064, + "grad_norm": 0.4346248507499695, + "learning_rate": 2.998319198211823e-05, + "loss": 0.1209, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03126058373600245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016639840230345725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07916100770235061, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9898512363433838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3860, + "weighted_flow_matching_loss": 0.04250144585967064, + "weighted_rkd_loss": 0.07835762724280357 + }, + { + "epoch": 1.1259819610125108, + "flow_matching_loss": 0.041782664507627486, + "grad_norm": 0.4396354556083679, + "learning_rate": 2.998279844030786e-05, + "loss": 0.1167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029963991045951842, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01576798325404525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07569596543908119, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9897986829280854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3870, + "weighted_flow_matching_loss": 0.041782664507627486, + "weighted_rkd_loss": 0.07492381557822228 + }, + { + "epoch": 1.1288914751236543, + "flow_matching_loss": 0.04293069504201412, + "grad_norm": 0.44866999983787537, + "learning_rate": 2.9982400347118926e-05, + "loss": 0.1216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03144175540655851, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016639978624880314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07952348962426185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9897460043430328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3880, + "weighted_flow_matching_loss": 0.04293069504201412, + "weighted_rkd_loss": 0.07870804518461227 + }, + { + "epoch": 1.1318009892347978, + "flow_matching_loss": 0.04410448297858238, + "grad_norm": 0.4898183047771454, + "learning_rate": 2.9981997702672353e-05, + "loss": 0.1255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032570486143231395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01714574946090579, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08228671997785568, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9896931827068329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3890, + "weighted_flow_matching_loss": 0.04410448297858238, + "weighted_rkd_loss": 0.08143864944577217 + }, + { + "epoch": 1.1347105033459413, + "flow_matching_loss": 0.041243035718798636, + "grad_norm": 0.46817559003829956, + "learning_rate": 2.9981590507090454e-05, + "loss": 0.1167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030179235339164733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01589019615203142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07624866664409638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9896402418613434, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3900, + "weighted_flow_matching_loss": 0.041243035718798636, + "weighted_rkd_loss": 0.07545878291130066 + }, + { + "epoch": 1.1376200174570847, + "flow_matching_loss": 0.04447698965668678, + "grad_norm": 0.46651068329811096, + "learning_rate": 2.9981178760496927e-05, + "loss": 0.1219, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030947642214596272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016391894407570363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07828717902302743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9895871460437775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3910, + "weighted_flow_matching_loss": 0.04447698965668678, + "weighted_rkd_loss": 0.07747204527258873 + }, + { + "epoch": 1.140529531568228, + "flow_matching_loss": 0.04400472827255726, + "grad_norm": 0.44160833954811096, + "learning_rate": 2.9980762463016848e-05, + "loss": 0.1189, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029798135720193387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01609282484278083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07568909600377083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.989533931016922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3920, + "weighted_flow_matching_loss": 0.04400472827255726, + "weighted_rkd_loss": 0.07489696517586708 + }, + { + "epoch": 1.1434390456793715, + "flow_matching_loss": 0.04250936768949032, + "grad_norm": 0.529961109161377, + "learning_rate": 2.9980341614776673e-05, + "loss": 0.124, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.032554307393729685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.017260128632187843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08236874341964721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9894805788993836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3930, + "weighted_flow_matching_loss": 0.04250936768949032, + "weighted_rkd_loss": 0.08150230199098588 + }, + { + "epoch": 1.146348559790515, + "flow_matching_loss": 0.04363785907626152, + "grad_norm": 0.5008299946784973, + "learning_rate": 2.9979916215904247e-05, + "loss": 0.12, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0308285778388381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01554818768054247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07720534279942512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9894270896911621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3940, + "weighted_flow_matching_loss": 0.04363785907626152, + "weighted_rkd_loss": 0.07638896927237511 + }, + { + "epoch": 1.1492580739016585, + "flow_matching_loss": 0.045806118100881574, + "grad_norm": 0.4383701682090759, + "learning_rate": 2.99794862665288e-05, + "loss": 0.1205, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030234663188457488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015000730473548173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07547005712985992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9893734753131866, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3950, + "weighted_flow_matching_loss": 0.045806118100881574, + "weighted_rkd_loss": 0.07466804459691048 + }, + { + "epoch": 1.152167588012802, + "flow_matching_loss": 0.04325656443834305, + "grad_norm": 0.4562664031982422, + "learning_rate": 2.9979051766780938e-05, + "loss": 0.1186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030191951617598533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015808488056063653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07619239315390587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9893197119235992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3960, + "weighted_flow_matching_loss": 0.04325656443834305, + "weighted_rkd_loss": 0.07537864036858082 + }, + { + "epoch": 1.1550771021239452, + "flow_matching_loss": 0.04156697317957878, + "grad_norm": 0.4995730519294739, + "learning_rate": 2.9978612716792647e-05, + "loss": 0.1176, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030470831133425236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015932001825422048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0768736645579338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9892658293247223, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3970, + "weighted_flow_matching_loss": 0.04156697317957878, + "weighted_rkd_loss": 0.07604842111468316 + }, + { + "epoch": 1.1579866162350887, + "flow_matching_loss": 0.04305365048348904, + "grad_norm": 0.5578781366348267, + "learning_rate": 2.99781691166973e-05, + "loss": 0.1246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0327368400990963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016954601835459472, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08242828287184238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9892118036746979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3980, + "weighted_flow_matching_loss": 0.04305365048348904, + "weighted_rkd_loss": 0.0815389271825552 + }, + { + "epoch": 1.1608961303462322, + "flow_matching_loss": 0.04190938360989094, + "grad_norm": 0.5652178525924683, + "learning_rate": 2.9977720966629646e-05, + "loss": 0.1217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03207990787923336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016551281046122312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08071109652519226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.989157646894455, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 3990, + "weighted_flow_matching_loss": 0.04190938360989094, + "weighted_rkd_loss": 0.07983597815036773 + }, + { + "epoch": 1.1638056444573757, + "flow_matching_loss": 0.04288073554635048, + "grad_norm": 0.4291142225265503, + "learning_rate": 2.997726826672583e-05, + "loss": 0.1215, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031275549158453944, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016957372520118953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07950847074389458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.989103353023529, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4000, + "weighted_flow_matching_loss": 0.04288073554635048, + "weighted_rkd_loss": 0.07864213213324547 + }, + { + "epoch": 1.1667151585685192, + "flow_matching_loss": 0.043655458837747574, + "grad_norm": 0.49047231674194336, + "learning_rate": 2.9976811017123368e-05, + "loss": 0.121, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.031140678375959397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01595188844949007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0782332442700863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9890489220619202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4010, + "weighted_flow_matching_loss": 0.043655458837747574, + "weighted_rkd_loss": 0.0773764930665493 + }, + { + "epoch": 1.1696246726796624, + "flow_matching_loss": 0.042031294852495196, + "grad_norm": 0.4296210706233978, + "learning_rate": 2.9976349217961158e-05, + "loss": 0.1167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029732710123062132, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016078327782452106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07554374933242798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9889943599700928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4020, + "weighted_flow_matching_loss": 0.042031294852495196, + "weighted_rkd_loss": 0.074712385982275 + }, + { + "epoch": 1.172534186790806, + "flow_matching_loss": 0.04321780242025852, + "grad_norm": 0.46372491121292114, + "learning_rate": 2.9975882869379482e-05, + "loss": 0.1195, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03051628414541483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016093122120946646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07712569162249565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9889396727085114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4030, + "weighted_flow_matching_loss": 0.04321780242025852, + "weighted_rkd_loss": 0.07627263516187668 + }, + { + "epoch": 1.1754437009019494, + "flow_matching_loss": 0.04241251777857542, + "grad_norm": 0.48793748021125793, + "learning_rate": 2.997541197152001e-05, + "loss": 0.1134, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028438241221010685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014878420624881983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07175490260124207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9888848423957824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4040, + "weighted_flow_matching_loss": 0.04241251777857542, + "weighted_rkd_loss": 0.07095733731985092 + }, + { + "epoch": 1.1783532150130929, + "flow_matching_loss": 0.04265654869377613, + "grad_norm": 0.4759970009326935, + "learning_rate": 2.9974936524525774e-05, + "loss": 0.1231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03231102842837572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016754102893173695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.08137615919113159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9888298809528351, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4050, + "weighted_flow_matching_loss": 0.04265654869377613, + "weighted_rkd_loss": 0.08046722933650016 + }, + { + "epoch": 1.1812627291242364, + "flow_matching_loss": 0.04580349698662758, + "grad_norm": 0.43773001432418823, + "learning_rate": 2.9974456528541222e-05, + "loss": 0.1231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030988800711929797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016200535371899603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07817813605070115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9887747704982758, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4060, + "weighted_flow_matching_loss": 0.04580349698662758, + "weighted_rkd_loss": 0.07730061709880828 + }, + { + "epoch": 1.1841722432353796, + "flow_matching_loss": 0.04137607999145985, + "grad_norm": 0.41860976815223694, + "learning_rate": 2.9973971983712147e-05, + "loss": 0.1161, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030119987204670906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01529259718954563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07553257197141647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9887195527553558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4070, + "weighted_flow_matching_loss": 0.04137607999145985, + "weighted_rkd_loss": 0.07468054741621018 + }, + { + "epoch": 1.187081757346523, + "flow_matching_loss": 0.040230756625533104, + "grad_norm": 0.5386835336685181, + "learning_rate": 2.9973482890185753e-05, + "loss": 0.1149, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03008071593940258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015335477329790593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07549690902233124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9886641800403595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4080, + "weighted_flow_matching_loss": 0.040230756625533104, + "weighted_rkd_loss": 0.07464106008410454 + }, + { + "epoch": 1.1899912714576666, + "flow_matching_loss": 0.04297571666538715, + "grad_norm": 0.46928226947784424, + "learning_rate": 2.9972989248110606e-05, + "loss": 0.12, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030781496316194534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01638573156669736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07794872373342514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9886086881160736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4090, + "weighted_flow_matching_loss": 0.04297571666538715, + "weighted_rkd_loss": 0.0770608089864254 + }, + { + "epoch": 1.19290078556881, + "flow_matching_loss": 0.04121788777410984, + "grad_norm": 0.46700266003608704, + "learning_rate": 2.9972491057636665e-05, + "loss": 0.1169, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030060683749616145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016416792292147875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07653815969824791, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9885530591011047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4100, + "weighted_flow_matching_loss": 0.04121788777410984, + "weighted_rkd_loss": 0.0756620593369007 + }, + { + "epoch": 1.1958102996799536, + "flow_matching_loss": 0.04361990988254547, + "grad_norm": 0.46604979038238525, + "learning_rate": 2.9971988318915268e-05, + "loss": 0.1188, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03022413942962885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015588521398603917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07603680044412613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9884972870349884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4110, + "weighted_flow_matching_loss": 0.04361990988254547, + "weighted_rkd_loss": 0.07516221031546592 + }, + { + "epoch": 1.1987198137910968, + "flow_matching_loss": 0.04198716655373573, + "grad_norm": 0.5023899078369141, + "learning_rate": 2.997148103209913e-05, + "loss": 0.1185, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03055605012923479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016311628650873898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0774237286299467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.988441389799118, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4120, + "weighted_flow_matching_loss": 0.04198716655373573, + "weighted_rkd_loss": 0.07652883045375347 + }, + { + "epoch": 1.2016293279022403, + "flow_matching_loss": 0.043975270539522174, + "grad_norm": 0.4679531455039978, + "learning_rate": 2.9970969197342356e-05, + "loss": 0.1198, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03032873198390007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.016015106253325938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07667257115244866, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9883853554725647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4130, + "weighted_flow_matching_loss": 0.043975270539522174, + "weighted_rkd_loss": 0.07578203529119491 + }, + { + "epoch": 1.2045388420133838, + "flow_matching_loss": 0.04023208096623421, + "grad_norm": 0.4782048761844635, + "learning_rate": 2.9970452814800422e-05, + "loss": 0.1159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030482966639101505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01555713452398777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07652306705713272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9883291900157929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4140, + "weighted_flow_matching_loss": 0.04023208096623421, + "weighted_rkd_loss": 0.07562991678714752 + }, + { + "epoch": 1.2074483561245273, + "flow_matching_loss": 0.04089884087443352, + "grad_norm": 0.4799026846885681, + "learning_rate": 2.99699318846302e-05, + "loss": 0.1167, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0305167343467474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0156369904987514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07667045891284943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.988272887468338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4150, + "weighted_flow_matching_loss": 0.04089884087443352, + "weighted_rkd_loss": 0.07577135190367698 + }, + { + "epoch": 1.2103578702356708, + "flow_matching_loss": 0.04185736812651157, + "grad_norm": 0.47331735491752625, + "learning_rate": 2.9969406406989927e-05, + "loss": 0.1159, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02970480117946863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015497720893472433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07490732297301292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9882164537906647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4160, + "weighted_flow_matching_loss": 0.04185736812651157, + "weighted_rkd_loss": 0.07402465790510178 + }, + { + "epoch": 1.213267384346814, + "flow_matching_loss": 0.0413203340023756, + "grad_norm": 0.46581435203552246, + "learning_rate": 2.996887638203923e-05, + "loss": 0.1128, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028909534588456152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01450604321435094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07232511155307293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9881598651409149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4170, + "weighted_flow_matching_loss": 0.0413203340023756, + "weighted_rkd_loss": 0.07146877013146877 + }, + { + "epoch": 1.2161768984579575, + "flow_matching_loss": 0.0418656550347805, + "grad_norm": 0.5281745791435242, + "learning_rate": 2.9968341809939116e-05, + "loss": 0.1187, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030777533538639545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01615919144824147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07771425917744637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9881031811237335, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4180, + "weighted_flow_matching_loss": 0.0418656550347805, + "weighted_rkd_loss": 0.07678968608379363 + }, + { + "epoch": 1.219086412569101, + "flow_matching_loss": 0.04445470198988914, + "grad_norm": 0.4979766309261322, + "learning_rate": 2.9967802690851985e-05, + "loss": 0.1163, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028794347867369653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015112528204917907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07270122356712819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9880463421344757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4190, + "weighted_flow_matching_loss": 0.04445470198988914, + "weighted_rkd_loss": 0.07183216474950313 + }, + { + "epoch": 1.2219959266802445, + "flow_matching_loss": 0.0424899622797966, + "grad_norm": 0.44374218583106995, + "learning_rate": 2.9967259024941595e-05, + "loss": 0.1182, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030442866310477257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015737684350460766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07662341743707657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9879893720149994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4200, + "weighted_flow_matching_loss": 0.0424899622797966, + "weighted_rkd_loss": 0.07570311315357685 + }, + { + "epoch": 1.2249054407913877, + "flow_matching_loss": 0.041402389854192735, + "grad_norm": 0.43194690346717834, + "learning_rate": 2.9966710812373097e-05, + "loss": 0.1155, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029690224677324295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015591346099972725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07497179582715034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9879322588443756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4210, + "weighted_flow_matching_loss": 0.041402389854192735, + "weighted_rkd_loss": 0.07406710758805275 + }, + { + "epoch": 1.2278149549025312, + "flow_matching_loss": 0.04359724670648575, + "grad_norm": 0.382061243057251, + "learning_rate": 2.996615805331303e-05, + "loss": 0.1108, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027046102099120617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01394684361293912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0680390477180481, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9878750205039978, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4220, + "weighted_flow_matching_loss": 0.04359724670648575, + "weighted_rkd_loss": 0.06721407845616341 + }, + { + "epoch": 1.2307244690136747, + "flow_matching_loss": 0.040113484114408494, + "grad_norm": 0.4165836572647095, + "learning_rate": 2.99656007479293e-05, + "loss": 0.1085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027345597557723523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014551467821002006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06924266256392002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9878176510334015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4230, + "weighted_flow_matching_loss": 0.040113484114408494, + "weighted_rkd_loss": 0.06839914321899414 + }, + { + "epoch": 1.2336339831248182, + "flow_matching_loss": 0.041287968307733534, + "grad_norm": 0.46773919463157654, + "learning_rate": 2.9965038896391213e-05, + "loss": 0.1117, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02816312275826931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014993552304804324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07131979763507842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9877601444721222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4240, + "weighted_flow_matching_loss": 0.041287968307733534, + "weighted_rkd_loss": 0.07044684365391732 + }, + { + "epoch": 1.2365434972359617, + "flow_matching_loss": 0.042412680387496945, + "grad_norm": 0.43552693724632263, + "learning_rate": 2.9964472498869434e-05, + "loss": 0.117, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.03004682995378971, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01545442147180438, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07554808109998704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9877025008201599, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4250, + "weighted_flow_matching_loss": 0.042412680387496945, + "weighted_rkd_loss": 0.07461904883384704 + }, + { + "epoch": 1.239453011347105, + "flow_matching_loss": 0.03958940841257572, + "grad_norm": 0.467896044254303, + "learning_rate": 2.996390155553603e-05, + "loss": 0.1089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027920983359217645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014351496752351523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07019346244633198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9876447379589081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4260, + "weighted_flow_matching_loss": 0.03958940841257572, + "weighted_rkd_loss": 0.06932613104581833 + }, + { + "epoch": 1.2423625254582484, + "flow_matching_loss": 0.04227674212306738, + "grad_norm": 0.4251887798309326, + "learning_rate": 2.9963326066564426e-05, + "loss": 0.1079, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026273897476494314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013936576433479786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0664843712002039, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9875868260860443, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4270, + "weighted_flow_matching_loss": 0.04227674212306738, + "weighted_rkd_loss": 0.06565910317003727 + }, + { + "epoch": 1.245272039569392, + "flow_matching_loss": 0.041303524374961854, + "grad_norm": 0.41843855381011963, + "learning_rate": 2.996274603212945e-05, + "loss": 0.1131, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02890071552246809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014872029051184654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07267345935106277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.987528783082962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4280, + "weighted_flow_matching_loss": 0.041303524374961854, + "weighted_rkd_loss": 0.07176710441708564 + }, + { + "epoch": 1.2481815536805354, + "flow_matching_loss": 0.042651550844311714, + "grad_norm": 0.3982480764389038, + "learning_rate": 2.9962161452407296e-05, + "loss": 0.1119, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027869426272809507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014435934368520975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0701747864484787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9874706089496612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4290, + "weighted_flow_matching_loss": 0.042651550844311714, + "weighted_rkd_loss": 0.06929559633135796 + }, + { + "epoch": 1.2510910677916787, + "flow_matching_loss": 0.04174447394907475, + "grad_norm": 0.4543277323246002, + "learning_rate": 2.996157232757555e-05, + "loss": 0.112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028277757950127123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014545171521604061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07110068611800671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.987412303686142, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4300, + "weighted_flow_matching_loss": 0.04174447394907475, + "weighted_rkd_loss": 0.07020573690533638 + }, + { + "epoch": 1.2540005819028222, + "flow_matching_loss": 0.040239982306957245, + "grad_norm": 0.40806347131729126, + "learning_rate": 2.996097865781316e-05, + "loss": 0.1122, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029124313034117222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014670508354902268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07291913442313672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9873538553714752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4310, + "weighted_flow_matching_loss": 0.040239982306957245, + "weighted_rkd_loss": 0.07199698835611343 + }, + { + "epoch": 1.2569100960139656, + "flow_matching_loss": 0.03904929421842098, + "grad_norm": 0.4609018862247467, + "learning_rate": 2.996038044330048e-05, + "loss": 0.1087, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028200980462133884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01417368222028017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07057564333081245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9872952818870544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4320, + "weighted_flow_matching_loss": 0.03904929421842098, + "weighted_rkd_loss": 0.06967897824943066 + }, + { + "epoch": 1.2598196101251091, + "flow_matching_loss": 0.04105313159525394, + "grad_norm": 0.45566943287849426, + "learning_rate": 2.9959777684219233e-05, + "loss": 0.1084, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027282741479575635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01361660286784172, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06818208545446396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9872365713119506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4330, + "weighted_flow_matching_loss": 0.04105313159525394, + "weighted_rkd_loss": 0.0673118658363819 + }, + { + "epoch": 1.2627291242362526, + "flow_matching_loss": 0.04016308225691319, + "grad_norm": 0.40696367621421814, + "learning_rate": 2.9959170380752508e-05, + "loss": 0.1055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02635580599308014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013456965424120426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06616857722401619, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9871777355670929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4340, + "weighted_flow_matching_loss": 0.04016308225691319, + "weighted_rkd_loss": 0.0653202086687088 + }, + { + "epoch": 1.2656386383473959, + "flow_matching_loss": 0.04097651410847902, + "grad_norm": 0.48358821868896484, + "learning_rate": 2.99585585330848e-05, + "loss": 0.1099, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027564342319965362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014719576481729746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06984826102852822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9871187508106232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4350, + "weighted_flow_matching_loss": 0.04097651410847902, + "weighted_rkd_loss": 0.06894847229123116 + }, + { + "epoch": 1.2685481524585394, + "flow_matching_loss": 0.04292388819158077, + "grad_norm": 0.41242003440856934, + "learning_rate": 2.9957942141401967e-05, + "loss": 0.113, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028075540065765382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014801801741123199, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07095288261771202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9870596408843995, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4360, + "weighted_flow_matching_loss": 0.04292388819158077, + "weighted_rkd_loss": 0.07003476284444332 + }, + { + "epoch": 1.2714576665696828, + "flow_matching_loss": 0.038737544789910316, + "grad_norm": 0.48865798115730286, + "learning_rate": 2.995732120589125e-05, + "loss": 0.1085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028114462457597255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014425077941268682, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0706540022045374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9870004057884216, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4370, + "weighted_flow_matching_loss": 0.038737544789910316, + "weighted_rkd_loss": 0.06973544955253601 + }, + { + "epoch": 1.2743671806808263, + "flow_matching_loss": 0.041081962361931804, + "grad_norm": 0.4036014974117279, + "learning_rate": 2.9956695726741276e-05, + "loss": 0.1117, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028674367256462573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014220249094069003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07156898379325867, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9869410216808319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4380, + "weighted_flow_matching_loss": 0.041081962361931804, + "weighted_rkd_loss": 0.07063439153134823 + }, + { + "epoch": 1.2772766947919698, + "flow_matching_loss": 0.043696265667676926, + "grad_norm": 0.4805940091609955, + "learning_rate": 2.995606570414205e-05, + "loss": 0.1149, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028604856319725515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014936226326972247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07214593961834907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9868814945220947, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4390, + "weighted_flow_matching_loss": 0.043696265667676926, + "weighted_rkd_loss": 0.0711994707584381 + }, + { + "epoch": 1.280186208903113, + "flow_matching_loss": 0.04269919618964195, + "grad_norm": 0.4406993091106415, + "learning_rate": 2.9955431138284955e-05, + "loss": 0.112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02786965500563383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01448278920724988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07022209800779819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9868218660354614, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4400, + "weighted_flow_matching_loss": 0.04269919618964195, + "weighted_rkd_loss": 0.0692967303097248 + }, + { + "epoch": 1.2830957230142566, + "flow_matching_loss": 0.042334765940904615, + "grad_norm": 0.4727557301521301, + "learning_rate": 2.9954792029362754e-05, + "loss": 0.1127, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028184352815151213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014979125559329986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07134783156216144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9867620885372161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4410, + "weighted_flow_matching_loss": 0.042334765940904615, + "weighted_rkd_loss": 0.0704033188521862 + }, + { + "epoch": 1.2860052371254, + "flow_matching_loss": 0.042285578697919844, + "grad_norm": 0.4344619810581207, + "learning_rate": 2.9954148377569598e-05, + "loss": 0.11, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027363698557019235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01389720756560564, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06862460486590863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9867021799087524, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4420, + "weighted_flow_matching_loss": 0.042285578697919844, + "weighted_rkd_loss": 0.06771210059523583 + }, + { + "epoch": 1.2889147512365435, + "flow_matching_loss": 0.04320454336702824, + "grad_norm": 0.514594554901123, + "learning_rate": 2.9953500183101002e-05, + "loss": 0.1145, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028526553884148598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015190055593848228, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07224316447973252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9866421341896057, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4430, + "weighted_flow_matching_loss": 0.04320454336702824, + "weighted_rkd_loss": 0.07127803340554237 + }, + { + "epoch": 1.291824265347687, + "flow_matching_loss": 0.043726541101932526, + "grad_norm": 0.5330427289009094, + "learning_rate": 2.9952847446153877e-05, + "loss": 0.1186, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.030229745991528036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015394726116210223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07585421800613404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.986581951379776, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4440, + "weighted_flow_matching_loss": 0.043726541101932526, + "weighted_rkd_loss": 0.07483639046549798 + }, + { + "epoch": 1.2947337794588303, + "flow_matching_loss": 0.04362403117120266, + "grad_norm": 0.4384973645210266, + "learning_rate": 2.9952190166926498e-05, + "loss": 0.1156, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029140431433916092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014727239031344652, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0730081021785736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9865216493606568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4450, + "weighted_flow_matching_loss": 0.04362403117120266, + "weighted_rkd_loss": 0.07202408984303474 + }, + { + "epoch": 1.2976432935699738, + "flow_matching_loss": 0.041142093017697334, + "grad_norm": 0.36471474170684814, + "learning_rate": 2.9951528345618546e-05, + "loss": 0.1141, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029276941530406476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01541943047195673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07397331371903419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9864611983299255, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4460, + "weighted_flow_matching_loss": 0.041142093017697334, + "weighted_rkd_loss": 0.0729718629270792 + }, + { + "epoch": 1.3005528076811173, + "flow_matching_loss": 0.04097845666110515, + "grad_norm": 0.4118562340736389, + "learning_rate": 2.995086198243105e-05, + "loss": 0.1084, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027170504815876483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01400219164788723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06834320127964019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9864006221294404, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4470, + "weighted_flow_matching_loss": 0.04097845666110515, + "weighted_rkd_loss": 0.06741386651992798 + }, + { + "epoch": 1.3034623217922607, + "flow_matching_loss": 0.044914058223366735, + "grad_norm": 0.4491333067417145, + "learning_rate": 2.9950191077566434e-05, + "loss": 0.1183, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.029665614664554595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.015067136008292436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07439836487174034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9863399267196655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4480, + "weighted_flow_matching_loss": 0.044914058223366735, + "weighted_rkd_loss": 0.07338209748268128 + }, + { + "epoch": 1.3063718359034042, + "flow_matching_loss": 0.03919374924153089, + "grad_norm": 0.49771034717559814, + "learning_rate": 2.9949515631228515e-05, + "loss": 0.1064, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026994217373430728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014147477131336928, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06813591122627258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9862790703773499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4490, + "weighted_flow_matching_loss": 0.03919374924153089, + "weighted_rkd_loss": 0.06720103472471237 + }, + { + "epoch": 1.3092813500145475, + "flow_matching_loss": 0.039618587121367455, + "grad_norm": 0.4027614891529083, + "learning_rate": 2.994883564362246e-05, + "loss": 0.1052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026504097133874895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013459660392254592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0664678554981947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9862180888652802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4500, + "weighted_flow_matching_loss": 0.039618587121367455, + "weighted_rkd_loss": 0.06555180437862873 + }, + { + "epoch": 1.312190864125691, + "flow_matching_loss": 0.0405099131166935, + "grad_norm": 0.4294455349445343, + "learning_rate": 2.994815111495484e-05, + "loss": 0.109, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0275558490306139, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014306957367807627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06941865645349025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9861569762229919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4510, + "weighted_flow_matching_loss": 0.0405099131166935, + "weighted_rkd_loss": 0.06845764592289924 + }, + { + "epoch": 1.3151003782368345, + "flow_matching_loss": 0.04086163081228733, + "grad_norm": 0.45923319458961487, + "learning_rate": 2.9947462045433594e-05, + "loss": 0.1104, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028071087412536146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014368514996021986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07051068954169751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9860957384109497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4520, + "weighted_flow_matching_loss": 0.04086163081228733, + "weighted_rkd_loss": 0.06953031122684479 + }, + { + "epoch": 1.318009892347978, + "flow_matching_loss": 0.04295903891324997, + "grad_norm": 0.4854811429977417, + "learning_rate": 2.9946768435268045e-05, + "loss": 0.1146, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028838179260492324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014977999310940505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0726543590426445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.98603435754776, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4530, + "weighted_flow_matching_loss": 0.04295903891324997, + "weighted_rkd_loss": 0.07163961715996266 + }, + { + "epoch": 1.3209194064591214, + "flow_matching_loss": 0.03891957346349954, + "grad_norm": 0.4800679683685303, + "learning_rate": 2.9946070284668884e-05, + "loss": 0.1055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026887090504169465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013755768816918135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06752995066344739, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9859728455543518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4540, + "weighted_flow_matching_loss": 0.03891957346349954, + "weighted_rkd_loss": 0.0665826540440321 + }, + { + "epoch": 1.3238289205702647, + "flow_matching_loss": 0.04052949510514736, + "grad_norm": 0.4137799143791199, + "learning_rate": 2.994536759384821e-05, + "loss": 0.1091, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027704453468322753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014097627718001604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0695065338164568, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9859112083911896, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4550, + "weighted_flow_matching_loss": 0.04052949510514736, + "weighted_rkd_loss": 0.06852730214595795 + }, + { + "epoch": 1.3267384346814082, + "flow_matching_loss": 0.04076945558190346, + "grad_norm": 0.48877280950546265, + "learning_rate": 2.994466036301946e-05, + "loss": 0.1086, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02739419974386692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014011024218052625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06879942417144776, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9858494281768799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4560, + "weighted_flow_matching_loss": 0.04076945558190346, + "weighted_rkd_loss": 0.06782585829496383 + }, + { + "epoch": 1.3296479487925517, + "flow_matching_loss": 0.041454768553376195, + "grad_norm": 0.5112590193748474, + "learning_rate": 2.9943948592397494e-05, + "loss": 0.1054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025605504959821702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013649093825370074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06486010439693927, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9857875108718872, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4570, + "weighted_flow_matching_loss": 0.041454768553376195, + "weighted_rkd_loss": 0.06393823474645614 + }, + { + "epoch": 1.3325574629036951, + "flow_matching_loss": 0.04097128100693226, + "grad_norm": 0.4572969377040863, + "learning_rate": 2.994323228219851e-05, + "loss": 0.1089, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027447852492332458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014042982086539268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06893868707120418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9857254743576049, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4580, + "weighted_flow_matching_loss": 0.04097128100693226, + "weighted_rkd_loss": 0.06795461103320122 + }, + { + "epoch": 1.3354669770148386, + "flow_matching_loss": 0.0398044690489769, + "grad_norm": 0.4520373046398163, + "learning_rate": 2.9942511432640113e-05, + "loss": 0.1071, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02717262078076601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013971641007810832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06831688284873963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9856632888317108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4590, + "weighted_flow_matching_loss": 0.0398044690489769, + "weighted_rkd_loss": 0.06733739376068115 + }, + { + "epoch": 1.338376491125982, + "flow_matching_loss": 0.04133272469043732, + "grad_norm": 0.4329788088798523, + "learning_rate": 2.9941786043941278e-05, + "loss": 0.1087, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02728904169052839, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0138183300383389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06839641332626342, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9856009840965271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4600, + "weighted_flow_matching_loss": 0.04133272469043732, + "weighted_rkd_loss": 0.0674116149544716 + }, + { + "epoch": 1.3412860052371254, + "flow_matching_loss": 0.041466443240642546, + "grad_norm": 0.4292128384113312, + "learning_rate": 2.9941056116322357e-05, + "loss": 0.1057, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025957281328737737, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013227002136409283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06514156423509121, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9855385422706604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4610, + "weighted_flow_matching_loss": 0.041466443240642546, + "weighted_rkd_loss": 0.06419958546757698 + }, + { + "epoch": 1.3441955193482689, + "flow_matching_loss": 0.04135647229850292, + "grad_norm": 0.39206042885780334, + "learning_rate": 2.994032165000508e-05, + "loss": 0.1109, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02804343234747648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014434822928160429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07052168659865857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9854759693145752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4620, + "weighted_flow_matching_loss": 0.04135647229850292, + "weighted_rkd_loss": 0.06949745453894138 + }, + { + "epoch": 1.3471050334594123, + "flow_matching_loss": 0.037559503130614756, + "grad_norm": 0.4498572051525116, + "learning_rate": 2.9939582645212566e-05, + "loss": 0.1025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026198693178594114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013534838054329156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06593222618103027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9854132533073425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4630, + "weighted_flow_matching_loss": 0.037559503130614756, + "weighted_rkd_loss": 0.0649705033749342 + }, + { + "epoch": 1.3500145475705558, + "flow_matching_loss": 0.04246874488890171, + "grad_norm": 0.409853994846344, + "learning_rate": 2.9938839102169303e-05, + "loss": 0.1085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02681354098021984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013421417586505413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06704849973320962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9853504121303558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4640, + "weighted_flow_matching_loss": 0.04246874488890171, + "weighted_rkd_loss": 0.0660662218928337 + }, + { + "epoch": 1.352924061681699, + "flow_matching_loss": 0.03827174492180348, + "grad_norm": 0.4265771210193634, + "learning_rate": 2.993809102110116e-05, + "loss": 0.1043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026645805314183237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01373158311471343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06702319271862507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9852874398231506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4650, + "weighted_flow_matching_loss": 0.03827174492180348, + "weighted_rkd_loss": 0.06603714562952519 + }, + { + "epoch": 1.3558335757928426, + "flow_matching_loss": 0.03881617709994316, + "grad_norm": 0.4072340726852417, + "learning_rate": 2.9937338402235373e-05, + "loss": 0.104, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026326931826770306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01345862178131938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06611248590052128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9852243065834045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4660, + "weighted_flow_matching_loss": 0.03881617709994316, + "weighted_rkd_loss": 0.06513563990592956 + }, + { + "epoch": 1.358743089903986, + "flow_matching_loss": 0.04136151298880577, + "grad_norm": 0.4327808916568756, + "learning_rate": 2.993658124580058e-05, + "loss": 0.1112, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028060470521450043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014733490068465471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0708544310182333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9851610898971558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4670, + "weighted_flow_matching_loss": 0.04136151298880577, + "weighted_rkd_loss": 0.06980303525924683 + }, + { + "epoch": 1.3616526040151296, + "flow_matching_loss": 0.039094870537519456, + "grad_norm": 0.4557742774486542, + "learning_rate": 2.9935819552026785e-05, + "loss": 0.1105, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028822448663413526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01487146271392703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07251636087894439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9850977063179016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4680, + "weighted_flow_matching_loss": 0.039094870537519456, + "weighted_rkd_loss": 0.0714357241988182 + }, + { + "epoch": 1.364562118126273, + "flow_matching_loss": 0.03837416358292103, + "grad_norm": 0.43929725885391235, + "learning_rate": 2.9935053321145368e-05, + "loss": 0.1052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026941587589681147, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013988150097429753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06787132546305656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9850341975688934, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4690, + "weighted_flow_matching_loss": 0.03837416358292103, + "weighted_rkd_loss": 0.06685548052191734 + }, + { + "epoch": 1.3674716322374163, + "flow_matching_loss": 0.04343388862907886, + "grad_norm": 0.4563772976398468, + "learning_rate": 2.9934282553389088e-05, + "loss": 0.1093, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026734127104282378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013447730615735055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0669159848242998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9849705696105957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4700, + "weighted_flow_matching_loss": 0.04343388862907886, + "weighted_rkd_loss": 0.06591027006506919 + }, + { + "epoch": 1.3703811463485598, + "flow_matching_loss": 0.03951000925153494, + "grad_norm": 0.37235066294670105, + "learning_rate": 2.9933507248992084e-05, + "loss": 0.1032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025906560011208057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012827602121978999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06464072242379189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.984906792640686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4710, + "weighted_flow_matching_loss": 0.03951000925153494, + "weighted_rkd_loss": 0.06366510279476642 + }, + { + "epoch": 1.3732906604597033, + "flow_matching_loss": 0.0375518623739481, + "grad_norm": 0.44291627407073975, + "learning_rate": 2.9932727408189876e-05, + "loss": 0.1077, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028278872929513456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014682894106954336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07124064043164254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9848428785800933, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4720, + "weighted_flow_matching_loss": 0.0375518623739481, + "weighted_rkd_loss": 0.07016084454953671 + }, + { + "epoch": 1.3762001745708465, + "flow_matching_loss": 0.039480778574943545, + "grad_norm": 0.4159018397331238, + "learning_rate": 2.9931943031219356e-05, + "loss": 0.1029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02596455905586481, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012471696641296149, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06440081521868705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9847788393497467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4730, + "weighted_flow_matching_loss": 0.039480778574943545, + "weighted_rkd_loss": 0.06342057064175606 + }, + { + "epoch": 1.3791096886819902, + "flow_matching_loss": 0.03996344730257988, + "grad_norm": 0.3824985921382904, + "learning_rate": 2.9931154118318803e-05, + "loss": 0.1061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026720623858273028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013764702342450618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06720595024526119, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9847146689891815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4740, + "weighted_flow_matching_loss": 0.03996344730257988, + "weighted_rkd_loss": 0.06617872193455696 + }, + { + "epoch": 1.3820192027931335, + "flow_matching_loss": 0.03823336064815521, + "grad_norm": 0.3596716523170471, + "learning_rate": 2.9930360669727858e-05, + "loss": 0.1002, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02514624074101448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01263132570311427, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06292380727827548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9846503734588623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4750, + "weighted_flow_matching_loss": 0.03823336064815521, + "weighted_rkd_loss": 0.06195801831781864 + }, + { + "epoch": 1.384928716904277, + "flow_matching_loss": 0.040426566079258916, + "grad_norm": 0.4376072883605957, + "learning_rate": 2.992956268568755e-05, + "loss": 0.1024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025267177633941172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01240742700174451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06294178292155266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9845859229564666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4760, + "weighted_flow_matching_loss": 0.040426566079258916, + "weighted_rkd_loss": 0.061971528083086015 + }, + { + "epoch": 1.3878382310154205, + "flow_matching_loss": 0.03988154679536819, + "grad_norm": 0.5496772527694702, + "learning_rate": 2.99287601664403e-05, + "loss": 0.1049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026235096715390684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013563226256519557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06603341922163963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9845213532447815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4770, + "weighted_flow_matching_loss": 0.03988154679536819, + "weighted_rkd_loss": 0.06501128748059273 + }, + { + "epoch": 1.3907477451265637, + "flow_matching_loss": 0.040035593509674075, + "grad_norm": 0.4650663137435913, + "learning_rate": 2.992795311222988e-05, + "loss": 0.1061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02684625834226608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013422523159533739, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06711503937840461, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9844566524028778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4780, + "weighted_flow_matching_loss": 0.040035593509674075, + "weighted_rkd_loss": 0.06607181243598462 + }, + { + "epoch": 1.3936572592377072, + "flow_matching_loss": 0.04110653102397919, + "grad_norm": 0.4546741545200348, + "learning_rate": 2.9927141523301453e-05, + "loss": 0.1045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02550454568117857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013356122840195894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06436521336436271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9843918144702911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4790, + "weighted_flow_matching_loss": 0.04110653102397919, + "weighted_rkd_loss": 0.06336060576140881 + }, + { + "epoch": 1.3965667733488507, + "flow_matching_loss": 0.038567556999623775, + "grad_norm": 0.43216562271118164, + "learning_rate": 2.992632539990156e-05, + "loss": 0.1, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024953257106244565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012533063255250453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0624395776540041, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9843268454074859, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4800, + "weighted_flow_matching_loss": 0.038567556999623775, + "weighted_rkd_loss": 0.06146101653575897 + }, + { + "epoch": 1.3994762874599942, + "flow_matching_loss": 0.03886025100946426, + "grad_norm": 0.4616755247116089, + "learning_rate": 2.9925504742278117e-05, + "loss": 0.1048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026552744396030902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013855868857353926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0669613566249609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9842617452144623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4810, + "weighted_flow_matching_loss": 0.03886025100946426, + "weighted_rkd_loss": 0.06590750403702259 + }, + { + "epoch": 1.4023858015711377, + "flow_matching_loss": 0.03916996736079455, + "grad_norm": 0.47149035334587097, + "learning_rate": 2.9924679550680416e-05, + "loss": 0.1047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02672057244926691, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013185685221105813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06662682965397834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9841965079307556, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4820, + "weighted_flow_matching_loss": 0.03916996736079455, + "weighted_rkd_loss": 0.06557396911084652 + }, + { + "epoch": 1.405295315682281, + "flow_matching_loss": 0.037647131644189355, + "grad_norm": 0.42603933811187744, + "learning_rate": 2.9923849825359133e-05, + "loss": 0.1007, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025706953555345535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012636322900652886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06405023001134395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9841311395168304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4830, + "weighted_flow_matching_loss": 0.037647131644189355, + "weighted_rkd_loss": 0.06303381398320199 + }, + { + "epoch": 1.4082048297934244, + "flow_matching_loss": 0.04187978357076645, + "grad_norm": 0.47731995582580566, + "learning_rate": 2.9923015566566317e-05, + "loss": 0.1069, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02619638666510582, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013654941506683827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06604771502315998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9840656518936157, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4840, + "weighted_flow_matching_loss": 0.04187978357076645, + "weighted_rkd_loss": 0.06499530263245105 + }, + { + "epoch": 1.411114343904568, + "flow_matching_loss": 0.039414003118872644, + "grad_norm": 0.42546018958091736, + "learning_rate": 2.992217677455539e-05, + "loss": 0.106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027086144499480724, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01350749284029007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06767978146672249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9840000033378601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4850, + "weighted_flow_matching_loss": 0.039414003118872644, + "weighted_rkd_loss": 0.06659688390791416 + }, + { + "epoch": 1.4140238580157114, + "flow_matching_loss": 0.042620157822966576, + "grad_norm": 0.4686235189437866, + "learning_rate": 2.9921333449581153e-05, + "loss": 0.1066, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02590783629566431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013164776749908924, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06498044915497303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9839342415332795, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4860, + "weighted_flow_matching_loss": 0.042620157822966576, + "weighted_rkd_loss": 0.063936485350132 + }, + { + "epoch": 1.4169333721268549, + "flow_matching_loss": 0.03685944303870201, + "grad_norm": 0.4050063490867615, + "learning_rate": 2.9920485591899795e-05, + "loss": 0.1004, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025714655593037605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013186510931700469, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06461582221090793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9838683426380157, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4870, + "weighted_flow_matching_loss": 0.03685944303870201, + "weighted_rkd_loss": 0.06357339806854725 + }, + { + "epoch": 1.4198428862379981, + "flow_matching_loss": 0.03904719054698944, + "grad_norm": 0.4272739291191101, + "learning_rate": 2.991963320176886e-05, + "loss": 0.1043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026537316292524336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01326729441061616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06634192653000355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9838023126125336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4880, + "weighted_flow_matching_loss": 0.03904719054698944, + "weighted_rkd_loss": 0.06526742577552795 + }, + { + "epoch": 1.4227524003491416, + "flow_matching_loss": 0.04001774415373802, + "grad_norm": 0.46247681975364685, + "learning_rate": 2.991877627944729e-05, + "loss": 0.1062, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02680060304701328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013725867681205272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0673270732164383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9837361514568329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4890, + "weighted_flow_matching_loss": 0.04001774415373802, + "weighted_rkd_loss": 0.06623211614787579 + }, + { + "epoch": 1.4256619144602851, + "flow_matching_loss": 0.041063864901661876, + "grad_norm": 0.44328993558883667, + "learning_rate": 2.9917914825195393e-05, + "loss": 0.1106, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.028053283505141736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014562540780752897, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.07066910862922668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9836698472499847, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4900, + "weighted_flow_matching_loss": 0.041063864901661876, + "weighted_rkd_loss": 0.06951502300798892 + }, + { + "epoch": 1.4285714285714286, + "flow_matching_loss": 0.0408224456012249, + "grad_norm": 0.49067190289497375, + "learning_rate": 2.991704883927486e-05, + "loss": 0.1069, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026823862455785275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013510931748896837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06715865731239319, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9836034178733826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4910, + "weighted_flow_matching_loss": 0.0408224456012249, + "weighted_rkd_loss": 0.06605741754174232 + }, + { + "epoch": 1.431480942682572, + "flow_matching_loss": 0.03813911490142345, + "grad_norm": 0.4630507826805115, + "learning_rate": 2.991617832194875e-05, + "loss": 0.1025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025931907817721367, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013624184019863605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06548799946904182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9835368514060974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4920, + "weighted_flow_matching_loss": 0.03813911490142345, + "weighted_rkd_loss": 0.06440986953675747 + }, + { + "epoch": 1.4343904567937154, + "flow_matching_loss": 0.04210890308022499, + "grad_norm": 0.37804946303367615, + "learning_rate": 2.9915303273481504e-05, + "loss": 0.1048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025407460145652296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012902684509754181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06371760442852974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9834701418876648, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4930, + "weighted_flow_matching_loss": 0.04210890308022499, + "weighted_rkd_loss": 0.06266433261334896 + }, + { + "epoch": 1.4372999709048588, + "flow_matching_loss": 0.03836460169404745, + "grad_norm": 0.4211759865283966, + "learning_rate": 2.9914423694138942e-05, + "loss": 0.1035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02647423166781664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013273945916444064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06622240953147411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9834033370018005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4940, + "weighted_flow_matching_loss": 0.03836460169404745, + "weighted_rkd_loss": 0.06512336172163487 + }, + { + "epoch": 1.4402094850160023, + "flow_matching_loss": 0.037355498597025874, + "grad_norm": 0.4406851530075073, + "learning_rate": 2.9913539584188253e-05, + "loss": 0.1039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026682006567716597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.014288565423339606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06765257827937603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9833363831043244, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4950, + "weighted_flow_matching_loss": 0.037355498597025874, + "weighted_rkd_loss": 0.06652528531849385 + }, + { + "epoch": 1.4431189991271458, + "flow_matching_loss": 0.03961227722465992, + "grad_norm": 0.4035019874572754, + "learning_rate": 2.9912650943898008e-05, + "loss": 0.1004, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024579704366624357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01261345250532031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06177286170423031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9832692861557006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4960, + "weighted_flow_matching_loss": 0.03961227722465992, + "weighted_rkd_loss": 0.060739366710186 + }, + { + "epoch": 1.4460285132382893, + "flow_matching_loss": 0.03720871973782778, + "grad_norm": 0.5231278538703918, + "learning_rate": 2.9911757773538148e-05, + "loss": 0.1018, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026341461576521397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012995444424450397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06567836701869964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9832020580768586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4970, + "weighted_flow_matching_loss": 0.03720871973782778, + "weighted_rkd_loss": 0.06457497961819172 + }, + { + "epoch": 1.4489380273494326, + "flow_matching_loss": 0.039245129004120825, + "grad_norm": 0.41985610127449036, + "learning_rate": 2.9910860073380005e-05, + "loss": 0.101, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025008610263466836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012749699503183365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06276692003011704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9831346988677978, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4980, + "weighted_flow_matching_loss": 0.039245129004120825, + "weighted_rkd_loss": 0.06170830577611923 + }, + { + "epoch": 1.451847541460576, + "flow_matching_loss": 0.03888173662126064, + "grad_norm": 0.4283619225025177, + "learning_rate": 2.9909957843696267e-05, + "loss": 0.1023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025701947323977948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013079081196337938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06448297612369061, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9830672144889832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 4990, + "weighted_flow_matching_loss": 0.03888173662126064, + "weighted_rkd_loss": 0.06339113861322403 + }, + { + "epoch": 1.4547570555717195, + "flow_matching_loss": 0.03679296914488077, + "grad_norm": 0.4620370864868164, + "learning_rate": 2.990905108476101e-05, + "loss": 0.0972, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024742061831057073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011975339706987142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061459462717175485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.982999587059021, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5000, + "weighted_flow_matching_loss": 0.03679296914488077, + "weighted_rkd_loss": 0.060414588823914525 + }, + { + "epoch": 1.457666569682863, + "flow_matching_loss": 0.040688663721084595, + "grad_norm": 0.46285924315452576, + "learning_rate": 2.9908139796849683e-05, + "loss": 0.1004, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02414251081645489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012430767994374037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06071578934788704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9829318344593048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5010, + "weighted_flow_matching_loss": 0.040688663721084595, + "weighted_rkd_loss": 0.05967944152653217 + }, + { + "epoch": 1.4605760837940065, + "flow_matching_loss": 0.03880066331475973, + "grad_norm": 0.4988221526145935, + "learning_rate": 2.9907223980239114e-05, + "loss": 0.1048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026856032013893128, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013479731883853674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06719179600477218, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9828639626502991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5020, + "weighted_flow_matching_loss": 0.03880066331475973, + "weighted_rkd_loss": 0.06604033820331097 + }, + { + "epoch": 1.4634855979051498, + "flow_matching_loss": 0.03724528029561043, + "grad_norm": 0.4256477952003479, + "learning_rate": 2.99063036352075e-05, + "loss": 0.0979, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024730156175792217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012220188230276107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06168050020933151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9827959299087524, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5030, + "weighted_flow_matching_loss": 0.03724528029561043, + "weighted_rkd_loss": 0.06061935313045978 + }, + { + "epoch": 1.4663951120162932, + "flow_matching_loss": 0.03818691372871399, + "grad_norm": 0.40557265281677246, + "learning_rate": 2.9905378762034424e-05, + "loss": 0.1053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.027429739013314247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013401541765779257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06826101914048195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9827277839183808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5040, + "weighted_flow_matching_loss": 0.03818691372871399, + "weighted_rkd_loss": 0.06708204634487629 + }, + { + "epoch": 1.4693046261274367, + "flow_matching_loss": 0.039044552855193614, + "grad_norm": 0.5059331059455872, + "learning_rate": 2.9904449361000833e-05, + "loss": 0.1028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02557769138365984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013752916362136602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06490829847753048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9826594948768616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5050, + "weighted_flow_matching_loss": 0.039044552855193614, + "weighted_rkd_loss": 0.06378269158303737 + }, + { + "epoch": 1.4722141402385802, + "flow_matching_loss": 0.038382695987820625, + "grad_norm": 0.48256978392601013, + "learning_rate": 2.9903515432389056e-05, + "loss": 0.1033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.026089817099273205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013927699532359838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06610733345150947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9825910866260529, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5060, + "weighted_flow_matching_loss": 0.038382695987820625, + "weighted_rkd_loss": 0.06495642103254795 + }, + { + "epoch": 1.4751236543497237, + "flow_matching_loss": 0.037533212825655936, + "grad_norm": 0.4667920768260956, + "learning_rate": 2.990257697648279e-05, + "loss": 0.0999, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025289548560976982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012865993566811085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06344509199261665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9825225353240967, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5070, + "weighted_flow_matching_loss": 0.037533212825655936, + "weighted_rkd_loss": 0.062336158752441403 + }, + { + "epoch": 1.478033168460867, + "flow_matching_loss": 0.03989128768444061, + "grad_norm": 0.5042613744735718, + "learning_rate": 2.9901633993567125e-05, + "loss": 0.104, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025974439270794392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013307216763496398, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06525609605014324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.982453852891922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5080, + "weighted_flow_matching_loss": 0.03989128768444061, + "weighted_rkd_loss": 0.06411106362938881 + }, + { + "epoch": 1.4809426825720104, + "flow_matching_loss": 0.03768734112381935, + "grad_norm": 0.40050598978996277, + "learning_rate": 2.99006864839285e-05, + "loss": 0.0994, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02512609977275133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01257230043411255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06282449997961521, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9823850333690644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5090, + "weighted_flow_matching_loss": 0.03768734112381935, + "weighted_rkd_loss": 0.0617179099470377 + }, + { + "epoch": 1.483852196683154, + "flow_matching_loss": 0.04028908740729094, + "grad_norm": 0.4460805654525757, + "learning_rate": 2.989973444785476e-05, + "loss": 0.1022, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02504252940416336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012976690474897624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06306174993515015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9823160886764526, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5100, + "weighted_flow_matching_loss": 0.04028908740729094, + "weighted_rkd_loss": 0.061946555972099304 + }, + { + "epoch": 1.4867617107942974, + "flow_matching_loss": 0.03799768146127462, + "grad_norm": 0.4471319615840912, + "learning_rate": 2.989877788563509e-05, + "loss": 0.0978, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024378984235227108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01216724030673504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060925208032131195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.982246994972229, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5110, + "weighted_flow_matching_loss": 0.03799768146127462, + "weighted_rkd_loss": 0.0598436251282692 + }, + { + "epoch": 1.489671224905441, + "flow_matching_loss": 0.039705405756831166, + "grad_norm": 0.4083530008792877, + "learning_rate": 2.9897816797560085e-05, + "loss": 0.1025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025502632185816766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012945177406072617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06395044066011905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9821778059005737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5120, + "weighted_flow_matching_loss": 0.039705405756831166, + "weighted_rkd_loss": 0.06281070671975612 + }, + { + "epoch": 1.4925807390165842, + "flow_matching_loss": 0.037256214395165445, + "grad_norm": 0.4931636452674866, + "learning_rate": 2.989685118392169e-05, + "loss": 0.0998, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025215070880949497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013293280266225338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06372342184185982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9821084678173065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5130, + "weighted_flow_matching_loss": 0.037256214395165445, + "weighted_rkd_loss": 0.0625832837074995 + }, + { + "epoch": 1.4954902531277277, + "flow_matching_loss": 0.038626445829868315, + "grad_norm": 0.4622655212879181, + "learning_rate": 2.9895881045013232e-05, + "loss": 0.0976, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023837588727474213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012391215097159148, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0600663922727108, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9820389926433564, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5140, + "weighted_flow_matching_loss": 0.038626445829868315, + "weighted_rkd_loss": 0.05898752808570862 + }, + { + "epoch": 1.4983997672388711, + "flow_matching_loss": 0.0378149913623929, + "grad_norm": 0.38545170426368713, + "learning_rate": 2.9894906381129414e-05, + "loss": 0.1006, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025705932453274728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012515522167086601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.063927386328578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9819693803787232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5150, + "weighted_flow_matching_loss": 0.0378149913623929, + "weighted_rkd_loss": 0.06277473717927932 + }, + { + "epoch": 1.5013092813500144, + "flow_matching_loss": 0.036994903907179834, + "grad_norm": 0.48021233081817627, + "learning_rate": 2.9893927192566316e-05, + "loss": 0.1004, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025678676925599574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.013183191139250993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06454054526984691, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9818996369838715, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5160, + "weighted_flow_matching_loss": 0.036994903907179834, + "weighted_rkd_loss": 0.06337228938937187 + }, + { + "epoch": 1.504218795461158, + "flow_matching_loss": 0.038667124882340434, + "grad_norm": 0.4305771291255951, + "learning_rate": 2.989294347962139e-05, + "loss": 0.0994, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024605037830770015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012691394239664078, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06190146990120411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9818297743797302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5170, + "weighted_flow_matching_loss": 0.038667124882340434, + "weighted_rkd_loss": 0.060776736959815024 + }, + { + "epoch": 1.5071283095723014, + "flow_matching_loss": 0.03762441016733646, + "grad_norm": 0.43619170784950256, + "learning_rate": 2.989195524259346e-05, + "loss": 0.1006, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025654651597142218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01284049078822136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0641497939825058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.981759774684906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5180, + "weighted_flow_matching_loss": 0.03762441016733646, + "weighted_rkd_loss": 0.06297967620193959 + }, + { + "epoch": 1.5100378236834449, + "flow_matching_loss": 0.03680469393730164, + "grad_norm": 0.48322251439094543, + "learning_rate": 2.9890962481782723e-05, + "loss": 0.0977, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024799341149628163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012443962413817644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06204264499247074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9816896378993988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5190, + "weighted_flow_matching_loss": 0.03680469393730164, + "weighted_rkd_loss": 0.060906555876135825 + }, + { + "epoch": 1.5129473377945883, + "flow_matching_loss": 0.03745276369154453, + "grad_norm": 0.46845874190330505, + "learning_rate": 2.9889965197490757e-05, + "loss": 0.097, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02428786475211382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012098416686058044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060674145445227624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9816193580627441, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5200, + "weighted_flow_matching_loss": 0.03745276369154453, + "weighted_rkd_loss": 0.05955891385674476 + }, + { + "epoch": 1.5158568519057316, + "flow_matching_loss": 0.040203763917088506, + "grad_norm": 0.4100789427757263, + "learning_rate": 2.988896339002052e-05, + "loss": 0.1036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.025592600740492345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01342218928039074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06460739150643349, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9815489768981933, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5210, + "weighted_flow_matching_loss": 0.040203763917088506, + "weighted_rkd_loss": 0.06341534852981567 + }, + { + "epoch": 1.5187663660168753, + "flow_matching_loss": 0.040494243055582045, + "grad_norm": 0.3835209012031555, + "learning_rate": 2.9887957059676315e-05, + "loss": 0.1014, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024950594641268253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012164094485342503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.062065284699201584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9814784467220307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5220, + "weighted_flow_matching_loss": 0.040494243055582045, + "weighted_rkd_loss": 0.0609157495200634 + }, + { + "epoch": 1.5216758801280186, + "flow_matching_loss": 0.0372025229036808, + "grad_norm": 0.4481339156627655, + "learning_rate": 2.9886946206763855e-05, + "loss": 0.0963, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024021228216588496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012163219135254621, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06020567454397678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.981407779455185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5230, + "weighted_flow_matching_loss": 0.0372025229036808, + "weighted_rkd_loss": 0.059086304157972336 + }, + { + "epoch": 1.524585394239162, + "flow_matching_loss": 0.038127701729536057, + "grad_norm": 0.4292542636394501, + "learning_rate": 2.9885930831590202e-05, + "loss": 0.0994, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024898134544491766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0126053087413311, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.062401578202843665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9813369750976563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5240, + "weighted_flow_matching_loss": 0.038127701729536057, + "weighted_rkd_loss": 0.061236971989274025 + }, + { + "epoch": 1.5274949083503055, + "flow_matching_loss": 0.038643788732588294, + "grad_norm": 0.3977719247341156, + "learning_rate": 2.98849109344638e-05, + "loss": 0.1006, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02528718952089548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012549947574734687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06312432698905468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9812660455703736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5250, + "weighted_flow_matching_loss": 0.038643788732588294, + "weighted_rkd_loss": 0.06194181963801384 + }, + { + "epoch": 1.5304044224614488, + "flow_matching_loss": 0.03774735033512115, + "grad_norm": 0.394496887922287, + "learning_rate": 2.9883886515694474e-05, + "loss": 0.0988, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0249767679721117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012245538365095853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06219907402992249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9811949908733368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5260, + "weighted_flow_matching_loss": 0.03774735033512115, + "weighted_rkd_loss": 0.06102945655584335 + }, + { + "epoch": 1.5333139365725925, + "flow_matching_loss": 0.03790297545492649, + "grad_norm": 0.47917625308036804, + "learning_rate": 2.988285757559341e-05, + "loss": 0.0976, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024305842630565167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012265745364129544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060877430438995364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9811238050460815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5270, + "weighted_flow_matching_loss": 0.03790297545492649, + "weighted_rkd_loss": 0.05972816348075867 + }, + { + "epoch": 1.5362234506837358, + "flow_matching_loss": 0.03984878435730934, + "grad_norm": 0.44771116971969604, + "learning_rate": 2.988182411447317e-05, + "loss": 0.0993, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024148341827094556, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012252321373671294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0605490043759346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9810524821281433, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5280, + "weighted_flow_matching_loss": 0.03984878435730934, + "weighted_rkd_loss": 0.059401804208755495 + }, + { + "epoch": 1.5391329647948793, + "flow_matching_loss": 0.039753039740025996, + "grad_norm": 0.3916635513305664, + "learning_rate": 2.98807861326477e-05, + "loss": 0.0992, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024126654118299486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012317868135869502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060571176931262016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.98098104596138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5290, + "weighted_flow_matching_loss": 0.039753039740025996, + "weighted_rkd_loss": 0.0594191238284111 + }, + { + "epoch": 1.5420424789060228, + "flow_matching_loss": 0.0365497849881649, + "grad_norm": 0.397889107465744, + "learning_rate": 2.9879743630432307e-05, + "loss": 0.0954, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024072807841002943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011898695770651103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06004431247711182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9809094429016113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5300, + "weighted_flow_matching_loss": 0.0365497849881649, + "weighted_rkd_loss": 0.05889810472726822 + }, + { + "epoch": 1.544951993017166, + "flow_matching_loss": 0.036635611206293106, + "grad_norm": 0.39773014187812805, + "learning_rate": 2.9878696608143677e-05, + "loss": 0.0956, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023853684775531292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012406249344348908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060113619267940524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9808377265930176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5310, + "weighted_flow_matching_loss": 0.036635611206293106, + "weighted_rkd_loss": 0.058961677551269534 + }, + { + "epoch": 1.5478615071283097, + "flow_matching_loss": 0.03824678547680378, + "grad_norm": 0.36657726764678955, + "learning_rate": 2.9877645066099862e-05, + "loss": 0.0975, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024047474935650827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012353042885661126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06044799238443375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9807658791542053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5320, + "weighted_flow_matching_loss": 0.03824678547680378, + "weighted_rkd_loss": 0.05928533524274826 + }, + { + "epoch": 1.550771021239453, + "flow_matching_loss": 0.03936916384845972, + "grad_norm": 0.44824305176734924, + "learning_rate": 2.9876589004620298e-05, + "loss": 0.0977, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02372058480978012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01203631004318595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05947747975587845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9806938946247101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5330, + "weighted_flow_matching_loss": 0.03936916384845972, + "weighted_rkd_loss": 0.05832922197878361 + }, + { + "epoch": 1.5536805353505965, + "flow_matching_loss": 0.038834993913769725, + "grad_norm": 0.42328977584838867, + "learning_rate": 2.987552842402579e-05, + "loss": 0.0948, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023045673593878747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010944809857755899, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05703615695238114, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9806217789649964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5340, + "weighted_flow_matching_loss": 0.038834993913769725, + "weighted_rkd_loss": 0.05593097023665905 + }, + { + "epoch": 1.55659004946174, + "flow_matching_loss": 0.03822529949247837, + "grad_norm": 0.41375255584716797, + "learning_rate": 2.9874463324638515e-05, + "loss": 0.093, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02241830453276634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01098050009459257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05581710934638977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9805495321750641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5350, + "weighted_flow_matching_loss": 0.03822529949247837, + "weighted_rkd_loss": 0.05473146215081215 + }, + { + "epoch": 1.5594995635728832, + "flow_matching_loss": 0.035648852773010734, + "grad_norm": 0.4589329957962036, + "learning_rate": 2.9873393706782016e-05, + "loss": 0.0957, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02455149181187153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012126588169485331, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06122957170009613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9804771542549133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5360, + "weighted_flow_matching_loss": 0.035648852773010734, + "weighted_rkd_loss": 0.06003409065306187 + }, + { + "epoch": 1.562409077684027, + "flow_matching_loss": 0.03889153525233269, + "grad_norm": 0.4426143169403076, + "learning_rate": 2.987231957078122e-05, + "loss": 0.0964, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023460125364363194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011701626051217318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.058621876314282416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9804046511650085, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5370, + "weighted_flow_matching_loss": 0.03889153525233269, + "weighted_rkd_loss": 0.05747315771877766 + }, + { + "epoch": 1.5653185917951702, + "flow_matching_loss": 0.03685066886246204, + "grad_norm": 0.4030165374279022, + "learning_rate": 2.9871240916962414e-05, + "loss": 0.0972, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024436791054904462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012654936499893665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061528518050909045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9803320169448853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5380, + "weighted_flow_matching_loss": 0.03685066886246204, + "weighted_rkd_loss": 0.06031843759119511 + }, + { + "epoch": 1.5682281059063137, + "flow_matching_loss": 0.037727971002459525, + "grad_norm": 0.4216347634792328, + "learning_rate": 2.9870157745653274e-05, + "loss": 0.0943, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022918199002742768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011906491871923209, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05774289034307003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.98025923371315, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5390, + "weighted_flow_matching_loss": 0.037727971002459525, + "weighted_rkd_loss": 0.05660303570330143 + }, + { + "epoch": 1.5711376200174572, + "flow_matching_loss": 0.03769857138395309, + "grad_norm": 0.4383501708507538, + "learning_rate": 2.9869070057182838e-05, + "loss": 0.094, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02298201620578766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011450037453323602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05741407014429569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9801863372325897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5400, + "weighted_flow_matching_loss": 0.03769857138395309, + "weighted_rkd_loss": 0.05627644062042236 + }, + { + "epoch": 1.5740471341286004, + "flow_matching_loss": 0.03728417679667473, + "grad_norm": 0.3822955787181854, + "learning_rate": 2.986797785188151e-05, + "loss": 0.0932, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022975297644734383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011076207645237446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0570268027484417, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.980113297700882, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5410, + "weighted_flow_matching_loss": 0.03728417679667473, + "weighted_rkd_loss": 0.05589275024831295 + }, + { + "epoch": 1.576956648239744, + "flow_matching_loss": 0.03609065320342779, + "grad_norm": 0.36715808510780334, + "learning_rate": 2.9866881130081074e-05, + "loss": 0.0931, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023411919735372065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011340500880032777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05816434025764465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9800401329994202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5420, + "weighted_flow_matching_loss": 0.03609065320342779, + "weighted_rkd_loss": 0.05700347721576691 + }, + { + "epoch": 1.5798661623508874, + "flow_matching_loss": 0.037938546016812326, + "grad_norm": 0.4082779288291931, + "learning_rate": 2.986577989211469e-05, + "loss": 0.094, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02281799092888832, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011525858752429485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05716184042394161, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9799668371677399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5430, + "weighted_flow_matching_loss": 0.037938546016812326, + "weighted_rkd_loss": 0.05601667501032352 + }, + { + "epoch": 1.5827756764620309, + "flow_matching_loss": 0.03643720671534538, + "grad_norm": 0.3664437532424927, + "learning_rate": 2.9864674138316887e-05, + "loss": 0.0965, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024518690444529055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012224481534212827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0612618625164032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9798934042453766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5440, + "weighted_flow_matching_loss": 0.03643720671534538, + "weighted_rkd_loss": 0.06003022044897079 + }, + { + "epoch": 1.5856851905731744, + "flow_matching_loss": 0.0393749188631773, + "grad_norm": 0.43108558654785156, + "learning_rate": 2.9863563869023554e-05, + "loss": 0.0993, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0245073726400733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0121216656640172, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06113641038537025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9798198342323303, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5450, + "weighted_flow_matching_loss": 0.0393749188631773, + "weighted_rkd_loss": 0.059902597591280936 + }, + { + "epoch": 1.5885947046843176, + "flow_matching_loss": 0.03779102396219969, + "grad_norm": 0.41169679164886475, + "learning_rate": 2.9862449084571972e-05, + "loss": 0.0966, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024039996415376665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01198167335242033, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06006166711449623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9797461450099945, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5460, + "weighted_flow_matching_loss": 0.03779102396219969, + "weighted_rkd_loss": 0.05884519144892693 + }, + { + "epoch": 1.5915042187954611, + "flow_matching_loss": 0.037828844785690305, + "grad_norm": 0.4722120761871338, + "learning_rate": 2.9861329785300774e-05, + "loss": 0.0948, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02330929283052683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011506149824708701, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05812473557889462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9796723127365112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5470, + "weighted_flow_matching_loss": 0.037828844785690305, + "weighted_rkd_loss": 0.056943224370479585 + }, + { + "epoch": 1.5944137329066046, + "flow_matching_loss": 0.03684357758611441, + "grad_norm": 0.5328090190887451, + "learning_rate": 2.9860205971549978e-05, + "loss": 0.0955, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023913415893912316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012006043829023838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05983287543058395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9795983672142029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5480, + "weighted_flow_matching_loss": 0.03684357758611441, + "weighted_rkd_loss": 0.058612025901675226 + }, + { + "epoch": 1.597323247017748, + "flow_matching_loss": 0.03797118365764618, + "grad_norm": 0.4341059625148773, + "learning_rate": 2.9859077643660974e-05, + "loss": 0.0953, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023320830054581167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011901525594294072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.058543185517191884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9795242726802826, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5490, + "weighted_flow_matching_loss": 0.03797118365764618, + "weighted_rkd_loss": 0.05734447836875915 + }, + { + "epoch": 1.6002327611288916, + "flow_matching_loss": 0.03662909977138042, + "grad_norm": 0.4062139093875885, + "learning_rate": 2.985794480197651e-05, + "loss": 0.0923, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022741270996630192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011358684580773114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056841226294636726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9794500589370727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5500, + "weighted_flow_matching_loss": 0.03662909977138042, + "weighted_rkd_loss": 0.055673114582896234 + }, + { + "epoch": 1.6031422752400348, + "flow_matching_loss": 0.03838112279772758, + "grad_norm": 0.4464319050312042, + "learning_rate": 2.9856807446840716e-05, + "loss": 0.0967, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023659176751971246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012230623234063386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.059548976644873616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.97937570810318, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5510, + "weighted_flow_matching_loss": 0.03838112279772758, + "weighted_rkd_loss": 0.05832092873752117 + }, + { + "epoch": 1.6060517893511783, + "flow_matching_loss": 0.03727037254720926, + "grad_norm": 0.4103446900844574, + "learning_rate": 2.9855665578599093e-05, + "loss": 0.096, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023980303853750228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01206009928137064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.06002070754766464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9793012320995331, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5520, + "weighted_flow_matching_loss": 0.03727037254720926, + "weighted_rkd_loss": 0.058778277784585956 + }, + { + "epoch": 1.6089613034623218, + "flow_matching_loss": 0.04066900797188282, + "grad_norm": 0.4046085476875305, + "learning_rate": 2.9854519197598504e-05, + "loss": 0.1, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02407106924802065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.012460892740637063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.060603031888604164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9792266070842743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5530, + "weighted_flow_matching_loss": 0.04066900797188282, + "weighted_rkd_loss": 0.05934420824050903 + }, + { + "epoch": 1.6118708175734653, + "flow_matching_loss": 0.03515064734965563, + "grad_norm": 0.3758999705314636, + "learning_rate": 2.9853368304187197e-05, + "loss": 0.0928, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023649968206882477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01160175520926714, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05890169180929661, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9791518688201905, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5540, + "weighted_flow_matching_loss": 0.03515064734965563, + "weighted_rkd_loss": 0.05767372585833073 + }, + { + "epoch": 1.6147803316846088, + "flow_matching_loss": 0.0373912762850523, + "grad_norm": 0.4624415934085846, + "learning_rate": 2.9852212898714783e-05, + "loss": 0.092, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022250082716345786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011244214233011007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05574437975883484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9790769875049591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5550, + "weighted_flow_matching_loss": 0.0373912762850523, + "weighted_rkd_loss": 0.0545779537409544 + }, + { + "epoch": 1.617689845795752, + "flow_matching_loss": 0.03777967244386673, + "grad_norm": 0.4266854524612427, + "learning_rate": 2.9851052981532233e-05, + "loss": 0.0943, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023311092890799044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01109158555045724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057713770866394044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9790019869804383, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5560, + "weighted_flow_matching_loss": 0.03777967244386673, + "weighted_rkd_loss": 0.05650190003216267 + }, + { + "epoch": 1.6205993599068955, + "flow_matching_loss": 0.037483400851488116, + "grad_norm": 0.375227153301239, + "learning_rate": 2.9849888552991908e-05, + "loss": 0.093, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022602795250713827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01153657902032137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0567421693354845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9789268434047699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5570, + "weighted_flow_matching_loss": 0.037483400851488116, + "weighted_rkd_loss": 0.055546394735574725 + }, + { + "epoch": 1.623508874018039, + "flow_matching_loss": 0.03655656967312097, + "grad_norm": 0.3826400339603424, + "learning_rate": 2.9848719613447535e-05, + "loss": 0.0925, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02296092323958874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01120041711255908, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057122263312339785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9788515746593476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5580, + "weighted_flow_matching_loss": 0.03655656967312097, + "weighted_rkd_loss": 0.0559141967445612 + }, + { + "epoch": 1.6264183881291823, + "flow_matching_loss": 0.03699306100606918, + "grad_norm": 0.4135676920413971, + "learning_rate": 2.9847546163254194e-05, + "loss": 0.0938, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0232479689642787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011510175839066505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05800611339509487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9787761688232421, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5590, + "weighted_flow_matching_loss": 0.03699306100606918, + "weighted_rkd_loss": 0.05677499808371067 + }, + { + "epoch": 1.629327902240326, + "flow_matching_loss": 0.03579703737050295, + "grad_norm": 0.47463735938072205, + "learning_rate": 2.984636820276836e-05, + "loss": 0.0927, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023165633156895637, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011804823763668537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.058136089518666265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9787006378173828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5600, + "weighted_flow_matching_loss": 0.03579703737050295, + "weighted_rkd_loss": 0.056897790357470515 + }, + { + "epoch": 1.6322374163514692, + "flow_matching_loss": 0.03485032431781292, + "grad_norm": 0.4381447434425354, + "learning_rate": 2.984518573234786e-05, + "loss": 0.0911, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023028936795890333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011425600107759237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05748347379267216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9786249697208405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5610, + "weighted_flow_matching_loss": 0.03485032431781292, + "weighted_rkd_loss": 0.056254716217517854 + }, + { + "epoch": 1.6351469304626127, + "flow_matching_loss": 0.037240192852914335, + "grad_norm": 0.4117930233478546, + "learning_rate": 2.98439987523519e-05, + "loss": 0.0944, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023369436897337437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011636110115796327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.058374984562397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9785491704940796, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5620, + "weighted_flow_matching_loss": 0.037240192852914335, + "weighted_rkd_loss": 0.0571228351444006 + }, + { + "epoch": 1.6380564445737562, + "flow_matching_loss": 0.0357122790068388, + "grad_norm": 0.3691185712814331, + "learning_rate": 2.9842807263141052e-05, + "loss": 0.0928, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0233423113822937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011618276219815015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05830289907753468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9784732460975647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5630, + "weighted_flow_matching_loss": 0.0357122790068388, + "weighted_rkd_loss": 0.05704783499240875 + }, + { + "epoch": 1.6409659586848995, + "flow_matching_loss": 0.03776847645640373, + "grad_norm": 0.4398343861103058, + "learning_rate": 2.9841611265077256e-05, + "loss": 0.0929, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02258529718965292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011178664863109589, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056349259614944455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9783971965312958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5640, + "weighted_flow_matching_loss": 0.03776847645640373, + "weighted_rkd_loss": 0.05513188131153583 + }, + { + "epoch": 1.6438754727960432, + "flow_matching_loss": 0.03715270534157753, + "grad_norm": 0.41913750767707825, + "learning_rate": 2.9840410758523832e-05, + "loss": 0.0915, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0222957257181406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010951480455696582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.055542932078242305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9783210098743439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5650, + "weighted_flow_matching_loss": 0.03715270534157753, + "weighted_rkd_loss": 0.05433876030147076 + }, + { + "epoch": 1.6467849869071864, + "flow_matching_loss": 0.03749148566275835, + "grad_norm": 0.3894643783569336, + "learning_rate": 2.9839205743845453e-05, + "loss": 0.0926, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0228111756965518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01069833030924201, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05632068142294884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9782446920871735, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5660, + "weighted_flow_matching_loss": 0.03749148566275835, + "weighted_rkd_loss": 0.055095480754971504 + }, + { + "epoch": 1.64969450101833, + "flow_matching_loss": 0.0376986563205719, + "grad_norm": 0.4031471908092499, + "learning_rate": 2.983799622140818e-05, + "loss": 0.0894, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02105775587260723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010710784746333957, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05282629653811455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9781682312488555, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5670, + "weighted_flow_matching_loss": 0.0376986563205719, + "weighted_rkd_loss": 0.05167299397289753 + }, + { + "epoch": 1.6526040151294734, + "flow_matching_loss": 0.03713206574320793, + "grad_norm": 0.42452287673950195, + "learning_rate": 2.9836782191579425e-05, + "loss": 0.094, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02323022112250328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011659289989620448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05811973176896572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9780916452407837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5680, + "weighted_flow_matching_loss": 0.03713206574320793, + "weighted_rkd_loss": 0.05684639513492584 + }, + { + "epoch": 1.6555135292406167, + "flow_matching_loss": 0.03538382388651371, + "grad_norm": 0.37493830919265747, + "learning_rate": 2.983556365472799e-05, + "loss": 0.0895, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02231658063828945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010710146371275186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.055343307182192804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9780149400234223, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5690, + "weighted_flow_matching_loss": 0.03538382388651371, + "weighted_rkd_loss": 0.05412655211985111 + }, + { + "epoch": 1.6584230433517604, + "flow_matching_loss": 0.03593739550560713, + "grad_norm": 0.3751412630081177, + "learning_rate": 2.9834340611224022e-05, + "loss": 0.0913, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022710893861949445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0111888337880373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056610621884465216, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9779380857944489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5700, + "weighted_flow_matching_loss": 0.03593739550560713, + "weighted_rkd_loss": 0.05536171607673168 + }, + { + "epoch": 1.6613325574629036, + "flow_matching_loss": 0.0361731231212616, + "grad_norm": 0.46776527166366577, + "learning_rate": 2.9833113061439057e-05, + "loss": 0.0925, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023187369853258134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011192838475108147, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05756757855415344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9778611123561859, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5710, + "weighted_flow_matching_loss": 0.0361731231212616, + "weighted_rkd_loss": 0.05629306174814701 + }, + { + "epoch": 1.6642420715740471, + "flow_matching_loss": 0.036257537081837654, + "grad_norm": 0.3976631760597229, + "learning_rate": 2.9831881005745986e-05, + "loss": 0.0881, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02131763715296984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010406923294067384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05304219722747803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9777840077877045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5720, + "weighted_flow_matching_loss": 0.036257537081837654, + "weighted_rkd_loss": 0.051863788068294524 + }, + { + "epoch": 1.6671515856851906, + "flow_matching_loss": 0.04034061953425407, + "grad_norm": 0.40524348616600037, + "learning_rate": 2.983064444451908e-05, + "loss": 0.0924, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021302003785967825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0106141971424222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053218204900622366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9777067720890045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5730, + "weighted_flow_matching_loss": 0.04034061953425407, + "weighted_rkd_loss": 0.052031799033284185 + }, + { + "epoch": 1.6700610997963339, + "flow_matching_loss": 0.03907028995454311, + "grad_norm": 0.4360905587673187, + "learning_rate": 2.9829403378133974e-05, + "loss": 0.0952, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023031186684966086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011382807604968547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05744518153369427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.977629405260086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5740, + "weighted_flow_matching_loss": 0.03907028995454311, + "weighted_rkd_loss": 0.056160058826208115 + }, + { + "epoch": 1.6729706139074776, + "flow_matching_loss": 0.036353734135627744, + "grad_norm": 0.45874541997909546, + "learning_rate": 2.9828157806967668e-05, + "loss": 0.0914, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022490431740880012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011372860986739397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056353725120425226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9775519073009491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5750, + "weighted_flow_matching_loss": 0.036353734135627744, + "weighted_rkd_loss": 0.05508869141340256 + }, + { + "epoch": 1.6758801280186209, + "flow_matching_loss": 0.039936267957091334, + "grad_norm": 0.36217859387397766, + "learning_rate": 2.9826907731398533e-05, + "loss": 0.0964, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023009010963141917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01171679412946105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057734815403819084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9774742722511292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5760, + "weighted_flow_matching_loss": 0.039936267957091334, + "weighted_rkd_loss": 0.056434324756264684 + }, + { + "epoch": 1.6787896421297643, + "flow_matching_loss": 0.03833948951214552, + "grad_norm": 0.3751000463962555, + "learning_rate": 2.9825653151806315e-05, + "loss": 0.0923, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022205008007586003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01081879660487175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05522881299257278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9773965120315552, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5770, + "weighted_flow_matching_loss": 0.03833948951214552, + "weighted_rkd_loss": 0.05398041121661663 + }, + { + "epoch": 1.6816991562409078, + "flow_matching_loss": 0.03996134325861931, + "grad_norm": 0.4207327663898468, + "learning_rate": 2.9824394068572114e-05, + "loss": 0.0939, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022342258505523203, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010501146130263806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05518566370010376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9773186206817627, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5780, + "weighted_flow_matching_loss": 0.03996134325861931, + "weighted_rkd_loss": 0.05393389351665974 + }, + { + "epoch": 1.684608670352051, + "flow_matching_loss": 0.036430385522544385, + "grad_norm": 0.39892542362213135, + "learning_rate": 2.9823130482078415e-05, + "loss": 0.0913, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02249682229012251, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011129209958016872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05612285509705543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9772405922412872, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5790, + "weighted_flow_matching_loss": 0.036430385522544385, + "weighted_rkd_loss": 0.0548455148935318 + }, + { + "epoch": 1.6875181844631948, + "flow_matching_loss": 0.03498385399580002, + "grad_norm": 0.39505499601364136, + "learning_rate": 2.9821862392709054e-05, + "loss": 0.0872, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021375369839370252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010705317463725805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05345605760812759, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9771624445915222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5800, + "weighted_flow_matching_loss": 0.03498385399580002, + "weighted_rkd_loss": 0.052235215157270434 + }, + { + "epoch": 1.690427698574338, + "flow_matching_loss": 0.03716672789305449, + "grad_norm": 0.4082459509372711, + "learning_rate": 2.982058980084925e-05, + "loss": 0.0949, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.023580172657966615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011881747376173734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05904209278523922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9770841538906098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5810, + "weighted_flow_matching_loss": 0.03716672789305449, + "weighted_rkd_loss": 0.05768914632499218 + }, + { + "epoch": 1.6933372126854815, + "flow_matching_loss": 0.03667423613369465, + "grad_norm": 0.5143720507621765, + "learning_rate": 2.9819312706885577e-05, + "loss": 0.0921, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022687271423637866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011324629466980696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056699172407388684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9770057439804077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5820, + "weighted_flow_matching_loss": 0.03667423613369465, + "weighted_rkd_loss": 0.05539533942937851 + }, + { + "epoch": 1.696246726796625, + "flow_matching_loss": 0.035062461905181405, + "grad_norm": 0.4177018702030182, + "learning_rate": 2.9818031111205986e-05, + "loss": 0.0896, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02251962833106518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01082799406722188, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05586725063621998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9769271969795227, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5830, + "weighted_flow_matching_loss": 0.035062461905181405, + "weighted_rkd_loss": 0.054578233510255814 + }, + { + "epoch": 1.6991562409077683, + "flow_matching_loss": 0.03752928245812655, + "grad_norm": 0.42697176337242126, + "learning_rate": 2.981674501419979e-05, + "loss": 0.0908, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021977921202778816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010572082176804543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05452792420983314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9768485248088836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5840, + "weighted_flow_matching_loss": 0.03752928245812655, + "weighted_rkd_loss": 0.05326553769409657 + }, + { + "epoch": 1.702065755018912, + "flow_matching_loss": 0.03692398816347122, + "grad_norm": 0.3683511018753052, + "learning_rate": 2.981545441625767e-05, + "loss": 0.0884, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021166450530290603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010341350641101598, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052674252167344095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9767697155475616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5850, + "weighted_flow_matching_loss": 0.03692398816347122, + "weighted_rkd_loss": 0.05145058184862137 + }, + { + "epoch": 1.7049752691300553, + "flow_matching_loss": 0.035530702024698255, + "grad_norm": 0.35556739568710327, + "learning_rate": 2.9814159317771682e-05, + "loss": 0.0954, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.024743451923131942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011796152032911778, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.061283056065440177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9766907632350922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5860, + "weighted_flow_matching_loss": 0.035530702024698255, + "weighted_rkd_loss": 0.059854619204998016 + }, + { + "epoch": 1.7078847832411987, + "flow_matching_loss": 0.03819911740720272, + "grad_norm": 0.37562277913093567, + "learning_rate": 2.9812859719135236e-05, + "loss": 0.0927, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022524680010974407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010796223487704992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05584558360278606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9766117036342621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5870, + "weighted_flow_matching_loss": 0.03819911740720272, + "weighted_rkd_loss": 0.05453943498432636 + }, + { + "epoch": 1.7107942973523422, + "flow_matching_loss": 0.03849171698093414, + "grad_norm": 0.44854211807250977, + "learning_rate": 2.981155562074312e-05, + "loss": 0.0938, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022745923325419425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011107041127979755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.056598888337612153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9765324890613556, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5880, + "weighted_flow_matching_loss": 0.03849171698093414, + "weighted_rkd_loss": 0.05527056977152824 + }, + { + "epoch": 1.7137038114634855, + "flow_matching_loss": 0.03832842409610748, + "grad_norm": 0.412710040807724, + "learning_rate": 2.9810247022991483e-05, + "loss": 0.0895, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021012228541076185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010343912430107594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052368369325995444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.976453173160553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5890, + "weighted_flow_matching_loss": 0.03832842409610748, + "weighted_rkd_loss": 0.051135281845927236 + }, + { + "epoch": 1.7166133255746292, + "flow_matching_loss": 0.03846380971372128, + "grad_norm": 0.41747644543647766, + "learning_rate": 2.9808933926277842e-05, + "loss": 0.0919, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02212189882993698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010520851984620095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.054764649271965025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9763737142086029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5900, + "weighted_flow_matching_loss": 0.03846380971372128, + "weighted_rkd_loss": 0.05347085632383823 + }, + { + "epoch": 1.7195228396857725, + "flow_matching_loss": 0.03709434662014246, + "grad_norm": 0.37919434905052185, + "learning_rate": 2.9807616331001078e-05, + "loss": 0.0923, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022492865845561028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011610216461122036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0565959483385086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9762941241264343, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5910, + "weighted_flow_matching_loss": 0.03709434662014246, + "weighted_rkd_loss": 0.05525427311658859 + }, + { + "epoch": 1.722432353796916, + "flow_matching_loss": 0.03668814245611429, + "grad_norm": 0.40567269921302795, + "learning_rate": 2.9806294237561452e-05, + "loss": 0.0911, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022459054179489613, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010802246630191803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05572035498917103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9762143909931182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5920, + "weighted_flow_matching_loss": 0.03668814245611429, + "weighted_rkd_loss": 0.05439498648047447 + }, + { + "epoch": 1.7253418679080594, + "flow_matching_loss": 0.03907383196055889, + "grad_norm": 0.5039904713630676, + "learning_rate": 2.980496764636057e-05, + "loss": 0.0918, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02151535591110587, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010998116713017226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05402882806956768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9761345386505127, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5930, + "weighted_flow_matching_loss": 0.03907383196055889, + "weighted_rkd_loss": 0.05273932069540024 + }, + { + "epoch": 1.7282513820192027, + "flow_matching_loss": 0.03866733871400356, + "grad_norm": 0.4149782061576843, + "learning_rate": 2.9803636557801417e-05, + "loss": 0.0895, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021080330573022367, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009940591407939792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05210125260055065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9760545551776886, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5940, + "weighted_flow_matching_loss": 0.03866733871400356, + "weighted_rkd_loss": 0.05085368156433105 + }, + { + "epoch": 1.7311608961303462, + "flow_matching_loss": 0.037264126166701314, + "grad_norm": 0.4070252478122711, + "learning_rate": 2.980230097228835e-05, + "loss": 0.0877, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020659624971449374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010319750104099513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05163900032639503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.975974440574646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5950, + "weighted_flow_matching_loss": 0.037264126166701314, + "weighted_rkd_loss": 0.05039832033216953 + }, + { + "epoch": 1.7340704102414897, + "flow_matching_loss": 0.03691701851785183, + "grad_norm": 0.38185736536979675, + "learning_rate": 2.9800960890227076e-05, + "loss": 0.0899, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021798978000879286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010643882770091295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05424183867871761, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9758942008018494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5960, + "weighted_flow_matching_loss": 0.03691701851785183, + "weighted_rkd_loss": 0.052934328094124794 + }, + { + "epoch": 1.7369799243526332, + "flow_matching_loss": 0.038545583561062816, + "grad_norm": 0.3724169433116913, + "learning_rate": 2.979961631202468e-05, + "loss": 0.091, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021551843360066415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010611898265779018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05371558517217636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9758138358592987, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5970, + "weighted_flow_matching_loss": 0.038545583561062816, + "weighted_rkd_loss": 0.05241644680500031 + }, + { + "epoch": 1.7398894384637766, + "flow_matching_loss": 0.035736244171857834, + "grad_norm": 0.36272382736206055, + "learning_rate": 2.9798267238089613e-05, + "loss": 0.0878, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021426220051944257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010500846430659294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05335328653454781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9757333278656006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5980, + "weighted_flow_matching_loss": 0.035736244171857834, + "weighted_rkd_loss": 0.05205857716500759 + }, + { + "epoch": 1.74279895257492, + "flow_matching_loss": 0.03676814008504152, + "grad_norm": 0.3936516344547272, + "learning_rate": 2.9796913668831684e-05, + "loss": 0.0895, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02168995626270771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010712461732327937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05409237444400787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9756526827812195, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 5990, + "weighted_flow_matching_loss": 0.03676814008504152, + "weighted_rkd_loss": 0.05277540385723114 + }, + { + "epoch": 1.7457084666860634, + "flow_matching_loss": 0.03546370714902878, + "grad_norm": 0.3783160150051117, + "learning_rate": 2.979555560466207e-05, + "loss": 0.0889, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021980854868888854, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010829890798777341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05479159951210022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9755719125270843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6000, + "weighted_flow_matching_loss": 0.03546370714902878, + "weighted_rkd_loss": 0.053453102707862854 + }, + { + "epoch": 1.7486179807972069, + "flow_matching_loss": 0.035827735997736455, + "grad_norm": 0.35940855741500854, + "learning_rate": 2.9794193045993317e-05, + "loss": 0.0868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02089379373937845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010434544924646617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05222213268280029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9754910230636596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6010, + "weighted_flow_matching_loss": 0.035827735997736455, + "weighted_rkd_loss": 0.050942272320389746 + }, + { + "epoch": 1.7515274949083504, + "flow_matching_loss": 0.037853668257594106, + "grad_norm": 0.3854762315750122, + "learning_rate": 2.9792825993239337e-05, + "loss": 0.09, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021442280896008015, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010534418234601616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05341897942125797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.975409984588623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6020, + "weighted_flow_matching_loss": 0.037853668257594106, + "weighted_rkd_loss": 0.052105285972356793 + }, + { + "epoch": 1.7544370090194938, + "flow_matching_loss": 0.035087612085044385, + "grad_norm": 0.36852049827575684, + "learning_rate": 2.9791454446815404e-05, + "loss": 0.0855, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02082800567150116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010013650450855494, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051669662445783617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9753288447856903, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6030, + "weighted_flow_matching_loss": 0.035087612085044385, + "weighted_rkd_loss": 0.05039496347308159 + }, + { + "epoch": 1.757346523130637, + "flow_matching_loss": 0.03617609292268753, + "grad_norm": 0.38739725947380066, + "learning_rate": 2.9790078407138154e-05, + "loss": 0.0939, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02372868452221155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011701289284974336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05915865860879421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9752475559711457, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6040, + "weighted_flow_matching_loss": 0.03617609292268753, + "weighted_rkd_loss": 0.05769433192908764 + }, + { + "epoch": 1.7602560372417806, + "flow_matching_loss": 0.038469836860895154, + "grad_norm": 0.4533136188983917, + "learning_rate": 2.9788697874625588e-05, + "loss": 0.0903, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02133137471973896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010443946532905101, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053106696158647534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9751661360263825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6050, + "weighted_flow_matching_loss": 0.038469836860895154, + "weighted_rkd_loss": 0.0517878282815218 + }, + { + "epoch": 1.763165551352924, + "flow_matching_loss": 0.03484505508095026, + "grad_norm": 0.4200778305530548, + "learning_rate": 2.9787312849697088e-05, + "loss": 0.0857, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020991743355989457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010146399214863776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05212988555431366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9750845730304718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6060, + "weighted_flow_matching_loss": 0.03484505508095026, + "weighted_rkd_loss": 0.05083104893565178 + }, + { + "epoch": 1.7660750654640673, + "flow_matching_loss": 0.03941704817116261, + "grad_norm": 0.427607923746109, + "learning_rate": 2.978592333277338e-05, + "loss": 0.0944, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02267814353108406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011000354029238223, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05635664090514183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9750029027462006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6070, + "weighted_flow_matching_loss": 0.03941704817116261, + "weighted_rkd_loss": 0.05494788512587547 + }, + { + "epoch": 1.768984579575211, + "flow_matching_loss": 0.03725051824003458, + "grad_norm": 0.362937867641449, + "learning_rate": 2.9784529324276557e-05, + "loss": 0.088, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020964011177420618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010150615498423577, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05207863822579384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9749211013317108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6080, + "weighted_flow_matching_loss": 0.03725051824003458, + "weighted_rkd_loss": 0.050772585347294806 + }, + { + "epoch": 1.7718940936863543, + "flow_matching_loss": 0.03666094951331615, + "grad_norm": 0.44038328528404236, + "learning_rate": 2.9783130824630093e-05, + "loss": 0.0903, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02196533903479576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011104946304112673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05503562428057194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.974839162826538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6090, + "weighted_flow_matching_loss": 0.03666094951331615, + "weighted_rkd_loss": 0.05365087203681469 + }, + { + "epoch": 1.7748036077974978, + "flow_matching_loss": 0.03708468824625015, + "grad_norm": 0.4499649107456207, + "learning_rate": 2.9781727834258806e-05, + "loss": 0.0908, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022111629694700242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010928972903639079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0551522321999073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9747570931911469, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6100, + "weighted_flow_matching_loss": 0.03708468824625015, + "weighted_rkd_loss": 0.05376003421843052 + }, + { + "epoch": 1.7777131219086413, + "flow_matching_loss": 0.033557116612792016, + "grad_norm": 0.41981878876686096, + "learning_rate": 2.97803203535889e-05, + "loss": 0.0894, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022953764349222184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.011354204453527927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.057261732220649716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9746748805046082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6110, + "weighted_flow_matching_loss": 0.033557116612792016, + "weighted_rkd_loss": 0.05581155009567738 + }, + { + "epoch": 1.7806226360197845, + "flow_matching_loss": 0.03789414092898369, + "grad_norm": 0.47421127557754517, + "learning_rate": 2.977890838304792e-05, + "loss": 0.0897, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021345651336014272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010458972305059433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05315027497708798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9745925664901733, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6120, + "weighted_flow_matching_loss": 0.03789414092898369, + "weighted_rkd_loss": 0.05179979577660561 + }, + { + "epoch": 1.7835321501309283, + "flow_matching_loss": 0.03558347541838884, + "grad_norm": 0.4735105633735657, + "learning_rate": 2.9777491923064782e-05, + "loss": 0.0878, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021372014097869395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010854591708630323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053598619624972345, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9745100975036621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6130, + "weighted_flow_matching_loss": 0.03558347541838884, + "weighted_rkd_loss": 0.05223235674202442 + }, + { + "epoch": 1.7864416642420715, + "flow_matching_loss": 0.039701738581061366, + "grad_norm": 0.42247602343559265, + "learning_rate": 2.977607097406978e-05, + "loss": 0.0939, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.022198880091309547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0111723393201828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05557009913027287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9744275152683258, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6140, + "weighted_flow_matching_loss": 0.039701738581061366, + "weighted_rkd_loss": 0.054149077460169795 + }, + { + "epoch": 1.789351178353215, + "flow_matching_loss": 0.034887865744531156, + "grad_norm": 0.3837222754955292, + "learning_rate": 2.9774645536494552e-05, + "loss": 0.0851, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020449652709066866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010640378762036561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05153968371450901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9743447959423065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6150, + "weighted_flow_matching_loss": 0.034887865744531156, + "weighted_rkd_loss": 0.0502174086868763 + }, + { + "epoch": 1.7922606924643585, + "flow_matching_loss": 0.0343173174187541, + "grad_norm": 0.3682175278663635, + "learning_rate": 2.9773215610772117e-05, + "loss": 0.086, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021213752590119837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010585167538374662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0530126728117466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9742619454860687, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6160, + "weighted_flow_matching_loss": 0.0343173174187541, + "weighted_rkd_loss": 0.051648151502013205 + }, + { + "epoch": 1.7951702065755017, + "flow_matching_loss": 0.03405263274908066, + "grad_norm": 0.368160218000412, + "learning_rate": 2.9771781197336837e-05, + "loss": 0.0818, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01977229956537485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009503898862749337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04904849827289581, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9741789698600769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6170, + "weighted_flow_matching_loss": 0.03405263274908066, + "weighted_rkd_loss": 0.047781992703676224 + }, + { + "epoch": 1.7980797206866455, + "flow_matching_loss": 0.035867098905146125, + "grad_norm": 0.42267584800720215, + "learning_rate": 2.977034229662446e-05, + "loss": 0.0858, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02057680394500494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010116531047970057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051270139217376706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9740958452224732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6180, + "weighted_flow_matching_loss": 0.035867098905146125, + "weighted_rkd_loss": 0.049941998347640036 + }, + { + "epoch": 1.8009892347977887, + "flow_matching_loss": 0.03352090734988451, + "grad_norm": 0.35690852999687195, + "learning_rate": 2.976889890907208e-05, + "loss": 0.0845, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021155700925737618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010010358598083257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05232176035642624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9740126132965088, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6190, + "weighted_flow_matching_loss": 0.03352090734988451, + "weighted_rkd_loss": 0.05096210017800331 + }, + { + "epoch": 1.8038987489089322, + "flow_matching_loss": 0.0361460754647851, + "grad_norm": 0.4201953113079071, + "learning_rate": 2.9767451035118164e-05, + "loss": 0.0865, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020768795907497407, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010194521490484476, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05173211321234703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973929226398468, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6200, + "weighted_flow_matching_loss": 0.0361460754647851, + "weighted_rkd_loss": 0.050383371859788896 + }, + { + "epoch": 1.8068082630200757, + "flow_matching_loss": 0.034450991451740264, + "grad_norm": 0.39978575706481934, + "learning_rate": 2.9765998675202536e-05, + "loss": 0.0843, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020621568150818347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009983297856524586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051226434484124186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9738457441329956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6210, + "weighted_flow_matching_loss": 0.034450991451740264, + "weighted_rkd_loss": 0.049886612594127654 + }, + { + "epoch": 1.809717777131219, + "flow_matching_loss": 0.03611145447939634, + "grad_norm": 0.39505335688591003, + "learning_rate": 2.9764541829766382e-05, + "loss": 0.0859, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020427671447396278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010260046645998954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051115389168262484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9737621128559113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6220, + "weighted_flow_matching_loss": 0.03611145447939634, + "weighted_rkd_loss": 0.04977423064410687 + }, + { + "epoch": 1.8126272912423627, + "flow_matching_loss": 0.03424253035336733, + "grad_norm": 0.36810779571533203, + "learning_rate": 2.9763080499252253e-05, + "loss": 0.0867, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021810056082904337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01027437150478363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05389448367059231, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9736783623695373, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6230, + "weighted_flow_matching_loss": 0.03424253035336733, + "weighted_rkd_loss": 0.052475934848189354 + }, + { + "epoch": 1.815536805353506, + "flow_matching_loss": 0.03517394419759512, + "grad_norm": 0.3406098186969757, + "learning_rate": 2.9761614684104064e-05, + "loss": 0.0861, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02112982850521803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010094052739441394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052353710308670996, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973594468832016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6240, + "weighted_flow_matching_loss": 0.03517394419759512, + "weighted_rkd_loss": 0.050971309468150136 + }, + { + "epoch": 1.8184463194646494, + "flow_matching_loss": 0.03486404549330473, + "grad_norm": 0.3946590721607208, + "learning_rate": 2.9760144384767095e-05, + "loss": 0.0853, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02088659629225731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010048268921673298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0518214613199234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973510456085205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6250, + "weighted_flow_matching_loss": 0.03486404549330473, + "weighted_rkd_loss": 0.05044866278767586 + }, + { + "epoch": 1.821355833575793, + "flow_matching_loss": 0.035691522620618345, + "grad_norm": 0.43799978494644165, + "learning_rate": 2.9758669601687986e-05, + "loss": 0.0882, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021661111526191233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010667357593774796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.053989580273628233, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9734263002872467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6260, + "weighted_flow_matching_loss": 0.035691522620618345, + "weighted_rkd_loss": 0.05255484953522682 + }, + { + "epoch": 1.8242653476869362, + "flow_matching_loss": 0.0330562986433506, + "grad_norm": 0.4241708517074585, + "learning_rate": 2.9757190335314722e-05, + "loss": 0.085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02118962351232767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010989861562848091, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05336910896003246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9733420252799988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6270, + "weighted_flow_matching_loss": 0.0330562986433506, + "weighted_rkd_loss": 0.051946381106972694 + }, + { + "epoch": 1.8271748617980799, + "flow_matching_loss": 0.03617570716887712, + "grad_norm": 0.38745787739753723, + "learning_rate": 2.975570658609668e-05, + "loss": 0.0856, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02040631268173456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00994028290733695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0507529079914093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9732576131820678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6280, + "weighted_flow_matching_loss": 0.03617570716887712, + "weighted_rkd_loss": 0.04939567707479 + }, + { + "epoch": 1.8300843759092231, + "flow_matching_loss": 0.034715814888477324, + "grad_norm": 0.38150545954704285, + "learning_rate": 2.9754218354484582e-05, + "loss": 0.0811, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019137560762465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009431732119992375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047706853970885275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973173075914383, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6290, + "weighted_flow_matching_loss": 0.034715814888477324, + "weighted_rkd_loss": 0.04642696194350719 + }, + { + "epoch": 1.8329938900203666, + "flow_matching_loss": 0.03658540453761816, + "grad_norm": 0.38395223021507263, + "learning_rate": 2.9752725640930504e-05, + "loss": 0.0834, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01942904610186815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009288569632917642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04814666211605072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.973088413476944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6300, + "weighted_flow_matching_loss": 0.03658540453761816, + "weighted_rkd_loss": 0.04685096964240074 + }, + { + "epoch": 1.83590340413151, + "flow_matching_loss": 0.036114735156297685, + "grad_norm": 0.3911931812763214, + "learning_rate": 2.9751228445887905e-05, + "loss": 0.0871, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020944662019610406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010482614114880561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0523719385266304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9730036079883575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6310, + "weighted_flow_matching_loss": 0.036114735156297685, + "weighted_rkd_loss": 0.050958072766661644 + }, + { + "epoch": 1.8388129182426534, + "flow_matching_loss": 0.034885013662278655, + "grad_norm": 0.4091871976852417, + "learning_rate": 2.9749726769811588e-05, + "loss": 0.0828, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01996836541220546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00934917782433331, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04928590804338455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9729186773300171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6320, + "weighted_flow_matching_loss": 0.034885013662278655, + "weighted_rkd_loss": 0.047951143607497214 + }, + { + "epoch": 1.841722432353797, + "flow_matching_loss": 0.03813372030854225, + "grad_norm": 0.3996315598487854, + "learning_rate": 2.9748220613157717e-05, + "loss": 0.0854, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01950885895639658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009523658454418183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04854137673974037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9728336274623871, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6330, + "weighted_flow_matching_loss": 0.03813372030854225, + "weighted_rkd_loss": 0.04722268171608448 + }, + { + "epoch": 1.8446319464649403, + "flow_matching_loss": 0.034751942940056325, + "grad_norm": 0.42844605445861816, + "learning_rate": 2.9746709976383832e-05, + "loss": 0.0832, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019943023566156626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00990650411695242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04979255124926567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9727484285831451, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6340, + "weighted_flow_matching_loss": 0.034751942940056325, + "weighted_rkd_loss": 0.04843561723828316 + }, + { + "epoch": 1.8475414605760838, + "flow_matching_loss": 0.03620788659900427, + "grad_norm": 0.37418875098228455, + "learning_rate": 2.974519485994882e-05, + "loss": 0.0854, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02024734728038311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010102992225438357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05059768669307232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9726631224155426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6350, + "weighted_flow_matching_loss": 0.03620788659900427, + "weighted_rkd_loss": 0.04921452589333057 + }, + { + "epoch": 1.8504509746872273, + "flow_matching_loss": 0.03729523941874504, + "grad_norm": 0.37493810057640076, + "learning_rate": 2.9743675264312934e-05, + "loss": 0.0881, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020980684831738473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010302181541919708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05226355120539665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9725776791572571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6360, + "weighted_flow_matching_loss": 0.03729523941874504, + "weighted_rkd_loss": 0.05083038136363029 + }, + { + "epoch": 1.8533604887983706, + "flow_matching_loss": 0.03501616418361664, + "grad_norm": 0.45349863171577454, + "learning_rate": 2.9742151189937784e-05, + "loss": 0.0849, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02069475445896387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009880027547478677, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051269537210464476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9724920988082886, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6370, + "weighted_flow_matching_loss": 0.03501616418361664, + "weighted_rkd_loss": 0.04985920339822769 + }, + { + "epoch": 1.8562700029095143, + "flow_matching_loss": 0.03489003274589777, + "grad_norm": 0.41762223839759827, + "learning_rate": 2.974062263728635e-05, + "loss": 0.0809, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018980272859334946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009340428933501243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047300974279642104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9724063873291016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6380, + "weighted_flow_matching_loss": 0.03489003274589777, + "weighted_rkd_loss": 0.04599580876529217 + }, + { + "epoch": 1.8591795170206575, + "flow_matching_loss": 0.03543362095952034, + "grad_norm": 0.5072683095932007, + "learning_rate": 2.973908960682296e-05, + "loss": 0.0844, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020179011672735215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010037811193615197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05039583444595337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9723205506801605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6390, + "weighted_flow_matching_loss": 0.03543362095952034, + "weighted_rkd_loss": 0.04900084733963013 + }, + { + "epoch": 1.862089031131801, + "flow_matching_loss": 0.03730682320892811, + "grad_norm": 0.4439903199672699, + "learning_rate": 2.9737552099013313e-05, + "loss": 0.0863, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020251126401126383, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009936767537146807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05043902061879635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.97223459482193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6400, + "weighted_flow_matching_loss": 0.03730682320892811, + "weighted_rkd_loss": 0.04903857409954071 + }, + { + "epoch": 1.8649985452429445, + "flow_matching_loss": 0.03498919736593962, + "grad_norm": 0.42648279666900635, + "learning_rate": 2.9736010114324462e-05, + "loss": 0.0881, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02187827732414007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010860867239534855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05461742095649243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9721485018730164, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6410, + "weighted_flow_matching_loss": 0.03498919736593962, + "weighted_rkd_loss": 0.05309629440307617 + }, + { + "epoch": 1.8679080593540878, + "flow_matching_loss": 0.03597825802862644, + "grad_norm": 0.3746717870235443, + "learning_rate": 2.9734463653224813e-05, + "loss": 0.0882, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02142732385545969, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010914852004498243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05376949869096279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9720622777938843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6420, + "weighted_flow_matching_loss": 0.03597825802862644, + "weighted_rkd_loss": 0.05226732119917869 + }, + { + "epoch": 1.8708175734652313, + "flow_matching_loss": 0.03799619618803263, + "grad_norm": 0.43517211079597473, + "learning_rate": 2.9732912716184155e-05, + "loss": 0.0871, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0204435084015131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009657698590308428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05054471492767334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9719759404659272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6430, + "weighted_flow_matching_loss": 0.03799619618803263, + "weighted_rkd_loss": 0.04912818670272827 + }, + { + "epoch": 1.8737270875763747, + "flow_matching_loss": 0.03495470955967903, + "grad_norm": 0.4318709671497345, + "learning_rate": 2.9731357303673607e-05, + "loss": 0.0861, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021129072457551957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010328786075115204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052586931735277176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.971889442205429, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6440, + "weighted_flow_matching_loss": 0.03495470955967903, + "weighted_rkd_loss": 0.051108647137880325 + }, + { + "epoch": 1.8766366016875182, + "flow_matching_loss": 0.03501621633768082, + "grad_norm": 0.3603001832962036, + "learning_rate": 2.9729797416165677e-05, + "loss": 0.0812, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0190467806532979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009471155516803265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047564716264605525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9718028366565704, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6450, + "weighted_flow_matching_loss": 0.03501621633768082, + "weighted_rkd_loss": 0.04622353985905647 + }, + { + "epoch": 1.8795461157986617, + "flow_matching_loss": 0.03415846861898899, + "grad_norm": 0.4059363305568695, + "learning_rate": 2.9728233054134202e-05, + "loss": 0.0828, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02011628057807684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009778710734099149, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050011271983385085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9717160880565643, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6460, + "weighted_flow_matching_loss": 0.03415846861898899, + "weighted_rkd_loss": 0.04859671108424664 + }, + { + "epoch": 1.882455629909805, + "flow_matching_loss": 0.035375734232366084, + "grad_norm": 0.42152246832847595, + "learning_rate": 2.9726664218054405e-05, + "loss": 0.0854, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020754903368651868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010006147902458906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051515954360365865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9716292321681976, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6470, + "weighted_flow_matching_loss": 0.035375734232366084, + "weighted_rkd_loss": 0.050054401904344556 + }, + { + "epoch": 1.8853651440209485, + "flow_matching_loss": 0.03484472334384918, + "grad_norm": 0.3888736069202423, + "learning_rate": 2.9725090908402856e-05, + "loss": 0.0857, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021087253466248512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010140928160399198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05231543555855751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.971542227268219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6480, + "weighted_flow_matching_loss": 0.03484472334384918, + "weighted_rkd_loss": 0.050826696678996086 + }, + { + "epoch": 1.888274658132092, + "flow_matching_loss": 0.03495924919843674, + "grad_norm": 0.36946403980255127, + "learning_rate": 2.972351312565748e-05, + "loss": 0.0853, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020826555602252484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01012463988736272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.051777750998735425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9714551031589508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6490, + "weighted_flow_matching_loss": 0.03495924919843674, + "weighted_rkd_loss": 0.050299783051013944 + }, + { + "epoch": 1.8911841722432354, + "flow_matching_loss": 0.03475322369486093, + "grad_norm": 0.3951794505119324, + "learning_rate": 2.9721930870297567e-05, + "loss": 0.0855, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02112909685820341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009979712776839733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05223790667951107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9713678359985352, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6500, + "weighted_flow_matching_loss": 0.03475322369486093, + "weighted_rkd_loss": 0.0507422611117363 + }, + { + "epoch": 1.894093686354379, + "flow_matching_loss": 0.03627907894551754, + "grad_norm": 0.3459213078022003, + "learning_rate": 2.9720344142803766e-05, + "loss": 0.0867, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02080830717459321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010275875823572278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05189248993992805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.97128044962883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6510, + "weighted_flow_matching_loss": 0.03627907894551754, + "weighted_rkd_loss": 0.05040208958089352 + }, + { + "epoch": 1.8970032004655222, + "flow_matching_loss": 0.03513259366154671, + "grad_norm": 0.4157164692878723, + "learning_rate": 2.9718752943658085e-05, + "loss": 0.0841, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020172444358468055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010071868868544697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05041675753891468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9711929261684418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6520, + "weighted_flow_matching_loss": 0.03513259366154671, + "weighted_rkd_loss": 0.04896441251039505 + }, + { + "epoch": 1.8999127145766657, + "flow_matching_loss": 0.034189133904874325, + "grad_norm": 0.3791375458240509, + "learning_rate": 2.9717157273343885e-05, + "loss": 0.0788, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018493857327848674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00892163827084005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04590935297310352, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9711052894592285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6530, + "weighted_flow_matching_loss": 0.034189133904874325, + "weighted_rkd_loss": 0.04458279721438885 + }, + { + "epoch": 1.9028222286878091, + "flow_matching_loss": 0.03574183825403452, + "grad_norm": 0.3989207446575165, + "learning_rate": 2.971555713234589e-05, + "loss": 0.0868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02102062404155731, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01056131673976779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0526025652885437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9710175156593323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6540, + "weighted_flow_matching_loss": 0.03574183825403452, + "weighted_rkd_loss": 0.05107804015278816 + }, + { + "epoch": 1.9057317427989524, + "flow_matching_loss": 0.03479015287011862, + "grad_norm": 0.3976944386959076, + "learning_rate": 2.9713952521150176e-05, + "loss": 0.0873, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02175567578524351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010604281444102526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05411563329398632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.970929616689682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6550, + "weighted_flow_matching_loss": 0.03479015287011862, + "weighted_rkd_loss": 0.052542563527822495 + }, + { + "epoch": 1.9086412569100961, + "flow_matching_loss": 0.03694496601819992, + "grad_norm": 0.3784920573234558, + "learning_rate": 2.971234344024419e-05, + "loss": 0.0803, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018016443960368632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008623442379757762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04465632997453213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9708415806293488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6560, + "weighted_flow_matching_loss": 0.03694496601819992, + "weighted_rkd_loss": 0.043354231491684915 + }, + { + "epoch": 1.9115507710212394, + "flow_matching_loss": 0.03336806278675795, + "grad_norm": 0.4229895770549774, + "learning_rate": 2.971072989011672e-05, + "loss": 0.0817, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01998368576169014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009869604557752609, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04983697570860386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9707534313201904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6570, + "weighted_flow_matching_loss": 0.03336806278675795, + "weighted_rkd_loss": 0.048379404842853545 + }, + { + "epoch": 1.9144602851323829, + "flow_matching_loss": 0.035985504277050495, + "grad_norm": 0.46124109625816345, + "learning_rate": 2.9709111871257927e-05, + "loss": 0.0874, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.021153089590370655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010664508771151304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.052970688045024875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9706651270389557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6580, + "weighted_flow_matching_loss": 0.035985504277050495, + "weighted_rkd_loss": 0.05141684152185917 + }, + { + "epoch": 1.9173697992435264, + "flow_matching_loss": 0.03603915460407734, + "grad_norm": 0.45113369822502136, + "learning_rate": 2.9707489384159318e-05, + "loss": 0.0868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020943707413971425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01042410871013999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05231152363121509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9705767035484314, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6590, + "weighted_flow_matching_loss": 0.03603915460407734, + "weighted_rkd_loss": 0.0507722944021225 + }, + { + "epoch": 1.9202793133546696, + "flow_matching_loss": 0.0352060291916132, + "grad_norm": 0.40447017550468445, + "learning_rate": 2.9705862429313763e-05, + "loss": 0.0828, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019806251768022774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009389201225712896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04900170490145683, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9704881608486176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6600, + "weighted_flow_matching_loss": 0.0352060291916132, + "weighted_rkd_loss": 0.04755561128258705 + }, + { + "epoch": 1.9231888274658133, + "flow_matching_loss": 0.03518133368343115, + "grad_norm": 0.400789350271225, + "learning_rate": 2.9704231007215488e-05, + "loss": 0.084, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020083843357861043, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010108538018539549, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.050276225060224534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9703994870185852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6610, + "weighted_flow_matching_loss": 0.03518133368343115, + "weighted_rkd_loss": 0.048787977918982504 + }, + { + "epoch": 1.9260983415769566, + "flow_matching_loss": 0.03454417604953051, + "grad_norm": 0.4089585840702057, + "learning_rate": 2.9702595118360077e-05, + "loss": 0.0833, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020179435424506665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009938494116067887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05029736422002316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9703106820583344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6620, + "weighted_flow_matching_loss": 0.03454417604953051, + "weighted_rkd_loss": 0.048804013058543205 + }, + { + "epoch": 1.9290078556881, + "flow_matching_loss": 0.03713500238955021, + "grad_norm": 0.3316304683685303, + "learning_rate": 2.9700954763244465e-05, + "loss": 0.0833, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019094230607151985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009428867883980275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04761732965707779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9702217400074005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6630, + "weighted_flow_matching_loss": 0.03713500238955021, + "weighted_rkd_loss": 0.0461993683129549 + }, + { + "epoch": 1.9319173697992436, + "flow_matching_loss": 0.03303360212594271, + "grad_norm": 0.44264867901802063, + "learning_rate": 2.9699309942366953e-05, + "loss": 0.0826, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020517494529485703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010008336789906026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05104332529008389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9701326727867127, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6640, + "weighted_flow_matching_loss": 0.03303360212594271, + "weighted_rkd_loss": 0.04951871372759342 + }, + { + "epoch": 1.9348268839103868, + "flow_matching_loss": 0.03680181875824928, + "grad_norm": 0.3738258183002472, + "learning_rate": 2.969766065622719e-05, + "loss": 0.0841, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01963086808100343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009503821982070804, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048765558004379275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9700434803962708, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6650, + "weighted_flow_matching_loss": 0.03680181875824928, + "weighted_rkd_loss": 0.047304673492908476 + }, + { + "epoch": 1.9377363980215305, + "flow_matching_loss": 0.035417743399739264, + "grad_norm": 0.4027640223503113, + "learning_rate": 2.96960069053262e-05, + "loss": 0.0858, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02091404739767313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010164061281830072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0519921563565731, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9699541628360748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6660, + "weighted_flow_matching_loss": 0.035417743399739264, + "weighted_rkd_loss": 0.05043003223836422 + }, + { + "epoch": 1.9406459121326738, + "flow_matching_loss": 0.03581401463598013, + "grad_norm": 0.4205491244792938, + "learning_rate": 2.9694348690166327e-05, + "loss": 0.0815, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018840853311121463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009385714353993536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.047067420557141304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9698647022247314, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6670, + "weighted_flow_matching_loss": 0.03581401463598013, + "weighted_rkd_loss": 0.045648965239524844 + }, + { + "epoch": 1.9435554262438173, + "flow_matching_loss": 0.03754146359860897, + "grad_norm": 0.4342305660247803, + "learning_rate": 2.969268601125131e-05, + "loss": 0.0829, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018843424506485462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009060498606413604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046747347339987755, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9697751224040985, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6680, + "weighted_flow_matching_loss": 0.03754146359860897, + "weighted_rkd_loss": 0.04533439688384533 + }, + { + "epoch": 1.9464649403549608, + "flow_matching_loss": 0.03504956252872944, + "grad_norm": 0.37997448444366455, + "learning_rate": 2.969101886908622e-05, + "loss": 0.0783, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01794136334210634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008731839526444674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0446145661175251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9696854233741761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6690, + "weighted_flow_matching_loss": 0.03504956252872944, + "weighted_rkd_loss": 0.04326205179095268 + }, + { + "epoch": 1.949374454466104, + "flow_matching_loss": 0.03481202591210604, + "grad_norm": 0.3937007486820221, + "learning_rate": 2.968934726417749e-05, + "loss": 0.0843, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02048035301268101, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010058246972039341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0510189525783062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9695955812931061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6700, + "weighted_flow_matching_loss": 0.03481202591210604, + "weighted_rkd_loss": 0.04946766160428524 + }, + { + "epoch": 1.9522839685772477, + "flow_matching_loss": 0.03420040309429169, + "grad_norm": 0.42794084548950195, + "learning_rate": 2.968767119703291e-05, + "loss": 0.0828, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02019082847982645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00971710323356092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05009876042604446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9695056259632111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6710, + "weighted_flow_matching_loss": 0.03420040309429169, + "weighted_rkd_loss": 0.0485710933804512 + }, + { + "epoch": 1.955193482688391, + "flow_matching_loss": 0.0369144506752491, + "grad_norm": 0.4372098445892334, + "learning_rate": 2.9685990668161628e-05, + "loss": 0.0868, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020548059605062008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0103730374481529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05146915651857853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9694155216217041, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6720, + "weighted_flow_matching_loss": 0.0369144506752491, + "weighted_rkd_loss": 0.04989490769803524 + }, + { + "epoch": 1.9581029967995345, + "flow_matching_loss": 0.032645125314593314, + "grad_norm": 0.40103963017463684, + "learning_rate": 2.9684305678074137e-05, + "loss": 0.0785, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019039197638630868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009247504826635122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04732590056955814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9693252980709076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6730, + "weighted_flow_matching_loss": 0.032645125314593314, + "weighted_rkd_loss": 0.04587416760623455 + }, + { + "epoch": 1.961012510910678, + "flow_matching_loss": 0.03678116202354431, + "grad_norm": 0.3907322287559509, + "learning_rate": 2.9682616227282304e-05, + "loss": 0.0834, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01919612679630518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009707462368533015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.048099716752767564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.969234949350357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6740, + "weighted_flow_matching_loss": 0.03678116202354431, + "weighted_rkd_loss": 0.04661990255117417 + }, + { + "epoch": 1.9639220250218212, + "flow_matching_loss": 0.035586894303560254, + "grad_norm": 0.36051103472709656, + "learning_rate": 2.968092231629933e-05, + "loss": 0.0812, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01893600057810545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009156544227153063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04702854603528976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9691444575786591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6750, + "weighted_flow_matching_loss": 0.035586894303560254, + "weighted_rkd_loss": 0.045577412471175194 + }, + { + "epoch": 1.966831539132965, + "flow_matching_loss": 0.03520314004272222, + "grad_norm": 0.4234687089920044, + "learning_rate": 2.967922394563978e-05, + "loss": 0.0816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01941665168851614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009048811439424754, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04788211472332478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.969053852558136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6760, + "weighted_flow_matching_loss": 0.03520314004272222, + "weighted_rkd_loss": 0.04640032239258289 + }, + { + "epoch": 1.9697410532441082, + "flow_matching_loss": 0.03397627919912338, + "grad_norm": 0.40732041001319885, + "learning_rate": 2.9677521115819585e-05, + "loss": 0.0828, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020387749932706357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009659753134474158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05043525286018848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9689631164073944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6770, + "weighted_flow_matching_loss": 0.03397627919912338, + "weighted_rkd_loss": 0.0488699596375227 + }, + { + "epoch": 1.9726505673552517, + "flow_matching_loss": 0.03507388979196548, + "grad_norm": 0.36038386821746826, + "learning_rate": 2.967581382735601e-05, + "loss": 0.0835, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020070245675742625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00980914020910859, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049949631467461585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9688722431659699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6780, + "weighted_flow_matching_loss": 0.03507388979196548, + "weighted_rkd_loss": 0.048394788056612015 + }, + { + "epoch": 1.9755600814663952, + "flow_matching_loss": 0.035417002812027934, + "grad_norm": 0.422882616519928, + "learning_rate": 2.9674102080767685e-05, + "loss": 0.0851, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020384116657078268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.01047085588797927, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05123908966779709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9687812566757202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6790, + "weighted_flow_matching_loss": 0.035417002812027934, + "weighted_rkd_loss": 0.04963946640491486 + }, + { + "epoch": 1.9784695955775384, + "flow_matching_loss": 0.03491533119231462, + "grad_norm": 0.3932698667049408, + "learning_rate": 2.9672385876574597e-05, + "loss": 0.0856, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02105749063193798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010158219980075956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05227320082485676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9686901271343231, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6800, + "weighted_flow_matching_loss": 0.03491533119231462, + "weighted_rkd_loss": 0.050636491179466246 + }, + { + "epoch": 1.9813791096886821, + "flow_matching_loss": 0.03631148673593998, + "grad_norm": 0.341264009475708, + "learning_rate": 2.967066521529808e-05, + "loss": 0.0867, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020801112987101078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010451316693797707, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05205354206264019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9685988783836365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6810, + "weighted_flow_matching_loss": 0.03631148673593998, + "weighted_rkd_loss": 0.05041904672980309 + }, + { + "epoch": 1.9842886237998254, + "flow_matching_loss": 0.03540268428623676, + "grad_norm": 0.3979284465312958, + "learning_rate": 2.966894009746083e-05, + "loss": 0.0838, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019999766536056996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009927580878138543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049927113205194475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9685074925422669, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6820, + "weighted_flow_matching_loss": 0.03540268428623676, + "weighted_rkd_loss": 0.04835481904447079 + }, + { + "epoch": 1.9871981379109689, + "flow_matching_loss": 0.03381092213094235, + "grad_norm": 0.42159879207611084, + "learning_rate": 2.9667210523586888e-05, + "loss": 0.0819, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01986734624952078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009874842129647732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04960953369736672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9684159815311432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6830, + "weighted_flow_matching_loss": 0.03381092213094235, + "weighted_rkd_loss": 0.04804268814623356 + }, + { + "epoch": 1.9901076520221124, + "flow_matching_loss": 0.035085861943662165, + "grad_norm": 0.3857424259185791, + "learning_rate": 2.966547649420165e-05, + "loss": 0.0812, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019251326471567153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009117132052779198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04761978536844254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9683243453502655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6840, + "weighted_flow_matching_loss": 0.035085861943662165, + "weighted_rkd_loss": 0.04611138738691807 + }, + { + "epoch": 1.9930171661332556, + "flow_matching_loss": 0.03523168247193098, + "grad_norm": 0.4424073398113251, + "learning_rate": 2.9663738009831877e-05, + "loss": 0.0807, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019006266351789237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00891948058269918, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04693201407790184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9682325720787048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6850, + "weighted_flow_matching_loss": 0.03523168247193098, + "weighted_rkd_loss": 0.045441100373864174 + }, + { + "epoch": 1.9959266802443993, + "flow_matching_loss": 0.03533427696675062, + "grad_norm": 0.48295900225639343, + "learning_rate": 2.966199507100567e-05, + "loss": 0.0841, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0202303659170866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00986959571018815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05033032707870007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9681406855583191, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6860, + "weighted_flow_matching_loss": 0.03533427696675062, + "weighted_rkd_loss": 0.048726806044578554 + }, + { + "epoch": 1.9988361943555426, + "flow_matching_loss": 0.03350482489913702, + "grad_norm": 0.38842883706092834, + "learning_rate": 2.9660247678252484e-05, + "loss": 0.0831, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.02060692310333252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010033974656835198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05124782063066959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9680486619472504, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6870, + "weighted_flow_matching_loss": 0.03350482489913702, + "weighted_rkd_loss": 0.04961040876805782 + }, + { + "epoch": 2.001745708466686, + "flow_matching_loss": 0.036963261850178245, + "grad_norm": 0.4301987290382385, + "learning_rate": 2.965849583210314e-05, + "loss": 0.0845, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019632514193654062, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009829141059890389, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049094169214367864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9679565131664276, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6880, + "weighted_flow_matching_loss": 0.036963261850178245, + "weighted_rkd_loss": 0.04752103090286255 + }, + { + "epoch": 2.0046552225778296, + "flow_matching_loss": 0.03473290912806988, + "grad_norm": 0.41996920108795166, + "learning_rate": 2.9656739533089794e-05, + "loss": 0.0813, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019405173789709806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0093439482152462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0481542956084013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9678642332553864, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6890, + "weighted_flow_matching_loss": 0.03473290912806988, + "weighted_rkd_loss": 0.04660685732960701 + }, + { + "epoch": 2.007564736688973, + "flow_matching_loss": 0.0341061744838953, + "grad_norm": 0.44798457622528076, + "learning_rate": 2.9654978781745968e-05, + "loss": 0.079, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01869640816003084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008962025213986636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04635484144091606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9677718281745911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6900, + "weighted_flow_matching_loss": 0.0341061744838953, + "weighted_rkd_loss": 0.04486087597906589 + }, + { + "epoch": 2.0104742508001165, + "flow_matching_loss": 0.03572286907583475, + "grad_norm": 0.3505077660083771, + "learning_rate": 2.9653213578606534e-05, + "loss": 0.0811, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01875542001798749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009374655038118362, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046885494887828824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9676792860031128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6910, + "weighted_flow_matching_loss": 0.03572286907583475, + "weighted_rkd_loss": 0.04537009634077549 + }, + { + "epoch": 2.01338376491126, + "flow_matching_loss": 0.035804251208901405, + "grad_norm": 0.4043956398963928, + "learning_rate": 2.9651443924207704e-05, + "loss": 0.0816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018891695234924556, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00957197993993759, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04735537022352219, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9675866305828095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6920, + "weighted_flow_matching_loss": 0.035804251208901405, + "weighted_rkd_loss": 0.045820409432053566 + }, + { + "epoch": 2.016293279022403, + "flow_matching_loss": 0.03356452863663435, + "grad_norm": 0.3923085629940033, + "learning_rate": 2.964966981908706e-05, + "loss": 0.0797, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019259936921298505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009203790966421366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04772366508841515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9674938321113586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6930, + "weighted_flow_matching_loss": 0.03356452863663435, + "weighted_rkd_loss": 0.04617238976061344 + }, + { + "epoch": 2.019202793133547, + "flow_matching_loss": 0.035168423876166345, + "grad_norm": 0.35209891200065613, + "learning_rate": 2.964789126378353e-05, + "loss": 0.082, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01960935853421688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009185223001986742, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04840394072234631, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9674009084701538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6940, + "weighted_flow_matching_loss": 0.035168423876166345, + "weighted_rkd_loss": 0.04682600349187851 + }, + { + "epoch": 2.02211230724469, + "flow_matching_loss": 0.036476806923747065, + "grad_norm": 0.49004384875297546, + "learning_rate": 2.9646108258837388e-05, + "loss": 0.0795, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017872338835150003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008690559258684515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04443523697555065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9673078656196594, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6950, + "weighted_flow_matching_loss": 0.036476806923747065, + "weighted_rkd_loss": 0.042982491478323934 + }, + { + "epoch": 2.0250218213558338, + "flow_matching_loss": 0.03411474823951721, + "grad_norm": 0.41337886452674866, + "learning_rate": 2.9644320804790267e-05, + "loss": 0.0773, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018030437547713517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008550247363746166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0446111224591732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9672146916389466, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6960, + "weighted_flow_matching_loss": 0.03411474823951721, + "weighted_rkd_loss": 0.043148523569107054 + }, + { + "epoch": 2.027931335466977, + "flow_matching_loss": 0.033961812034249306, + "grad_norm": 0.34082502126693726, + "learning_rate": 2.9642528902185142e-05, + "loss": 0.0783, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018517169263213874, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008793775225058198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045828113704919814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9671213805675507, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6970, + "weighted_flow_matching_loss": 0.033961812034249306, + "weighted_rkd_loss": 0.04432131461799145 + }, + { + "epoch": 2.0308408495781203, + "flow_matching_loss": 0.03453426007181406, + "grad_norm": 0.3039707541465759, + "learning_rate": 2.9640732551566354e-05, + "loss": 0.0782, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018140660598874092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008892193855717778, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04517351500689983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9670279562473297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6980, + "weighted_flow_matching_loss": 0.03453426007181406, + "weighted_rkd_loss": 0.04368423521518707 + }, + { + "epoch": 2.033750363689264, + "flow_matching_loss": 0.032261612452566625, + "grad_norm": 0.4237155616283417, + "learning_rate": 2.9638931753479578e-05, + "loss": 0.0766, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018543172627687454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00876014307141304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04584648795425892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9669343948364257, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 6990, + "weighted_flow_matching_loss": 0.032261612452566625, + "weighted_rkd_loss": 0.044330519065260886 + }, + { + "epoch": 2.0366598778004072, + "flow_matching_loss": 0.034955838695168495, + "grad_norm": 0.366534560918808, + "learning_rate": 2.9637126508471858e-05, + "loss": 0.081, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019043572805821897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009538094513118267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04762524031102657, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9668406963348388, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7000, + "weighted_flow_matching_loss": 0.034955838695168495, + "weighted_rkd_loss": 0.04604605808854103 + }, + { + "epoch": 2.039569391911551, + "flow_matching_loss": 0.031194834411144255, + "grad_norm": 0.3929683268070221, + "learning_rate": 2.9635316817091577e-05, + "loss": 0.0763, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018960431218147278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008767441473901271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04668830409646034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9667468786239624, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7010, + "weighted_flow_matching_loss": 0.031194834411144255, + "weighted_rkd_loss": 0.04513569250702858 + }, + { + "epoch": 2.042478906022694, + "flow_matching_loss": 0.034612925723195076, + "grad_norm": 0.39219072461128235, + "learning_rate": 2.9633502679888467e-05, + "loss": 0.08, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018961084075272083, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009056024020537734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046978192031383516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9666529357433319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7020, + "weighted_flow_matching_loss": 0.034612925723195076, + "weighted_rkd_loss": 0.0454116228967905 + }, + { + "epoch": 2.0453884201338375, + "flow_matching_loss": 0.03502095025032759, + "grad_norm": 0.36104217171669006, + "learning_rate": 2.963168409741362e-05, + "loss": 0.0828, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019926333986222743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009606924653053284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049459592625498774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9665588676929474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7030, + "weighted_flow_matching_loss": 0.03502095025032759, + "weighted_rkd_loss": 0.04780560284852982 + }, + { + "epoch": 2.048297934244981, + "flow_matching_loss": 0.034093027003109455, + "grad_norm": 0.3438427448272705, + "learning_rate": 2.9629861070219467e-05, + "loss": 0.0812, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01948755793273449, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009728669608011841, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04870378561317921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9664646685123444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7040, + "weighted_flow_matching_loss": 0.034093027003109455, + "weighted_rkd_loss": 0.04707049615681171 + }, + { + "epoch": 2.0512074483561245, + "flow_matching_loss": 0.03489105477929115, + "grad_norm": 0.3714924156665802, + "learning_rate": 2.9628033598859807e-05, + "loss": 0.0801, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018713230080902577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009350696206092834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04677715674042702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9663703262805938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7050, + "weighted_flow_matching_loss": 0.03489105477929115, + "weighted_rkd_loss": 0.045204044133424756 + }, + { + "epoch": 2.054116962467268, + "flow_matching_loss": 0.0336213355883956, + "grad_norm": 0.4774661362171173, + "learning_rate": 2.9626201683889764e-05, + "loss": 0.0793, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019038770999759435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009202009811997414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04727955274283886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9662758767604828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7060, + "weighted_flow_matching_loss": 0.0336213355883956, + "weighted_rkd_loss": 0.04568508192896843 + }, + { + "epoch": 2.0570264765784114, + "flow_matching_loss": 0.033289106003940104, + "grad_norm": 0.41978442668914795, + "learning_rate": 2.9624365325865835e-05, + "loss": 0.0816, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019958500005304813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.010093076713383197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05001007616519928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9661812901496887, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7070, + "weighted_flow_matching_loss": 0.033289106003940104, + "weighted_rkd_loss": 0.04831882081925869 + }, + { + "epoch": 2.0599359906895547, + "flow_matching_loss": 0.03369449842721224, + "grad_norm": 0.3580319881439209, + "learning_rate": 2.962252452534585e-05, + "loss": 0.0777, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018191835563629865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009136104211211204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04551977589726448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9660865962505341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7080, + "weighted_flow_matching_loss": 0.03369449842721224, + "weighted_rkd_loss": 0.04397596679627895 + }, + { + "epoch": 2.0628455048006984, + "flow_matching_loss": 0.03643507920205593, + "grad_norm": 0.3822211027145386, + "learning_rate": 2.9620679282889006e-05, + "loss": 0.0824, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019149995222687722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009286399325355887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04758638925850391, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.965991735458374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7090, + "weighted_flow_matching_loss": 0.03643507920205593, + "weighted_rkd_loss": 0.04596803560853004 + }, + { + "epoch": 2.0657550189118417, + "flow_matching_loss": 0.03387464787811041, + "grad_norm": 0.3874756097793579, + "learning_rate": 2.9618829599055833e-05, + "loss": 0.0798, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01909602265805006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009405524516478182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04759756997227669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9658967733383179, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7100, + "weighted_flow_matching_loss": 0.03387464787811041, + "weighted_rkd_loss": 0.045974307134747504 + }, + { + "epoch": 2.0686645330229854, + "flow_matching_loss": 0.033403493277728555, + "grad_norm": 0.36407285928726196, + "learning_rate": 2.961697547440821e-05, + "loss": 0.0748, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01735933721065521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008128562662750482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04284723699092865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9658016741275788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7110, + "weighted_flow_matching_loss": 0.033403493277728555, + "weighted_rkd_loss": 0.04138191044330597 + }, + { + "epoch": 2.0715740471341286, + "flow_matching_loss": 0.0345667252317071, + "grad_norm": 0.45090535283088684, + "learning_rate": 2.9615116909509383e-05, + "loss": 0.0791, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018644759990274908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008836520975455642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04612604081630707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9657064497470855, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7120, + "weighted_flow_matching_loss": 0.0345667252317071, + "weighted_rkd_loss": 0.04454420395195484 + }, + { + "epoch": 2.074483561245272, + "flow_matching_loss": 0.03434982355684042, + "grad_norm": 0.35046786069869995, + "learning_rate": 2.9613253904923924e-05, + "loss": 0.0791, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01855154512450099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009279710333794355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04638280086219311, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9656111001968384, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7130, + "weighted_flow_matching_loss": 0.03434982355684042, + "weighted_rkd_loss": 0.04478773660957813 + }, + { + "epoch": 2.0773930753564156, + "flow_matching_loss": 0.03562445715069771, + "grad_norm": 0.33213111758232117, + "learning_rate": 2.9611386461217772e-05, + "loss": 0.0801, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01858010469004512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008857294591143728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04601750373840332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9655156195163727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7140, + "weighted_flow_matching_loss": 0.03562445715069771, + "weighted_rkd_loss": 0.044430628046393396 + }, + { + "epoch": 2.080302589467559, + "flow_matching_loss": 0.03516542427241802, + "grad_norm": 0.355943500995636, + "learning_rate": 2.96095145789582e-05, + "loss": 0.0792, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018426734767854214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00876715248450637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045620621740818025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9654200077056885, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7150, + "weighted_flow_matching_loss": 0.03516542427241802, + "weighted_rkd_loss": 0.04404304884374142 + }, + { + "epoch": 2.0832121035787026, + "flow_matching_loss": 0.033403361588716506, + "grad_norm": 0.3688553273677826, + "learning_rate": 2.9607638258713845e-05, + "loss": 0.0775, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018401410430669785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008904906967654825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04570772796869278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9653242766857147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7160, + "weighted_flow_matching_loss": 0.033403361588716506, + "weighted_rkd_loss": 0.044122814387083056 + }, + { + "epoch": 2.086121617689846, + "flow_matching_loss": 0.033595459349453446, + "grad_norm": 0.39695051312446594, + "learning_rate": 2.960575750105468e-05, + "loss": 0.0785, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018768616765737534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009017551690340042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04655478484928608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.965228408575058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7170, + "weighted_flow_matching_loss": 0.033595459349453446, + "weighted_rkd_loss": 0.04493599608540535 + }, + { + "epoch": 2.089031131800989, + "flow_matching_loss": 0.03391719236969948, + "grad_norm": 0.34939441084861755, + "learning_rate": 2.9603872306552025e-05, + "loss": 0.0765, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017919650301337242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00833344361744821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04417274482548237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9651324272155761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7180, + "weighted_flow_matching_loss": 0.03391719236969948, + "weighted_rkd_loss": 0.04263252690434456 + }, + { + "epoch": 2.091940645912133, + "flow_matching_loss": 0.03422503974288702, + "grad_norm": 0.38979706168174744, + "learning_rate": 2.9601982675778562e-05, + "loss": 0.0793, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018895884975790977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008902255352586508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04669402539730072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9650363087654114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7190, + "weighted_flow_matching_loss": 0.03422503974288702, + "weighted_rkd_loss": 0.04506141059100628 + }, + { + "epoch": 2.094850160023276, + "flow_matching_loss": 0.03411273527890444, + "grad_norm": 0.3580826222896576, + "learning_rate": 2.9600088609308294e-05, + "loss": 0.0803, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01928622992709279, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009302816074341535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04787527620792389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9649400651454926, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7200, + "weighted_flow_matching_loss": 0.03411273527890444, + "weighted_rkd_loss": 0.04619680494070053 + }, + { + "epoch": 2.0977596741344193, + "flow_matching_loss": 0.032101063430309294, + "grad_norm": 0.32040125131607056, + "learning_rate": 2.959819010771661e-05, + "loss": 0.078, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019075493700802325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009419449092820286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04757043644785881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9648437023162841, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7210, + "weighted_flow_matching_loss": 0.032101063430309294, + "weighted_rkd_loss": 0.04589806012809276 + }, + { + "epoch": 2.100669188245563, + "flow_matching_loss": 0.03484699074178934, + "grad_norm": 0.3420301675796509, + "learning_rate": 2.959628717158021e-05, + "loss": 0.0789, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01858229087665677, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008480756264179944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0456453375518322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9647471904754639, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7220, + "weighted_flow_matching_loss": 0.03484699074178934, + "weighted_rkd_loss": 0.044036278128623964 + }, + { + "epoch": 2.1035787023567063, + "flow_matching_loss": 0.031465462408959866, + "grad_norm": 0.4235519468784332, + "learning_rate": 2.959437980147716e-05, + "loss": 0.0783, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0196862543001771, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009144423017278313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04851693138480186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.964650571346283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7230, + "weighted_flow_matching_loss": 0.031465462408959866, + "weighted_rkd_loss": 0.04680180698633194 + }, + { + "epoch": 2.10648821646785, + "flow_matching_loss": 0.03397679831832647, + "grad_norm": 0.33208340406417847, + "learning_rate": 2.959246799798687e-05, + "loss": 0.078, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018564602732658385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008531352877616883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04566055834293366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9645538210868836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7240, + "weighted_flow_matching_loss": 0.03397679831832647, + "weighted_rkd_loss": 0.04404207617044449 + }, + { + "epoch": 2.1093977305789933, + "flow_matching_loss": 0.032040412351489064, + "grad_norm": 0.34892481565475464, + "learning_rate": 2.9590551761690095e-05, + "loss": 0.0764, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018425376899540425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009114165604114533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045964920148253444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9644569337368012, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7250, + "weighted_flow_matching_loss": 0.032040412351489064, + "weighted_rkd_loss": 0.04433109797537327 + }, + { + "epoch": 2.112307244690137, + "flow_matching_loss": 0.0331788931041956, + "grad_norm": 0.40169277787208557, + "learning_rate": 2.9588631093168936e-05, + "loss": 0.0766, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018102164287120102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008843155298382043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04504748396575451, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9643599331378937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7260, + "weighted_flow_matching_loss": 0.0331788931041956, + "weighted_rkd_loss": 0.04344196729362011 + }, + { + "epoch": 2.1152167588012802, + "flow_matching_loss": 0.03324162662029266, + "grad_norm": 0.3634774684906006, + "learning_rate": 2.9586705993006837e-05, + "loss": 0.0753, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01759700272232294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008448524959385395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04364253059029579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9642628014087677, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7270, + "weighted_flow_matching_loss": 0.03324162662029266, + "weighted_rkd_loss": 0.04208290837705135 + }, + { + "epoch": 2.1181262729124235, + "flow_matching_loss": 0.03509385846555233, + "grad_norm": 0.4179222881793976, + "learning_rate": 2.95847764617886e-05, + "loss": 0.0807, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019278411753475666, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008780545042827725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04733736850321293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9641655266284943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7280, + "weighted_flow_matching_loss": 0.03509385846555233, + "weighted_rkd_loss": 0.04564106836915016 + }, + { + "epoch": 2.121035787023567, + "flow_matching_loss": 0.03261497877538204, + "grad_norm": 0.394074410200119, + "learning_rate": 2.9582842500100364e-05, + "loss": 0.0756, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017882786691188812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008835549093782901, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04460112266242504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9640681445598602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7290, + "weighted_flow_matching_loss": 0.03261497877538204, + "weighted_rkd_loss": 0.04299848731607199 + }, + { + "epoch": 2.1239453011347105, + "flow_matching_loss": 0.033829766884446145, + "grad_norm": 0.42009425163269043, + "learning_rate": 2.958090410852961e-05, + "loss": 0.0809, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019498562440276146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009796785842627286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04879391118884087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9639706254005432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7300, + "weighted_flow_matching_loss": 0.033829766884446145, + "weighted_rkd_loss": 0.047035911306738856 + }, + { + "epoch": 2.1268548152458537, + "flow_matching_loss": 0.031458829529583454, + "grad_norm": 0.38007092475891113, + "learning_rate": 2.9578961287665175e-05, + "loss": 0.0788, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01971273347735405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009726780001074076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04915224686264992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9638729691505432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7310, + "weighted_flow_matching_loss": 0.031458829529583454, + "weighted_rkd_loss": 0.04737653732299805 + }, + { + "epoch": 2.1297643293569974, + "flow_matching_loss": 0.03461268413811922, + "grad_norm": 0.3397729694843292, + "learning_rate": 2.9577014038097238e-05, + "loss": 0.078, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018104040529578924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00879933312535286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04500741362571716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9637751996517181, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7320, + "weighted_flow_matching_loss": 0.03461268413811922, + "weighted_rkd_loss": 0.04337698705494404 + }, + { + "epoch": 2.1326738434681407, + "flow_matching_loss": 0.035374191403388974, + "grad_norm": 0.3337016999721527, + "learning_rate": 2.9575062360417324e-05, + "loss": 0.0831, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01988626690581441, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009751616278663278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.049524149671196935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9636773109436035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7330, + "weighted_flow_matching_loss": 0.035374191403388974, + "weighted_rkd_loss": 0.047725359350442885 + }, + { + "epoch": 2.1355833575792844, + "flow_matching_loss": 0.03346206620335579, + "grad_norm": 0.33237069845199585, + "learning_rate": 2.957310625521829e-05, + "loss": 0.0783, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01882558148354292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008914591604843735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04656575545668602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.963579285144806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7340, + "weighted_flow_matching_loss": 0.03346206620335579, + "weighted_rkd_loss": 0.04486974589526653 + }, + { + "epoch": 2.1384928716904277, + "flow_matching_loss": 0.03487737197428942, + "grad_norm": 0.4366443157196045, + "learning_rate": 2.957114572309436e-05, + "loss": 0.0747, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016616611555218697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00812426432967186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041357486695051196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9634811222553253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7350, + "weighted_flow_matching_loss": 0.03487737197428942, + "weighted_rkd_loss": 0.03984707333147526 + }, + { + "epoch": 2.141402385801571, + "flow_matching_loss": 0.03188405558466911, + "grad_norm": 0.3886711001396179, + "learning_rate": 2.956918076464109e-05, + "loss": 0.0747, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017965123988687992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008553115604445338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04448336362838745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9633828639984131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7360, + "weighted_flow_matching_loss": 0.03188405558466911, + "weighted_rkd_loss": 0.042854460328817366 + }, + { + "epoch": 2.1443118999127146, + "flow_matching_loss": 0.03445655629038811, + "grad_norm": 0.3686378002166748, + "learning_rate": 2.9567211380455376e-05, + "loss": 0.079, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018524365965276957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009206407750025392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046255139634013176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9632844388484955, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7370, + "weighted_flow_matching_loss": 0.03445655629038811, + "weighted_rkd_loss": 0.04455675929784775 + }, + { + "epoch": 2.147221414023858, + "flow_matching_loss": 0.03216169960796833, + "grad_norm": 0.36602476239204407, + "learning_rate": 2.9565237571135472e-05, + "loss": 0.0735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017456394620239735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008004505001008511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04291729368269444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9631859004497528, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7380, + "weighted_flow_matching_loss": 0.03216169960796833, + "weighted_rkd_loss": 0.04133731871843338 + }, + { + "epoch": 2.1501309281350016, + "flow_matching_loss": 0.03252312485128641, + "grad_norm": 0.3809986710548401, + "learning_rate": 2.9563259337280967e-05, + "loss": 0.0768, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018605638295412064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008812953671440481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04602422975003719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9630872428417205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7390, + "weighted_flow_matching_loss": 0.03252312485128641, + "weighted_rkd_loss": 0.044325248897075654 + }, + { + "epoch": 2.153040442246145, + "flow_matching_loss": 0.033495691046118735, + "grad_norm": 0.3786606192588806, + "learning_rate": 2.9561276679492794e-05, + "loss": 0.0764, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01804456003010273, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0084664068184793, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044555526599287985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9629884541034699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7400, + "weighted_flow_matching_loss": 0.033495691046118735, + "weighted_rkd_loss": 0.04290645495057106 + }, + { + "epoch": 2.155949956357288, + "flow_matching_loss": 0.03329925797879696, + "grad_norm": 0.37030482292175293, + "learning_rate": 2.9559289598373236e-05, + "loss": 0.0761, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01800095122307539, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008452019980177283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04445392191410065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9628895461559296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7410, + "weighted_flow_matching_loss": 0.03329925797879696, + "weighted_rkd_loss": 0.04280415996909141 + }, + { + "epoch": 2.158859470468432, + "flow_matching_loss": 0.03774644788354635, + "grad_norm": 0.43434208631515503, + "learning_rate": 2.9557298094525913e-05, + "loss": 0.085, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.019792080484330656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009466895554214715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04905105717480183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9627904951572418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7420, + "weighted_flow_matching_loss": 0.03774644788354635, + "weighted_rkd_loss": 0.04722588807344437 + }, + { + "epoch": 2.161768984579575, + "flow_matching_loss": 0.03363563157618046, + "grad_norm": 0.3790956437587738, + "learning_rate": 2.955530216855579e-05, + "loss": 0.0732, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016724972799420357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007643047627061606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04109299331903458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.962691330909729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7430, + "weighted_flow_matching_loss": 0.03363563157618046, + "weighted_rkd_loss": 0.03955989852547646 + }, + { + "epoch": 2.164678498690719, + "flow_matching_loss": 0.03396334387362003, + "grad_norm": 0.41917741298675537, + "learning_rate": 2.955330182106918e-05, + "loss": 0.0824, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.020381305366754532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009559571370482444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.05032218247652054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9625920295715332, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7440, + "weighted_flow_matching_loss": 0.03396334387362003, + "weighted_rkd_loss": 0.048439739271998405 + }, + { + "epoch": 2.167588012801862, + "flow_matching_loss": 0.03199628759175539, + "grad_norm": 0.3470848500728607, + "learning_rate": 2.9551297052673734e-05, + "loss": 0.0759, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018437081854790448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008748150523751974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045622315257787704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9624926090240479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7450, + "weighted_flow_matching_loss": 0.03199628759175539, + "weighted_rkd_loss": 0.043911101296544075 + }, + { + "epoch": 2.1704975269130053, + "flow_matching_loss": 0.03341722190380096, + "grad_norm": 0.3461305797100067, + "learning_rate": 2.954928786397845e-05, + "loss": 0.0755, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01771789714694023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0082637841347605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043699578195810315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.962393057346344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7460, + "weighted_flow_matching_loss": 0.03341722190380096, + "weighted_rkd_loss": 0.04205624461174011 + }, + { + "epoch": 2.173407041024149, + "flow_matching_loss": 0.03480984885245562, + "grad_norm": 0.3442961275577545, + "learning_rate": 2.9547274255593665e-05, + "loss": 0.0774, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01765624964609742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008928997721523047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044241496920585634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9622933864593506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7470, + "weighted_flow_matching_loss": 0.03480984885245562, + "weighted_rkd_loss": 0.042573298886418344 + }, + { + "epoch": 2.1763165551352923, + "flow_matching_loss": 0.03258742168545723, + "grad_norm": 0.36575546860694885, + "learning_rate": 2.9545256228131058e-05, + "loss": 0.0769, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018690828140825032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008642571326345206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046024227142333986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9621935725212097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7480, + "weighted_flow_matching_loss": 0.03258742168545723, + "weighted_rkd_loss": 0.044284379482269286 + }, + { + "epoch": 2.179226069246436, + "flow_matching_loss": 0.03397844154387712, + "grad_norm": 0.3069742023944855, + "learning_rate": 2.9543233782203653e-05, + "loss": 0.0778, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018278495501726867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008963612280786038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04552060328423977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9620936512947083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7490, + "weighted_flow_matching_loss": 0.03397844154387712, + "weighted_rkd_loss": 0.043795197270810604 + }, + { + "epoch": 2.1821355833575793, + "flow_matching_loss": 0.03325480502098799, + "grad_norm": 0.3877865970134735, + "learning_rate": 2.954120691842582e-05, + "loss": 0.0779, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018752238247543573, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008930718060582876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.046435194462537764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9619935750961304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7500, + "weighted_flow_matching_loss": 0.03325480502098799, + "weighted_rkd_loss": 0.0446702852845192 + }, + { + "epoch": 2.1850450974687226, + "flow_matching_loss": 0.03644457422196865, + "grad_norm": 0.38229697942733765, + "learning_rate": 2.9539175637413264e-05, + "loss": 0.0751, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016250348743051292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007733071921393275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04023376889526844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9618934154510498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7510, + "weighted_flow_matching_loss": 0.03644457422196865, + "weighted_rkd_loss": 0.03870059698820114 + }, + { + "epoch": 2.1879546115798663, + "flow_matching_loss": 0.03400642555207014, + "grad_norm": 0.32181164622306824, + "learning_rate": 2.9537139939783033e-05, + "loss": 0.0759, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017628361005336046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008350974088534713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043607696518301965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9617931008338928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7520, + "weighted_flow_matching_loss": 0.03400642555207014, + "weighted_rkd_loss": 0.04194166548550129 + }, + { + "epoch": 2.1908641256910095, + "flow_matching_loss": 0.03288216795772314, + "grad_norm": 0.3620411455631256, + "learning_rate": 2.9535099826153516e-05, + "loss": 0.0724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016564835514873266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00801170994527638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04114138074219227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9616926670074463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7530, + "weighted_flow_matching_loss": 0.03288216795772314, + "weighted_rkd_loss": 0.0395653635263443 + }, + { + "epoch": 2.1937736398021532, + "flow_matching_loss": 0.03426376786082983, + "grad_norm": 0.37896445393562317, + "learning_rate": 2.9533055297144454e-05, + "loss": 0.0735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016399007756263018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007961468119174243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04075948409736156, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9615920960903168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7540, + "weighted_flow_matching_loss": 0.03426376786082983, + "weighted_rkd_loss": 0.03919403664767742 + }, + { + "epoch": 2.1966831539132965, + "flow_matching_loss": 0.032712244614958766, + "grad_norm": 0.31625574827194214, + "learning_rate": 2.953100635337691e-05, + "loss": 0.0728, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016890323627740143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00786463925614953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04164528623223305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9614914238452912, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7550, + "weighted_flow_matching_loss": 0.032712244614958766, + "weighted_rkd_loss": 0.04004162214696407 + }, + { + "epoch": 2.1995926680244398, + "flow_matching_loss": 0.033312438242137435, + "grad_norm": 0.39913156628608704, + "learning_rate": 2.952895299547331e-05, + "loss": 0.0732, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016845340933650733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007769477972760797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04146015979349613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.961390596628189, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7560, + "weighted_flow_matching_loss": 0.033312438242137435, + "weighted_rkd_loss": 0.03985942527651787 + }, + { + "epoch": 2.2025021821355835, + "flow_matching_loss": 0.03339692670851946, + "grad_norm": 0.4121968746185303, + "learning_rate": 2.9526895224057402e-05, + "loss": 0.0754, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01768313320353627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008360719680786133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04372698590159416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9612896561622619, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7570, + "weighted_flow_matching_loss": 0.03339692670851946, + "weighted_rkd_loss": 0.042034311965107915 + }, + { + "epoch": 2.2054116962467267, + "flow_matching_loss": 0.03379805702716112, + "grad_norm": 0.31628602743148804, + "learning_rate": 2.9524833039754284e-05, + "loss": 0.0764, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017753866873681544, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008784780604764819, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04429251439869404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9611885845661163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7580, + "weighted_flow_matching_loss": 0.03379805702716112, + "weighted_rkd_loss": 0.04257351607084274 + }, + { + "epoch": 2.2083212103578704, + "flow_matching_loss": 0.034141728095710276, + "grad_norm": 0.31693509221076965, + "learning_rate": 2.9522766443190398e-05, + "loss": 0.0753, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017250902112573387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00833790353499353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04283970780670643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9610873878002166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7590, + "weighted_flow_matching_loss": 0.034141728095710276, + "weighted_rkd_loss": 0.04117272607982159 + }, + { + "epoch": 2.2112307244690137, + "flow_matching_loss": 0.034972178377211095, + "grad_norm": 0.3239327371120453, + "learning_rate": 2.9520695434993516e-05, + "loss": 0.0759, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01726019652560353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00807547364383936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042595867067575455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.960986065864563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7600, + "weighted_flow_matching_loss": 0.034972178377211095, + "weighted_rkd_loss": 0.040934079140424726 + }, + { + "epoch": 2.214140238580157, + "flow_matching_loss": 0.034637076407670976, + "grad_norm": 0.33352962136268616, + "learning_rate": 2.9518620015792757e-05, + "loss": 0.0748, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01695810491219163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007910444028675556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041826653853058816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9608846187591553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7610, + "weighted_flow_matching_loss": 0.034637076407670976, + "weighted_rkd_loss": 0.04019065015017986 + }, + { + "epoch": 2.2170497526913007, + "flow_matching_loss": 0.03180327732115984, + "grad_norm": 0.4324153661727905, + "learning_rate": 2.951654018621858e-05, + "loss": 0.0729, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017163297906517983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008463740302249789, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042790335789322856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9607830345630646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7620, + "weighted_flow_matching_loss": 0.03180327732115984, + "weighted_rkd_loss": 0.04111224412918091 + }, + { + "epoch": 2.219959266802444, + "flow_matching_loss": 0.031658276170492175, + "grad_norm": 0.4355967342853546, + "learning_rate": 2.951445594690278e-05, + "loss": 0.0754, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018287692312151195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0089538243599236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04552920870482922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9606813490390778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7630, + "weighted_flow_matching_loss": 0.031658276170492175, + "weighted_rkd_loss": 0.043738970160484315 + }, + { + "epoch": 2.2228687809135876, + "flow_matching_loss": 0.032620221562683585, + "grad_norm": 0.34414470195770264, + "learning_rate": 2.95123672984785e-05, + "loss": 0.0748, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017737141251564024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008418315695598722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04389259815216064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9605795204639435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7640, + "weighted_flow_matching_loss": 0.032620221562683585, + "weighted_rkd_loss": 0.04216240383684635 + }, + { + "epoch": 2.225778295024731, + "flow_matching_loss": 0.03336569331586361, + "grad_norm": 0.3518248200416565, + "learning_rate": 2.9510274241580207e-05, + "loss": 0.0753, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017625956516712903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008378335880115629, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04363024868071079, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9604775607585907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7650, + "weighted_flow_matching_loss": 0.03336569331586361, + "weighted_rkd_loss": 0.04190586470067501 + }, + { + "epoch": 2.228687809135874, + "flow_matching_loss": 0.03417206220328808, + "grad_norm": 0.35522112250328064, + "learning_rate": 2.9508176776843726e-05, + "loss": 0.0783, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018417415395379066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009077339665964246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04591217078268528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9603754878044128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7660, + "weighted_flow_matching_loss": 0.03417206220328808, + "weighted_rkd_loss": 0.044093067198991774 + }, + { + "epoch": 2.231597323247018, + "flow_matching_loss": 0.032523981668055056, + "grad_norm": 0.32585659623146057, + "learning_rate": 2.95060749049062e-05, + "loss": 0.0715, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016376268211752178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007861958211287856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04061449468135834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.960273277759552, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7670, + "weighted_flow_matching_loss": 0.032523981668055056, + "weighted_rkd_loss": 0.03900102898478508 + }, + { + "epoch": 2.234506837358161, + "flow_matching_loss": 0.0341058949008584, + "grad_norm": 0.414768248796463, + "learning_rate": 2.9503968626406133e-05, + "loss": 0.0725, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01617574440315366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007684773625805974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04003626219928265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9601709485054016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7680, + "weighted_flow_matching_loss": 0.0341058949008584, + "weighted_rkd_loss": 0.038441646844148636 + }, + { + "epoch": 2.237416351469305, + "flow_matching_loss": 0.0330090619623661, + "grad_norm": 0.37044522166252136, + "learning_rate": 2.9501857941983354e-05, + "loss": 0.0753, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0176485207863152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008739791158586741, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04403683245182037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9600684642791748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7690, + "weighted_flow_matching_loss": 0.0330090619623661, + "weighted_rkd_loss": 0.04227833598852158 + }, + { + "epoch": 2.240325865580448, + "flow_matching_loss": 0.03214245922863483, + "grad_norm": 0.3704112470149994, + "learning_rate": 2.9499742852279025e-05, + "loss": 0.0739, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01770283356308937, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008084005070850253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04348967224359512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9599658966064453, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7700, + "weighted_flow_matching_loss": 0.03214245922863483, + "weighted_rkd_loss": 0.04174863323569298 + }, + { + "epoch": 2.2432353796915914, + "flow_matching_loss": 0.03482238557189703, + "grad_norm": 0.33903658390045166, + "learning_rate": 2.9497623357935666e-05, + "loss": 0.0772, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017807562276721, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008496472192928196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04411159604787827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9598631918430328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7710, + "weighted_flow_matching_loss": 0.03482238557189703, + "weighted_rkd_loss": 0.0423411350697279 + }, + { + "epoch": 2.246144893802735, + "flow_matching_loss": 0.03434101659804582, + "grad_norm": 0.3487195372581482, + "learning_rate": 2.9495499459597116e-05, + "loss": 0.0791, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018804820813238622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.009005565149709582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04661520645022392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9597603499889373, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7720, + "weighted_flow_matching_loss": 0.03434101659804582, + "weighted_rkd_loss": 0.0447393823415041 + }, + { + "epoch": 2.2490544079138783, + "flow_matching_loss": 0.0331724276766181, + "grad_norm": 0.3860042691230774, + "learning_rate": 2.9493371157908563e-05, + "loss": 0.0749, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0174700110219419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008519453462213278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04345947541296482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9596573829650878, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7730, + "weighted_flow_matching_loss": 0.0331724276766181, + "weighted_rkd_loss": 0.04170623272657394 + }, + { + "epoch": 2.2519639220250216, + "flow_matching_loss": 0.03692702818661928, + "grad_norm": 0.30321282148361206, + "learning_rate": 2.9491238453516524e-05, + "loss": 0.0758, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01622352544218302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008044855343177915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040491906180977824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9595542848110199, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7740, + "weighted_flow_matching_loss": 0.03692702818661928, + "weighted_rkd_loss": 0.03885423541069031 + }, + { + "epoch": 2.2548734361361653, + "flow_matching_loss": 0.03295403532683849, + "grad_norm": 0.3917720317840576, + "learning_rate": 2.9489101347068868e-05, + "loss": 0.0759, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01808127872645855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008629227569326759, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04479178488254547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9594510912895202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7750, + "weighted_flow_matching_loss": 0.03295403532683849, + "weighted_rkd_loss": 0.042975519225001334 + }, + { + "epoch": 2.2577829502473086, + "flow_matching_loss": 0.031683667190372945, + "grad_norm": 0.3575385808944702, + "learning_rate": 2.948695983921478e-05, + "loss": 0.0748, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01810392439365387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008698430005460977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04490627869963646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9593477547168732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7760, + "weighted_flow_matching_loss": 0.031683667190372945, + "weighted_rkd_loss": 0.04308070167899132 + }, + { + "epoch": 2.2606924643584523, + "flow_matching_loss": 0.032226169109344484, + "grad_norm": 0.35279974341392517, + "learning_rate": 2.9484813930604804e-05, + "loss": 0.074, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017551325634121893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008456052606925368, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04355870373547077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9592442810535431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7770, + "weighted_flow_matching_loss": 0.032226169109344484, + "weighted_rkd_loss": 0.04178343489766121 + }, + { + "epoch": 2.2636019784695955, + "flow_matching_loss": 0.03143025115132332, + "grad_norm": 0.38792046904563904, + "learning_rate": 2.9482663621890804e-05, + "loss": 0.0738, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017780102975666522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00863134958781302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04419155567884445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9591406941413879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7780, + "weighted_flow_matching_loss": 0.03143025115132332, + "weighted_rkd_loss": 0.042385954037308694 + }, + { + "epoch": 2.2665114925807393, + "flow_matching_loss": 0.03320472333580256, + "grad_norm": 0.3365791141986847, + "learning_rate": 2.9480508913725986e-05, + "loss": 0.0755, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01801273413002491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008057197462767363, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04408266507089138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9590369760990143, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7790, + "weighted_flow_matching_loss": 0.03320472333580256, + "weighted_rkd_loss": 0.04227689728140831 + }, + { + "epoch": 2.2694210066918825, + "flow_matching_loss": 0.03262227885425091, + "grad_norm": 0.3856271207332611, + "learning_rate": 2.9478349806764895e-05, + "loss": 0.0751, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017788255773484706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008683844469487667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04426035545766353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9589331209659576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7800, + "weighted_flow_matching_loss": 0.03262227885425091, + "weighted_rkd_loss": 0.04244269020855427 + }, + { + "epoch": 2.2723305208030258, + "flow_matching_loss": 0.03173433095216751, + "grad_norm": 0.3517039120197296, + "learning_rate": 2.9476186301663414e-05, + "loss": 0.0703, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016383878234773875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007494083791971207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04026183970272541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.958829152584076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7810, + "weighted_flow_matching_loss": 0.03173433095216751, + "weighted_rkd_loss": 0.0386041946709156 + }, + { + "epoch": 2.2752400349141695, + "flow_matching_loss": 0.03275179676711559, + "grad_norm": 0.3300600051879883, + "learning_rate": 2.9474018399078752e-05, + "loss": 0.0711, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01610856121405959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007791440654546023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040008562803268435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9587250649929047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7820, + "weighted_flow_matching_loss": 0.03275179676711559, + "weighted_rkd_loss": 0.038357220590114594 + }, + { + "epoch": 2.2781495490253127, + "flow_matching_loss": 0.03194272927939892, + "grad_norm": 0.4135364294052124, + "learning_rate": 2.947184609966947e-05, + "loss": 0.074, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017515008244663476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008806832693517208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04383684918284416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9586208164691925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7830, + "weighted_flow_matching_loss": 0.03194272927939892, + "weighted_rkd_loss": 0.04202284850180149 + }, + { + "epoch": 2.281059063136456, + "flow_matching_loss": 0.03239751067012549, + "grad_norm": 0.3700578510761261, + "learning_rate": 2.9469669404095447e-05, + "loss": 0.074, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01761631332337856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008136961795389652, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04336958825588226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9585164844989776, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7840, + "weighted_flow_matching_loss": 0.03239751067012549, + "weighted_rkd_loss": 0.04157050922513008 + }, + { + "epoch": 2.2839685772475997, + "flow_matching_loss": 0.034364437125623225, + "grad_norm": 0.39839068055152893, + "learning_rate": 2.9467488313017908e-05, + "loss": 0.075, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01702834600582719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008346746116876603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04240343756973743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9584120154380799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7850, + "weighted_flow_matching_loss": 0.034364437125623225, + "weighted_rkd_loss": 0.04063987471163273 + }, + { + "epoch": 2.286878091358743, + "flow_matching_loss": 0.032409033924341205, + "grad_norm": 0.3812481462955475, + "learning_rate": 2.9465302827099412e-05, + "loss": 0.0704, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016034509893506764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007553612254559994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03962263204157353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.958307421207428, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7860, + "weighted_flow_matching_loss": 0.032409033924341205, + "weighted_rkd_loss": 0.037970618903636934 + }, + { + "epoch": 2.2897876054698867, + "flow_matching_loss": 0.03264165427535772, + "grad_norm": 0.3440989553928375, + "learning_rate": 2.9463112947003854e-05, + "loss": 0.0717, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01641871901229024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007931633200496436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04076907150447369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9582026958465576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7870, + "weighted_flow_matching_loss": 0.03264165427535772, + "weighted_rkd_loss": 0.039065053686499596 + }, + { + "epoch": 2.29269711958103, + "flow_matching_loss": 0.03250886332243681, + "grad_norm": 0.33270061016082764, + "learning_rate": 2.9460918673396455e-05, + "loss": 0.0766, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018424030113965272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00913628195412457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045984341949224475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9580978393554688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7880, + "weighted_flow_matching_loss": 0.03250886332243681, + "weighted_rkd_loss": 0.044057544134557244 + }, + { + "epoch": 2.295606633692173, + "flow_matching_loss": 0.034127171710133554, + "grad_norm": 0.3311072289943695, + "learning_rate": 2.9458720006943784e-05, + "loss": 0.0765, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01779647162184119, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008602736750617623, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04419567994773388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9579928636550903, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7890, + "weighted_flow_matching_loss": 0.034127171710133554, + "weighted_rkd_loss": 0.04233925119042396 + }, + { + "epoch": 2.298516147803317, + "flow_matching_loss": 0.03261054363101721, + "grad_norm": 0.3300142288208008, + "learning_rate": 2.9456516948313736e-05, + "loss": 0.0747, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017670369986444712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008615652564913034, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04395639225840568, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9578877627849579, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7900, + "weighted_flow_matching_loss": 0.03261054363101721, + "weighted_rkd_loss": 0.04210542142391205 + }, + { + "epoch": 2.30142566191446, + "flow_matching_loss": 0.03335427343845367, + "grad_norm": 0.3969387114048004, + "learning_rate": 2.9454309498175538e-05, + "loss": 0.0735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01689674789085984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008162226062268019, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04195572137832641, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9577825367450714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7910, + "weighted_flow_matching_loss": 0.03335427343845367, + "weighted_rkd_loss": 0.040184428170323375 + }, + { + "epoch": 2.304335176025604, + "flow_matching_loss": 0.03239164650440216, + "grad_norm": 0.3689165711402893, + "learning_rate": 2.9452097657199762e-05, + "loss": 0.0751, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01792128523811698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00870483135804534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04454740174114704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9576771914958954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7920, + "weighted_flow_matching_loss": 0.03239164650440216, + "weighted_rkd_loss": 0.042662058398127554 + }, + { + "epoch": 2.307244690136747, + "flow_matching_loss": 0.03214245270937681, + "grad_norm": 0.3638937175273895, + "learning_rate": 2.9449881426058303e-05, + "loss": 0.0723, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01690305955708027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008140778820961713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04194689840078354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9575717091560364, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7930, + "weighted_flow_matching_loss": 0.03214245270937681, + "weighted_rkd_loss": 0.040167171880602835 + }, + { + "epoch": 2.3101542042478904, + "flow_matching_loss": 0.031042477302253246, + "grad_norm": 0.35894063115119934, + "learning_rate": 2.9447660805424388e-05, + "loss": 0.0739, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018209440633654594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008323169918730855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044742051512002945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9574661016464233, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7940, + "weighted_flow_matching_loss": 0.031042477302253246, + "weighted_rkd_loss": 0.042838961258530615 + }, + { + "epoch": 2.313063718359034, + "flow_matching_loss": 0.03223907425999641, + "grad_norm": 0.39068061113357544, + "learning_rate": 2.94454357959726e-05, + "loss": 0.0738, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01749518308788538, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008465395122766495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04345576167106628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9573603808879853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7950, + "weighted_flow_matching_loss": 0.03223907425999641, + "weighted_rkd_loss": 0.04160277731716633 + }, + { + "epoch": 2.3159732324701774, + "flow_matching_loss": 0.0340087765827775, + "grad_norm": 0.35537683963775635, + "learning_rate": 2.9443206398378815e-05, + "loss": 0.0751, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01725318068638444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008397577609866858, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04290393926203251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9572545170783997, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7960, + "weighted_flow_matching_loss": 0.0340087765827775, + "weighted_rkd_loss": 0.04107000268995762 + }, + { + "epoch": 2.318882746581321, + "flow_matching_loss": 0.03223864231258631, + "grad_norm": 0.4509991705417633, + "learning_rate": 2.9440972613320282e-05, + "loss": 0.0759, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018393235467374324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008869893429800869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045656364411115646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9571485400199891, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7970, + "weighted_flow_matching_loss": 0.03223864231258631, + "weighted_rkd_loss": 0.04369995817542076 + }, + { + "epoch": 2.3217922606924644, + "flow_matching_loss": 0.03237993065267801, + "grad_norm": 0.34975457191467285, + "learning_rate": 2.943873444147556e-05, + "loss": 0.0714, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016357598826289177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008034673891961575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04074987210333347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9570424377918243, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7980, + "weighted_flow_matching_loss": 0.03237993065267801, + "weighted_rkd_loss": 0.03899936862289906 + }, + { + "epoch": 2.3247017748036076, + "flow_matching_loss": 0.03097814954817295, + "grad_norm": 0.3330213129520416, + "learning_rate": 2.9436491883524547e-05, + "loss": 0.0728, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017651294358074665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008377806656062603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.043680394813418386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9569362103939056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 7990, + "weighted_flow_matching_loss": 0.03097814954817295, + "weighted_rkd_loss": 0.041799357160925865 + }, + { + "epoch": 2.3276112889147513, + "flow_matching_loss": 0.0343880906701088, + "grad_norm": 0.39469173550605774, + "learning_rate": 2.9434244940148472e-05, + "loss": 0.0736, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016616786271333693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007780627394095063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04101420007646084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.956829845905304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8000, + "weighted_flow_matching_loss": 0.0343880906701088, + "weighted_rkd_loss": 0.039243639633059504 + }, + { + "epoch": 2.3305208030258946, + "flow_matching_loss": 0.03519913591444492, + "grad_norm": 0.3035915195941925, + "learning_rate": 2.9431993612029903e-05, + "loss": 0.0769, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017791378404945137, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008035259414464235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04361801669001579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9567233741283416, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8010, + "weighted_flow_matching_loss": 0.03519913591444492, + "weighted_rkd_loss": 0.041730436682701114 + }, + { + "epoch": 2.3334303171370383, + "flow_matching_loss": 0.029313852079212666, + "grad_norm": 0.31892067193984985, + "learning_rate": 2.9429737899852728e-05, + "loss": 0.0682, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016420198511332272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007774695800617337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04061509259045124, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9566167771816254, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8020, + "weighted_flow_matching_loss": 0.029313852079212666, + "weighted_rkd_loss": 0.038853125646710396 + }, + { + "epoch": 2.3363398312481816, + "flow_matching_loss": 0.03248544093221426, + "grad_norm": 0.33221137523651123, + "learning_rate": 2.9427477804302174e-05, + "loss": 0.0706, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01605578828603029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007685505645349621, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03979708254337311, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9565100371837616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8030, + "weighted_flow_matching_loss": 0.03248544093221426, + "weighted_rkd_loss": 0.038066329807043074 + }, + { + "epoch": 2.339249345359325, + "flow_matching_loss": 0.0331841304898262, + "grad_norm": 0.3722480237483978, + "learning_rate": 2.9425213326064797e-05, + "loss": 0.0735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017003228515386583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008182500442489981, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04218895733356476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9564031779766082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8040, + "weighted_flow_matching_loss": 0.0331841304898262, + "weighted_rkd_loss": 0.040349654853343964 + }, + { + "epoch": 2.3421588594704685, + "flow_matching_loss": 0.03284551724791527, + "grad_norm": 0.2905522882938385, + "learning_rate": 2.942294446582849e-05, + "loss": 0.0733, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01704081557691097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008243207493796945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04232483897358179, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9562961995601654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8050, + "weighted_flow_matching_loss": 0.03284551724791527, + "weighted_rkd_loss": 0.04047511611133814 + }, + { + "epoch": 2.345068373581612, + "flow_matching_loss": 0.03242737110704184, + "grad_norm": 0.39243438839912415, + "learning_rate": 2.9420671224282468e-05, + "loss": 0.0715, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01667421357706189, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007561205653473735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04090963266789913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.956189090013504, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8060, + "weighted_flow_matching_loss": 0.03242737110704184, + "weighted_rkd_loss": 0.039117347449064255 + }, + { + "epoch": 2.347977887692755, + "flow_matching_loss": 0.035371820256114, + "grad_norm": 0.37237077951431274, + "learning_rate": 2.9418393602117284e-05, + "loss": 0.0733, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01608869405463338, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007529511442407965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03970689959824085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9560818672180176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8070, + "weighted_flow_matching_loss": 0.035371820256114, + "weighted_rkd_loss": 0.03796304501593113 + }, + { + "epoch": 2.3508874018038988, + "flow_matching_loss": 0.03317793868482113, + "grad_norm": 0.40160128474235535, + "learning_rate": 2.9416111600024816e-05, + "loss": 0.0733, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016995518840849398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007955455221235752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04194649271667004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9559745013713836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8080, + "weighted_flow_matching_loss": 0.03317793868482113, + "weighted_rkd_loss": 0.040099728666245935 + }, + { + "epoch": 2.353796915915042, + "flow_matching_loss": 0.03312157988548279, + "grad_norm": 0.3622114062309265, + "learning_rate": 2.9413825218698284e-05, + "loss": 0.0721, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016323863249272107, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008109649969264865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040757376328110696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9558670163154602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8090, + "weighted_flow_matching_loss": 0.03312157988548279, + "weighted_rkd_loss": 0.038958636298775676 + }, + { + "epoch": 2.3567064300261857, + "flow_matching_loss": 0.03439767304807902, + "grad_norm": 0.38262251019477844, + "learning_rate": 2.9411534458832222e-05, + "loss": 0.0762, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01760722678154707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008518832316622139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04373328648507595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9557594180107116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8100, + "weighted_flow_matching_loss": 0.03439767304807902, + "weighted_rkd_loss": 0.041798485070467 + }, + { + "epoch": 2.359615944137329, + "flow_matching_loss": 0.037336367554962634, + "grad_norm": 0.338910847902298, + "learning_rate": 2.9409239321122513e-05, + "loss": 0.0762, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016442146245390178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007825741497799753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04071003384888172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9556516885757447, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8110, + "weighted_flow_matching_loss": 0.037336367554962634, + "weighted_rkd_loss": 0.03890470415353775 + }, + { + "epoch": 2.3625254582484727, + "flow_matching_loss": 0.033018526434898374, + "grad_norm": 0.3742712140083313, + "learning_rate": 2.940693980626634e-05, + "loss": 0.0725, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016737986728549002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007885318342596292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04136129170656204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9555438339710236, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8120, + "weighted_flow_matching_loss": 0.033018526434898374, + "weighted_rkd_loss": 0.03952254056930542 + }, + { + "epoch": 2.365434972359616, + "flow_matching_loss": 0.03258881289511919, + "grad_norm": 0.38390740752220154, + "learning_rate": 2.9404635914962258e-05, + "loss": 0.0737, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017463153321295978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008120269747450948, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04304657690227032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.955435848236084, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8130, + "weighted_flow_matching_loss": 0.03258881289511919, + "weighted_rkd_loss": 0.04112821370363236 + }, + { + "epoch": 2.3683444864707592, + "flow_matching_loss": 0.03196764010936022, + "grad_norm": 0.3590330183506012, + "learning_rate": 2.9402327647910113e-05, + "loss": 0.0708, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01633642874658108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008015944436192513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040688802301883695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9553277432918549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8140, + "weighted_flow_matching_loss": 0.03196764010936022, + "weighted_rkd_loss": 0.03887106403708458 + }, + { + "epoch": 2.371254000581903, + "flow_matching_loss": 0.03172668982297182, + "grad_norm": 0.3593321740627289, + "learning_rate": 2.94000150058111e-05, + "loss": 0.0708, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01655700085684657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007802619552239776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04091662131249905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9552195072174072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8150, + "weighted_flow_matching_loss": 0.03172668982297182, + "weighted_rkd_loss": 0.03908426761627197 + }, + { + "epoch": 2.374163514693046, + "flow_matching_loss": 0.032727453485131265, + "grad_norm": 0.33852070569992065, + "learning_rate": 2.939769798936774e-05, + "loss": 0.0701, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015994658693671227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007125371368601918, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039114689081907274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9551111459732056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8160, + "weighted_flow_matching_loss": 0.032727453485131265, + "weighted_rkd_loss": 0.03735893275588751 + }, + { + "epoch": 2.3770730288041895, + "flow_matching_loss": 0.030817514657974242, + "grad_norm": 0.35147467255592346, + "learning_rate": 2.9395376599283878e-05, + "loss": 0.0722, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01750267781317234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00833311527967453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04333847090601921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9550026655197144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8170, + "weighted_flow_matching_loss": 0.030817514657974242, + "weighted_rkd_loss": 0.04138827547430992 + }, + { + "epoch": 2.379982542915333, + "flow_matching_loss": 0.03251935392618179, + "grad_norm": 0.39684268832206726, + "learning_rate": 2.9393050836264697e-05, + "loss": 0.0753, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.018103280011564492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00858225142583251, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04478881135582924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9548940658569336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8180, + "weighted_flow_matching_loss": 0.03251935392618179, + "weighted_rkd_loss": 0.042768546007573606 + }, + { + "epoch": 2.3828920570264764, + "flow_matching_loss": 0.03344231881201267, + "grad_norm": 0.3892373740673065, + "learning_rate": 2.9390720701016693e-05, + "loss": 0.0758, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017802283726632596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008717433130368591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.044322000443935396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9547853410243988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8190, + "weighted_flow_matching_loss": 0.03344231881201267, + "weighted_rkd_loss": 0.04231791645288467 + }, + { + "epoch": 2.38580157113762, + "flow_matching_loss": 0.030659361742436886, + "grad_norm": 0.32246243953704834, + "learning_rate": 2.938838619424771e-05, + "loss": 0.068, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015579183213412761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007974681770429015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03913304805755615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9546764850616455, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8200, + "weighted_flow_matching_loss": 0.030659361742436886, + "weighted_rkd_loss": 0.03735946230590344 + }, + { + "epoch": 2.3887110852487634, + "flow_matching_loss": 0.0333844730630517, + "grad_norm": 0.34566640853881836, + "learning_rate": 2.93860473166669e-05, + "loss": 0.0747, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017409549187868835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008485888922587037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04330498799681663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9545675098896027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8210, + "weighted_flow_matching_loss": 0.0333844730630517, + "weighted_rkd_loss": 0.04133747443556786 + }, + { + "epoch": 2.391620599359907, + "flow_matching_loss": 0.02929150704294443, + "grad_norm": 0.31096524000167847, + "learning_rate": 2.938370406898476e-05, + "loss": 0.0651, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015101739950478078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00734920622780919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03755268603563309, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9544584035873414, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8220, + "weighted_flow_matching_loss": 0.02929150704294443, + "weighted_rkd_loss": 0.03584251664578915 + }, + { + "epoch": 2.3945301134710504, + "flow_matching_loss": 0.03065641112625599, + "grad_norm": 0.3466278612613678, + "learning_rate": 2.9381356451913103e-05, + "loss": 0.071, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01712091490626335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007984754955396056, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04222658462822437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9543491721153259, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8230, + "weighted_flow_matching_loss": 0.03065641112625599, + "weighted_rkd_loss": 0.040298952162265776 + }, + { + "epoch": 2.3974396275821936, + "flow_matching_loss": 0.03183155823498964, + "grad_norm": 0.35837361216545105, + "learning_rate": 2.9379004466165072e-05, + "loss": 0.0708, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016648046392947435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0075130736455321315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040809166431427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9542398095130921, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8240, + "weighted_flow_matching_loss": 0.03183155823498964, + "weighted_rkd_loss": 0.03894163966178894 + }, + { + "epoch": 2.4003491416933374, + "flow_matching_loss": 0.030655102618038655, + "grad_norm": 0.3295559883117676, + "learning_rate": 2.937664811245514e-05, + "loss": 0.0677, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015653148479759692, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007542168954387307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03884846642613411, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9541303455829621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8250, + "weighted_flow_matching_loss": 0.030655102618038655, + "weighted_rkd_loss": 0.03706652112305164 + }, + { + "epoch": 2.4032586558044806, + "flow_matching_loss": 0.03382215909659862, + "grad_norm": 0.3443688452243805, + "learning_rate": 2.9374287391499108e-05, + "loss": 0.0729, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016514144837856293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007912105368450283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04094039537012577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9540207386016846, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8260, + "weighted_flow_matching_loss": 0.03382215909659862, + "weighted_rkd_loss": 0.03905801475048065 + }, + { + "epoch": 2.406168169915624, + "flow_matching_loss": 0.03354391232132912, + "grad_norm": 0.3356019854545593, + "learning_rate": 2.9371922304014093e-05, + "loss": 0.0735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017007183376699687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007914035022258759, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04192840196192264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9539110124111175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8270, + "weighted_flow_matching_loss": 0.03354391232132912, + "weighted_rkd_loss": 0.03999602794647217 + }, + { + "epoch": 2.4090776840267676, + "flow_matching_loss": 0.0316433634608984, + "grad_norm": 0.35438263416290283, + "learning_rate": 2.9369552850718557e-05, + "loss": 0.0685, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015740037336945532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072092700749635695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038689345307648185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9538011610507965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8280, + "weighted_flow_matching_loss": 0.0316433634608984, + "weighted_rkd_loss": 0.036901980824768546 + }, + { + "epoch": 2.411987198137911, + "flow_matching_loss": 0.03135728482156992, + "grad_norm": 0.30237123370170593, + "learning_rate": 2.9367179032332262e-05, + "loss": 0.0686, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015848991088569166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007385864434763789, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03908384591341019, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9536911845207214, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8290, + "weighted_flow_matching_loss": 0.03135728482156992, + "weighted_rkd_loss": 0.03727393001317978 + }, + { + "epoch": 2.4148967122490546, + "flow_matching_loss": 0.03384285531938076, + "grad_norm": 0.41712459921836853, + "learning_rate": 2.936480084957633e-05, + "loss": 0.0735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016730746440589427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008122838521376251, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04158433079719544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9535810887813568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8300, + "weighted_flow_matching_loss": 0.03384285531938076, + "weighted_rkd_loss": 0.03965404685586691 + }, + { + "epoch": 2.417806226360198, + "flow_matching_loss": 0.03299286421388388, + "grad_norm": 0.34426429867744446, + "learning_rate": 2.9362418303173174e-05, + "loss": 0.0738, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017097117286175488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00864753145724535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0428417656570673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9534708559513092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8310, + "weighted_flow_matching_loss": 0.03299286421388388, + "weighted_rkd_loss": 0.04084844887256622 + }, + { + "epoch": 2.4207157404713415, + "flow_matching_loss": 0.031493622809648514, + "grad_norm": 0.28287598490715027, + "learning_rate": 2.9360031393846558e-05, + "loss": 0.0691, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01587941888719797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007669433206319809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03942827098071575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.953360515832901, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8320, + "weighted_flow_matching_loss": 0.031493622809648514, + "weighted_rkd_loss": 0.037589406967163085 + }, + { + "epoch": 2.423625254582485, + "flow_matching_loss": 0.030962265096604824, + "grad_norm": 0.32238632440567017, + "learning_rate": 2.9357640122321557e-05, + "loss": 0.0666, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015078532043844461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007187803462147713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03734486810863018, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9532500386238099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8330, + "weighted_flow_matching_loss": 0.030962265096604824, + "weighted_rkd_loss": 0.03559905420988798 + }, + { + "epoch": 2.426534768693628, + "flow_matching_loss": 0.03385345581918955, + "grad_norm": 0.3311382532119751, + "learning_rate": 2.935524448932458e-05, + "loss": 0.0716, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015985317714512347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007665631640702486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039636266976594926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9531394422054291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8340, + "weighted_flow_matching_loss": 0.03385345581918955, + "weighted_rkd_loss": 0.03777885288000107 + }, + { + "epoch": 2.4294442828047718, + "flow_matching_loss": 0.0314749613404274, + "grad_norm": 0.3699542284011841, + "learning_rate": 2.9352844495583356e-05, + "loss": 0.0698, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01625638296827674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0077250645961612465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040237830206751826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9530287206172943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8350, + "weighted_flow_matching_loss": 0.0314749613404274, + "weighted_rkd_loss": 0.03834776058793068 + }, + { + "epoch": 2.432353796915915, + "flow_matching_loss": 0.03201006315648556, + "grad_norm": 0.42538970708847046, + "learning_rate": 2.9350440141826938e-05, + "loss": 0.0757, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01849775779992342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008888492686673998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.045884007960557936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9529178738594055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8360, + "weighted_flow_matching_loss": 0.03201006315648556, + "weighted_rkd_loss": 0.04372363425791263 + }, + { + "epoch": 2.4352633110270583, + "flow_matching_loss": 0.032875629514455794, + "grad_norm": 0.4176555871963501, + "learning_rate": 2.9348031428785705e-05, + "loss": 0.073, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016887539066374302, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0082853093277663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.042060387134552, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9528069019317627, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8370, + "weighted_flow_matching_loss": 0.032875629514455794, + "weighted_rkd_loss": 0.04007549323141575 + }, + { + "epoch": 2.438172825138202, + "flow_matching_loss": 0.032503099925816056, + "grad_norm": 0.3413499891757965, + "learning_rate": 2.9345618357191363e-05, + "loss": 0.0699, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015946335159242153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007406451599672436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03929912224411965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9526958107948303, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8380, + "weighted_flow_matching_loss": 0.032503099925816056, + "weighted_rkd_loss": 0.03744013048708439 + }, + { + "epoch": 2.4410823392493453, + "flow_matching_loss": 0.029756935872137548, + "grad_norm": 0.41291749477386475, + "learning_rate": 2.934320092777694e-05, + "loss": 0.0673, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016044838540256023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007296268502250314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03938594572246075, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9525845944881439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8390, + "weighted_flow_matching_loss": 0.029756935872137548, + "weighted_rkd_loss": 0.037518464773893354 + }, + { + "epoch": 2.443991853360489, + "flow_matching_loss": 0.029634811729192734, + "grad_norm": 0.352495014667511, + "learning_rate": 2.9340779141276782e-05, + "loss": 0.0686, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016599893756210803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007716896152123809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04091668352484703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.952473247051239, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8400, + "weighted_flow_matching_loss": 0.029634811729192734, + "weighted_rkd_loss": 0.038972023874521255 + }, + { + "epoch": 2.4469013674716322, + "flow_matching_loss": 0.031720394641160964, + "grad_norm": 0.35490354895591736, + "learning_rate": 2.933835299842657e-05, + "loss": 0.0704, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016461350210011005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007714531710371375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04063723236322403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9523617804050446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8410, + "weighted_flow_matching_loss": 0.031720394641160964, + "weighted_rkd_loss": 0.03870139457285404 + }, + { + "epoch": 2.4498108815827755, + "flow_matching_loss": 0.03293209914118052, + "grad_norm": 0.38328903913497925, + "learning_rate": 2.93359224999633e-05, + "loss": 0.0716, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01653454964980483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007576291123405099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04064539112150669, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9522502005100251, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8420, + "weighted_flow_matching_loss": 0.03293209914118052, + "weighted_rkd_loss": 0.03870461918413639 + }, + { + "epoch": 2.452720395693919, + "flow_matching_loss": 0.03129397165030241, + "grad_norm": 0.3281248211860657, + "learning_rate": 2.933348764662529e-05, + "loss": 0.0705, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01671609254553914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00773100033402443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.041163185611367226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9521384835243225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8430, + "weighted_flow_matching_loss": 0.03129397165030241, + "weighted_rkd_loss": 0.039193087443709376 + }, + { + "epoch": 2.4556299098050625, + "flow_matching_loss": 0.02969861067831516, + "grad_norm": 0.32725316286087036, + "learning_rate": 2.933104843915219e-05, + "loss": 0.0698, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.017157090548425912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007804694818332791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04211887568235397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9520266473293304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8440, + "weighted_flow_matching_loss": 0.02969861067831516, + "weighted_rkd_loss": 0.04009828492999077 + }, + { + "epoch": 2.458539423916206, + "flow_matching_loss": 0.02918384876102209, + "grad_norm": 0.32282811403274536, + "learning_rate": 2.9328604878284963e-05, + "loss": 0.0703, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01729963654652238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008622796926647424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04322207048535347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9519146680831909, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8450, + "weighted_flow_matching_loss": 0.02918384876102209, + "weighted_rkd_loss": 0.0411437951028347 + }, + { + "epoch": 2.4614489380273494, + "flow_matching_loss": 0.03135015536099672, + "grad_norm": 0.34265145659446716, + "learning_rate": 2.9326156964765892e-05, + "loss": 0.07, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016510880179703237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007535311300307513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040557071566581726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9518025934696197, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8460, + "weighted_flow_matching_loss": 0.03135015536099672, + "weighted_rkd_loss": 0.038602330908179286 + }, + { + "epoch": 2.4643584521384927, + "flow_matching_loss": 0.03251570090651512, + "grad_norm": 0.3800177276134491, + "learning_rate": 2.9323704699338603e-05, + "loss": 0.0692, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015625875815749167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00726536768488586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03851711992174387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9516903817653656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8470, + "weighted_flow_matching_loss": 0.03251570090651512, + "weighted_rkd_loss": 0.03665636144578457 + }, + { + "epoch": 2.4672679662496364, + "flow_matching_loss": 0.03133134599775076, + "grad_norm": 0.32300153374671936, + "learning_rate": 2.932124808274802e-05, + "loss": 0.0711, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016895719710737465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007971699116751552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0417631383985281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9515780448913574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8480, + "weighted_flow_matching_loss": 0.03133134599775076, + "weighted_rkd_loss": 0.039740928635001185 + }, + { + "epoch": 2.4701774803607797, + "flow_matching_loss": 0.03267510272562504, + "grad_norm": 0.32643869519233704, + "learning_rate": 2.93187871157404e-05, + "loss": 0.0715, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016554754693061113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007699566474184394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04080907553434372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9514655888080596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8490, + "weighted_flow_matching_loss": 0.03267510272562504, + "weighted_rkd_loss": 0.0388285830616951 + }, + { + "epoch": 2.4730869944719234, + "flow_matching_loss": 0.0307666577398777, + "grad_norm": 0.33655574917793274, + "learning_rate": 2.931632179906332e-05, + "loss": 0.0675, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015459043625742196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007704068766906857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038622156530618665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9513530194759369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8500, + "weighted_flow_matching_loss": 0.0307666577398777, + "weighted_rkd_loss": 0.03674335982650519 + }, + { + "epoch": 2.4759965085830666, + "flow_matching_loss": 0.03296566773205996, + "grad_norm": 0.3257429599761963, + "learning_rate": 2.931385213346568e-05, + "loss": 0.0758, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01825954597443342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008507826691493392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04502691812813282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9512403070926666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8510, + "weighted_flow_matching_loss": 0.03296566773205996, + "weighted_rkd_loss": 0.0428314421325922 + }, + { + "epoch": 2.47890602269421, + "flow_matching_loss": 0.03421603180468082, + "grad_norm": 0.3504674732685089, + "learning_rate": 2.9311378119697694e-05, + "loss": 0.0723, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016161859314888716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0077400412410497665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04006375968456268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9511274814605712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8520, + "weighted_flow_matching_loss": 0.03421603180468082, + "weighted_rkd_loss": 0.03810568787157535 + }, + { + "epoch": 2.4818155368053536, + "flow_matching_loss": 0.030640304461121558, + "grad_norm": 0.3428850769996643, + "learning_rate": 2.9308899758510903e-05, + "loss": 0.07, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016833482403308155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007741945702582598, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0414089098572731, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9510145246982574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8530, + "weighted_flow_matching_loss": 0.030640304461121558, + "weighted_rkd_loss": 0.03938035778701306 + }, + { + "epoch": 2.484725050916497, + "flow_matching_loss": 0.03270660731941462, + "grad_norm": 0.4067997932434082, + "learning_rate": 2.9306417050658172e-05, + "loss": 0.072, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016620703507214785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00806449274532497, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04130589999258518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9509014546871185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8540, + "weighted_flow_matching_loss": 0.03270660731941462, + "weighted_rkd_loss": 0.03927785791456699 + }, + { + "epoch": 2.4876345650276406, + "flow_matching_loss": 0.031038209050893783, + "grad_norm": 0.33321037888526917, + "learning_rate": 2.9303929996893687e-05, + "loss": 0.0692, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01635372107848525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007428099540993572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04013554118573666, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9507882595062256, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8550, + "weighted_flow_matching_loss": 0.031038209050893783, + "weighted_rkd_loss": 0.03816034384071827 + }, + { + "epoch": 2.490544079138784, + "flow_matching_loss": 0.03344722632318735, + "grad_norm": 0.39429324865341187, + "learning_rate": 2.9301438597972932e-05, + "loss": 0.0726, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016584497597068548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00798619850538671, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04115519281476736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9506749272346496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8560, + "weighted_flow_matching_loss": 0.03344722632318735, + "weighted_rkd_loss": 0.03912527989596128 + }, + { + "epoch": 2.493453593249927, + "flow_matching_loss": 0.03207048084586859, + "grad_norm": 0.33763664960861206, + "learning_rate": 2.9298942854652744e-05, + "loss": 0.0705, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016257830802351235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007924817316234112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04044047966599464, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9505614817142487, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8570, + "weighted_flow_matching_loss": 0.03207048084586859, + "weighted_rkd_loss": 0.03844110146164894 + }, + { + "epoch": 2.496363107361071, + "flow_matching_loss": 0.03346795402467251, + "grad_norm": 0.34617000818252563, + "learning_rate": 2.9296442767691257e-05, + "loss": 0.0724, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016571225505322217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007841621804982424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04098407328128815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9504479169845581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8580, + "weighted_flow_matching_loss": 0.03346795402467251, + "weighted_rkd_loss": 0.03895324356853962 + }, + { + "epoch": 2.499272621472214, + "flow_matching_loss": 0.03113423176109791, + "grad_norm": 0.34472012519836426, + "learning_rate": 2.9293938337847936e-05, + "loss": 0.0694, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01631801212206483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007669024588540197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04030504934489727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9503342092037201, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8590, + "weighted_flow_matching_loss": 0.03113423176109791, + "weighted_rkd_loss": 0.03830324299633503 + }, + { + "epoch": 2.5021821355833573, + "flow_matching_loss": 0.03035027105361223, + "grad_norm": 0.3256068825721741, + "learning_rate": 2.9291429565883554e-05, + "loss": 0.0671, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015616033412516116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007447872264310718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03867993913590908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9502204060554504, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8600, + "weighted_flow_matching_loss": 0.03035027105361223, + "weighted_rkd_loss": 0.03675454370677471 + }, + { + "epoch": 2.505091649694501, + "flow_matching_loss": 0.03159577865153551, + "grad_norm": 0.3551051914691925, + "learning_rate": 2.9288916452560214e-05, + "loss": 0.0712, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01681108046323061, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008069743867963552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04169190414249897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9501064538955688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8610, + "weighted_flow_matching_loss": 0.03159577865153551, + "weighted_rkd_loss": 0.03961167000234127 + }, + { + "epoch": 2.5080011638056443, + "flow_matching_loss": 0.03351524006575346, + "grad_norm": 0.3367409110069275, + "learning_rate": 2.9286398998641334e-05, + "loss": 0.0698, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015540949534624815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007063458673655987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038145358487963674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9499923884868622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8620, + "weighted_flow_matching_loss": 0.03351524006575346, + "weighted_rkd_loss": 0.03623786997050047 + }, + { + "epoch": 2.510910677916788, + "flow_matching_loss": 0.03121789675205946, + "grad_norm": 0.3716631233692169, + "learning_rate": 2.9283877204891645e-05, + "loss": 0.0671, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015412306878715754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0069088835269212725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037733497098088264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.949878203868866, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8630, + "weighted_flow_matching_loss": 0.03121789675205946, + "weighted_rkd_loss": 0.03584218993782997 + }, + { + "epoch": 2.5138201920279313, + "flow_matching_loss": 0.029171780869364738, + "grad_norm": 0.41024622321128845, + "learning_rate": 2.9281351072077208e-05, + "loss": 0.0668, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016109131462872028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007357768109068274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039576031640172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9497638881206513, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8640, + "weighted_flow_matching_loss": 0.029171780869364738, + "weighted_rkd_loss": 0.0375878531485796 + }, + { + "epoch": 2.516729706139075, + "flow_matching_loss": 0.030772607773542404, + "grad_norm": 0.2787534296512604, + "learning_rate": 2.9278820600965393e-05, + "loss": 0.065, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01451915018260479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007001912593841553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036040212772786615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.949649453163147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8650, + "weighted_flow_matching_loss": 0.030772607773542404, + "weighted_rkd_loss": 0.03422566745430231 + }, + { + "epoch": 2.5196392202502182, + "flow_matching_loss": 0.031383704580366614, + "grad_norm": 0.3251194357872009, + "learning_rate": 2.9276285792324887e-05, + "loss": 0.0692, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016077778581529856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0077214238699525595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03987698033452034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9495348989963531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8660, + "weighted_flow_matching_loss": 0.031383704580366614, + "weighted_rkd_loss": 0.03786457292735577 + }, + { + "epoch": 2.5225487343613615, + "flow_matching_loss": 0.0329536022618413, + "grad_norm": 0.32556480169296265, + "learning_rate": 2.9273746646925703e-05, + "loss": 0.0711, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01619242625311017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007757285004481673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04014213792979717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9494202196598053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8670, + "weighted_flow_matching_loss": 0.0329536022618413, + "weighted_rkd_loss": 0.0381117906421423 + }, + { + "epoch": 2.525458248472505, + "flow_matching_loss": 0.029883015155792236, + "grad_norm": 0.2981410324573517, + "learning_rate": 2.9271203165539166e-05, + "loss": 0.0661, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01552324341610074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007083663670346141, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03813015073537827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9493054270744323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8680, + "weighted_flow_matching_loss": 0.029883015155792236, + "weighted_rkd_loss": 0.036197264492511746 + }, + { + "epoch": 2.5283677625836485, + "flow_matching_loss": 0.02890810966491699, + "grad_norm": 0.3457050323486328, + "learning_rate": 2.9268655348937913e-05, + "loss": 0.0662, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015829166024923326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007591396803036332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03924972862005234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9491904973983765, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8690, + "weighted_flow_matching_loss": 0.02890810966491699, + "weighted_rkd_loss": 0.0372555285692215 + }, + { + "epoch": 2.5312772766947917, + "flow_matching_loss": 0.03158041294664145, + "grad_norm": 0.35268205404281616, + "learning_rate": 2.926610319789591e-05, + "loss": 0.0679, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015520113706588744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007214362965896726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03825458995997906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.949075448513031, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8700, + "weighted_flow_matching_loss": 0.03158041294664145, + "weighted_rkd_loss": 0.03630648758262396 + }, + { + "epoch": 2.5341867908059355, + "flow_matching_loss": 0.03198325373232365, + "grad_norm": 0.3139726519584656, + "learning_rate": 2.926354671318843e-05, + "loss": 0.0674, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015106860175728798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007084659021347761, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03729837890714407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9489602625370026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8710, + "weighted_flow_matching_loss": 0.03198325373232365, + "weighted_rkd_loss": 0.03539470359683037 + }, + { + "epoch": 2.5370963049170787, + "flow_matching_loss": 0.029282843135297297, + "grad_norm": 0.32846036553382874, + "learning_rate": 2.926098589559206e-05, + "loss": 0.0653, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015411278884857892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0071235447190701965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03794610258191824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9488449692726135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8720, + "weighted_flow_matching_loss": 0.029282843135297297, + "weighted_rkd_loss": 0.03600497506558895 + }, + { + "epoch": 2.5400058190282224, + "flow_matching_loss": 0.03206729684025049, + "grad_norm": 0.3355114459991455, + "learning_rate": 2.925842074588472e-05, + "loss": 0.0674, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015128908026963473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007017070055007934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03727488629519939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9487295508384704, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8730, + "weighted_flow_matching_loss": 0.03206729684025049, + "weighted_rkd_loss": 0.03536387141793966 + }, + { + "epoch": 2.5429153331393657, + "flow_matching_loss": 0.03471379056572914, + "grad_norm": 0.3647654056549072, + "learning_rate": 2.925585126484563e-05, + "loss": 0.0735, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016486804839223622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007904626009985805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040878236293792725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9486140012741089, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8740, + "weighted_flow_matching_loss": 0.03471379056572914, + "weighted_rkd_loss": 0.03877778053283691 + }, + { + "epoch": 2.5458248472505094, + "flow_matching_loss": 0.03367012459784746, + "grad_norm": 0.35335859656333923, + "learning_rate": 2.9253277453255326e-05, + "loss": 0.0715, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016186448838561775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007486274093389511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03985917195677757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9484983444213867, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8750, + "weighted_flow_matching_loss": 0.03367012459784746, + "weighted_rkd_loss": 0.037806343659758566 + }, + { + "epoch": 2.5487343613616527, + "flow_matching_loss": 0.03284099642187357, + "grad_norm": 0.3477235436439514, + "learning_rate": 2.925069931189567e-05, + "loss": 0.0689, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01537166554480791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007265894953161478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038009225577116015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9483825623989105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8760, + "weighted_flow_matching_loss": 0.03284099642187357, + "weighted_rkd_loss": 0.03604724332690239 + }, + { + "epoch": 2.551643875472796, + "flow_matching_loss": 0.032672013342380526, + "grad_norm": 0.38557174801826477, + "learning_rate": 2.924811684154983e-05, + "loss": 0.0704, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01609064219519496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007657319866120815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039838604629039764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9482666492462158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8770, + "weighted_flow_matching_loss": 0.032672013342380526, + "weighted_rkd_loss": 0.03777758106589317 + }, + { + "epoch": 2.5545533895839396, + "flow_matching_loss": 0.029333700984716417, + "grad_norm": 0.3185732662677765, + "learning_rate": 2.9245530043002293e-05, + "loss": 0.0669, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016071074642241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0074772188439965245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03961936794221401, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9481506168842315, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8780, + "weighted_flow_matching_loss": 0.029333700984716417, + "weighted_rkd_loss": 0.03756520263850689 + }, + { + "epoch": 2.557462903695083, + "flow_matching_loss": 0.03377305306494236, + "grad_norm": 0.3802666664123535, + "learning_rate": 2.9242938917038863e-05, + "loss": 0.0717, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01592505192384124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008121538162231445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039971642196178436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9480344355106354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8790, + "weighted_flow_matching_loss": 0.03377305306494236, + "weighted_rkd_loss": 0.03789451383054256 + }, + { + "epoch": 2.560372417806226, + "flow_matching_loss": 0.03135838471353054, + "grad_norm": 0.31412988901138306, + "learning_rate": 2.9240343464446647e-05, + "loss": 0.0687, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015864082798361778, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007626489037647843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03935465477406978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9479181706905365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8800, + "weighted_flow_matching_loss": 0.03135838471353054, + "weighted_rkd_loss": 0.03730497192591429 + }, + { + "epoch": 2.56328193191737, + "flow_matching_loss": 0.03223849534988403, + "grad_norm": 0.36819788813591003, + "learning_rate": 2.923774368601409e-05, + "loss": 0.0706, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016338331438601018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00777209484949708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04044875781983137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9478017747402191, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8810, + "weighted_flow_matching_loss": 0.03223849534988403, + "weighted_rkd_loss": 0.038337374292314055 + }, + { + "epoch": 2.566191446028513, + "flow_matching_loss": 0.030669821240007878, + "grad_norm": 0.3751709759235382, + "learning_rate": 2.9235139582530918e-05, + "loss": 0.0706, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01694724867120385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008265816699713468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04216031394898891, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9476852476596832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8820, + "weighted_flow_matching_loss": 0.030669821240007878, + "weighted_rkd_loss": 0.03995478600263595 + }, + { + "epoch": 2.569100960139657, + "flow_matching_loss": 0.029130406863987445, + "grad_norm": 0.42053133249282837, + "learning_rate": 2.9232531154788203e-05, + "loss": 0.066, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01583139905706048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007235014904290437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03889781311154365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9475686013698578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8830, + "weighted_flow_matching_loss": 0.029130406863987445, + "weighted_rkd_loss": 0.03685825690627098 + }, + { + "epoch": 2.5720104742508, + "flow_matching_loss": 0.03000356797128916, + "grad_norm": 0.3339448869228363, + "learning_rate": 2.9229918403578313e-05, + "loss": 0.0686, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016430341824889182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007902201637625695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04076288528740406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9474518299102783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8840, + "weighted_flow_matching_loss": 0.03000356797128916, + "weighted_rkd_loss": 0.038620921596884725 + }, + { + "epoch": 2.574919988361944, + "flow_matching_loss": 0.029541725851595403, + "grad_norm": 0.31496599316596985, + "learning_rate": 2.922730132969493e-05, + "loss": 0.0658, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015302497800439597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007636875566095114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038241871446371076, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9473349392414093, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8850, + "weighted_flow_matching_loss": 0.029541725851595403, + "weighted_rkd_loss": 0.03622781094163656 + }, + { + "epoch": 2.577829502473087, + "flow_matching_loss": 0.029965277947485446, + "grad_norm": 0.35945355892181396, + "learning_rate": 2.922467993393305e-05, + "loss": 0.0671, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015822432842105626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00751978293992579, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03916464857757092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9472179174423218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8860, + "weighted_flow_matching_loss": 0.029965277947485446, + "weighted_rkd_loss": 0.03709740899503231 + }, + { + "epoch": 2.5807390165842303, + "flow_matching_loss": 0.03135298397392035, + "grad_norm": 0.33642885088920593, + "learning_rate": 2.9222054217088985e-05, + "loss": 0.0686, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015884397830814124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007589814299717545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03935861028730869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9471007704734802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8870, + "weighted_flow_matching_loss": 0.03135298397392035, + "weighted_rkd_loss": 0.03727648817002773 + }, + { + "epoch": 2.583648530695374, + "flow_matching_loss": 0.030701796151697637, + "grad_norm": 0.4107174575328827, + "learning_rate": 2.921942417996037e-05, + "loss": 0.0674, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015812208503484727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00715740080922842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03878181762993336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9469835162162781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8880, + "weighted_flow_matching_loss": 0.030701796151697637, + "weighted_rkd_loss": 0.03672565463930368 + }, + { + "epoch": 2.5865580448065173, + "flow_matching_loss": 0.03228181954473257, + "grad_norm": 0.3306277096271515, + "learning_rate": 2.921678982334612e-05, + "loss": 0.0695, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015767983347177505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007790659507736564, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03932662606239319, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9468661248683929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8890, + "weighted_flow_matching_loss": 0.03228181954473257, + "weighted_rkd_loss": 0.03723710812628269 + }, + { + "epoch": 2.5894675589176606, + "flow_matching_loss": 0.03121182955801487, + "grad_norm": 0.364623099565506, + "learning_rate": 2.92141511480465e-05, + "loss": 0.0685, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015961151476949452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0074697303585708145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.039392033405601976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9467486262321472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8900, + "weighted_flow_matching_loss": 0.03121182955801487, + "weighted_rkd_loss": 0.03729437962174416 + }, + { + "epoch": 2.5923770730288043, + "flow_matching_loss": 0.030199277587234973, + "grad_norm": 0.3631976544857025, + "learning_rate": 2.9211508154863066e-05, + "loss": 0.0685, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01640949947759509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007659187074750662, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040478186681866646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9466309964656829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8910, + "weighted_flow_matching_loss": 0.030199277587234973, + "weighted_rkd_loss": 0.03831789270043373 + }, + { + "epoch": 2.5952865871399475, + "flow_matching_loss": 0.03175779897719622, + "grad_norm": 0.3393990397453308, + "learning_rate": 2.9208860844598684e-05, + "loss": 0.0657, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014514350891113281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00684172180481255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0358704237267375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9465132415294647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8920, + "weighted_flow_matching_loss": 0.03175779897719622, + "weighted_rkd_loss": 0.033951827138662335 + }, + { + "epoch": 2.598196101251091, + "flow_matching_loss": 0.032794547453522685, + "grad_norm": 0.2911950945854187, + "learning_rate": 2.9206209218057543e-05, + "loss": 0.0689, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015455152746289969, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007205494679510593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03811580017209053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9463953733444214, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8930, + "weighted_flow_matching_loss": 0.032794547453522685, + "weighted_rkd_loss": 0.036072700284421444 + }, + { + "epoch": 2.6011056153622345, + "flow_matching_loss": 0.02976925075054169, + "grad_norm": 0.3344510793685913, + "learning_rate": 2.9203553276045135e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015293342340737582, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00705149001441896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03763817492872477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9462773740291596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8940, + "weighted_flow_matching_loss": 0.02976925075054169, + "weighted_rkd_loss": 0.035616178065538406 + }, + { + "epoch": 2.604015129473378, + "flow_matching_loss": 0.031793587282299994, + "grad_norm": 0.41457560658454895, + "learning_rate": 2.9200893019368263e-05, + "loss": 0.0685, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015696434397250415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007402102556079626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03879497144371271, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9461592555046081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8950, + "weighted_flow_matching_loss": 0.031793587282299994, + "weighted_rkd_loss": 0.03670618236064911 + }, + { + "epoch": 2.6069246435845215, + "flow_matching_loss": 0.028406400047242642, + "grad_norm": 0.33446139097213745, + "learning_rate": 2.9198228448835045e-05, + "loss": 0.0665, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016225469578057528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007824668707326055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04027560800313949, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9460410118103028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8960, + "weighted_flow_matching_loss": 0.028406400047242642, + "weighted_rkd_loss": 0.03810230977833271 + }, + { + "epoch": 2.6098341576956647, + "flow_matching_loss": 0.029612204246222972, + "grad_norm": 0.36200258135795593, + "learning_rate": 2.9195559565254908e-05, + "loss": 0.0662, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01565167512744665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007410589326173067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038713938929140566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9459226489067077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8970, + "weighted_flow_matching_loss": 0.029612204246222972, + "weighted_rkd_loss": 0.036620409972965715 + }, + { + "epoch": 2.6127436718068084, + "flow_matching_loss": 0.032059003785252574, + "grad_norm": 0.34872445464134216, + "learning_rate": 2.919288636943858e-05, + "loss": 0.0721, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01699757482856512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008343560062348843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04233871009200811, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9458041727542877, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8980, + "weighted_flow_matching_loss": 0.032059003785252574, + "weighted_rkd_loss": 0.04004411660134792 + }, + { + "epoch": 2.6156531859179517, + "flow_matching_loss": 0.034695441462099554, + "grad_norm": 0.3376346230506897, + "learning_rate": 2.9190208862198126e-05, + "loss": 0.0708, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01548608709126711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007210404612123966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0381825789809227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9456855595111847, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 8990, + "weighted_flow_matching_loss": 0.034695441462099554, + "weighted_rkd_loss": 0.03610873930156231 + }, + { + "epoch": 2.618562700029095, + "flow_matching_loss": 0.030159938707947732, + "grad_norm": 0.3708800971508026, + "learning_rate": 2.9187527044346883e-05, + "loss": 0.0648, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014805644564330579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006978504871949553, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03658979460597038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9455668270587921, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9000, + "weighted_flow_matching_loss": 0.030159938707947732, + "weighted_rkd_loss": 0.03459806460887194 + }, + { + "epoch": 2.6214722141402387, + "flow_matching_loss": 0.030403330363333224, + "grad_norm": 0.321186363697052, + "learning_rate": 2.918484091669953e-05, + "loss": 0.0668, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01562965465709567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007238580379635096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03849788978695869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9454479813575745, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9010, + "weighted_flow_matching_loss": 0.030403330363333224, + "weighted_rkd_loss": 0.03639778103679418 + }, + { + "epoch": 2.624381728251382, + "flow_matching_loss": 0.0288494361564517, + "grad_norm": 0.32818150520324707, + "learning_rate": 2.9182150480072026e-05, + "loss": 0.06, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013451970741152763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006086360337212682, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03299030177295208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9453289985656739, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9020, + "weighted_flow_matching_loss": 0.0288494361564517, + "weighted_rkd_loss": 0.031186701357364656 + }, + { + "epoch": 2.627291242362525, + "flow_matching_loss": 0.029924824088811874, + "grad_norm": 0.34332579374313354, + "learning_rate": 2.917945573528167e-05, + "loss": 0.0639, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01464094640687108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006657069968059659, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03593896273523569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9452099084854126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9030, + "weighted_flow_matching_loss": 0.029924824088811874, + "weighted_rkd_loss": 0.033969808369874954 + }, + { + "epoch": 2.630200756473669, + "flow_matching_loss": 0.03432887643575668, + "grad_norm": 0.32042166590690613, + "learning_rate": 2.9176756683147052e-05, + "loss": 0.0706, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01558813378214836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007174521591514349, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038350789621472356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9450906872749328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9040, + "weighted_flow_matching_loss": 0.03432887643575668, + "weighted_rkd_loss": 0.036245000921189784 + }, + { + "epoch": 2.633110270584812, + "flow_matching_loss": 0.032067866437137126, + "grad_norm": 0.3223123550415039, + "learning_rate": 2.9174053324488063e-05, + "loss": 0.0682, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015508077945560217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072584254667162895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03827458135783672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9449713468551636, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9050, + "weighted_flow_matching_loss": 0.032067866437137126, + "weighted_rkd_loss": 0.03616839125752449 + }, + { + "epoch": 2.636019784695956, + "flow_matching_loss": 0.03171088732779026, + "grad_norm": 0.3288215398788452, + "learning_rate": 2.917134566012592e-05, + "loss": 0.0683, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015887757949531078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006994518218562007, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03877003379166126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9448518753051758, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9060, + "weighted_flow_matching_loss": 0.03171088732779026, + "weighted_rkd_loss": 0.03663195930421352 + }, + { + "epoch": 2.638929298807099, + "flow_matching_loss": 0.030365868471562862, + "grad_norm": 0.2890784740447998, + "learning_rate": 2.916863369088314e-05, + "loss": 0.0679, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016168644279241563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007435341458767652, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03977263048291206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9447322964668274, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9070, + "weighted_flow_matching_loss": 0.030365868471562862, + "weighted_rkd_loss": 0.03757447525858879 + }, + { + "epoch": 2.641838812918243, + "flow_matching_loss": 0.030131563171744348, + "grad_norm": 0.3275986611843109, + "learning_rate": 2.9165917417583544e-05, + "loss": 0.0655, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015177699737250805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007117756502702832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0374731557443738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.944612592458725, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9080, + "weighted_flow_matching_loss": 0.030131563171744348, + "weighted_rkd_loss": 0.035397710651159285 + }, + { + "epoch": 2.644748327029386, + "flow_matching_loss": 0.030775259248912334, + "grad_norm": 0.3521469235420227, + "learning_rate": 2.916319684105227e-05, + "loss": 0.0675, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01575532918795943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007400295231491327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03891095370054245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9444927632808685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9090, + "weighted_flow_matching_loss": 0.030775259248912334, + "weighted_rkd_loss": 0.036751059256494044 + }, + { + "epoch": 2.6476578411405294, + "flow_matching_loss": 0.03023607712239027, + "grad_norm": 0.30129364132881165, + "learning_rate": 2.916047196211575e-05, + "loss": 0.0641, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01452222689986229, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0067969643045216795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035841417871415614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9443728148937225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9100, + "weighted_flow_matching_loss": 0.03023607712239027, + "weighted_rkd_loss": 0.03384772129356861 + }, + { + "epoch": 2.650567355251673, + "flow_matching_loss": 0.031124426610767843, + "grad_norm": 0.3401978015899658, + "learning_rate": 2.915774278160173e-05, + "loss": 0.0666, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015177742764353751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007247840752825141, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03760332632809878, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.944252747297287, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9110, + "weighted_flow_matching_loss": 0.031124426610767843, + "weighted_rkd_loss": 0.035506981424987316 + }, + { + "epoch": 2.6534768693628163, + "flow_matching_loss": 0.03144932296127081, + "grad_norm": 0.30977863073349, + "learning_rate": 2.9155009300339268e-05, + "loss": 0.0653, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014473723620176316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006894997647032141, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03584244456142187, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9441325485706329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9120, + "weighted_flow_matching_loss": 0.03144932296127081, + "weighted_rkd_loss": 0.03383996747434139 + }, + { + "epoch": 2.6563863834739596, + "flow_matching_loss": 0.030576928332448006, + "grad_norm": 0.33857786655426025, + "learning_rate": 2.9152271519158724e-05, + "loss": 0.0635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014202354662120342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006459996616467833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034864706173539164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9440122246742249, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9130, + "weighted_flow_matching_loss": 0.030576928332448006, + "weighted_rkd_loss": 0.032912643440067765 + }, + { + "epoch": 2.6592958975851033, + "flow_matching_loss": 0.031104054115712643, + "grad_norm": 0.3108706772327423, + "learning_rate": 2.9149529438891764e-05, + "loss": 0.0655, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014800366014242172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006844427390024066, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03644515909254551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9438917994499206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9140, + "weighted_flow_matching_loss": 0.031104054115712643, + "weighted_rkd_loss": 0.034400316141545775 + }, + { + "epoch": 2.6622054116962466, + "flow_matching_loss": 0.030883317813277245, + "grad_norm": 0.29013359546661377, + "learning_rate": 2.9146783060371352e-05, + "loss": 0.0645, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014344841614365578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006893632421270013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03558331560343504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9437712371349335, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9150, + "weighted_flow_matching_loss": 0.030883317813277245, + "weighted_rkd_loss": 0.033582530729472634 + }, + { + "epoch": 2.6651149258073903, + "flow_matching_loss": 0.03223025649785995, + "grad_norm": 0.31918755173683167, + "learning_rate": 2.914403238443177e-05, + "loss": 0.069, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015687570814043282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007575986254960298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03895112760365009, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9436505615711213, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9160, + "weighted_flow_matching_loss": 0.03223025649785995, + "weighted_rkd_loss": 0.03675616625696421 + }, + { + "epoch": 2.6680244399185336, + "flow_matching_loss": 0.03286211341619492, + "grad_norm": 0.33021536469459534, + "learning_rate": 2.9141277411908607e-05, + "loss": 0.0704, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01595820188522339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007878659339621664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03979506343603134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9435297548770905, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9170, + "weighted_flow_matching_loss": 0.03286211341619492, + "weighted_rkd_loss": 0.03754782900214195 + }, + { + "epoch": 2.6709339540296773, + "flow_matching_loss": 0.0313706811517477, + "grad_norm": 0.3429586589336395, + "learning_rate": 2.9138518143638744e-05, + "loss": 0.0664, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014943303540349007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007282812241464853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03716941922903061, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9434088170528412, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9180, + "weighted_flow_matching_loss": 0.0313706811517477, + "weighted_rkd_loss": 0.03506594505161047 + }, + { + "epoch": 2.6738434681408205, + "flow_matching_loss": 0.030599653907120227, + "grad_norm": 0.3643716871738434, + "learning_rate": 2.9135754580460373e-05, + "loss": 0.0671, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01567228203639388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0073511630296707155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03869572728872299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9432877838611603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9190, + "weighted_flow_matching_loss": 0.030599653907120227, + "weighted_rkd_loss": 0.03650118559598923 + }, + { + "epoch": 2.676752982251964, + "flow_matching_loss": 0.03134799748659134, + "grad_norm": 0.32405778765678406, + "learning_rate": 2.9132986723212992e-05, + "loss": 0.0632, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01369286896660924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006393544655293226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033779282867908475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9431666195392608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9200, + "weighted_flow_matching_loss": 0.03134799748659134, + "weighted_rkd_loss": 0.0318595252931118 + }, + { + "epoch": 2.6796624963631075, + "flow_matching_loss": 0.028905380330979825, + "grad_norm": 0.29557082056999207, + "learning_rate": 2.9130214572737403e-05, + "loss": 0.0644, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015342505183070898, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006914250692352653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03759926110506058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9430453300476074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9210, + "weighted_flow_matching_loss": 0.028905380330979825, + "weighted_rkd_loss": 0.035457808524370193 + }, + { + "epoch": 2.6825720104742508, + "flow_matching_loss": 0.03207075335085392, + "grad_norm": 0.36121273040771484, + "learning_rate": 2.9127438129875717e-05, + "loss": 0.0706, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016586375515908004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007678645197302103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0408513966947794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9429239213466645, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9220, + "weighted_flow_matching_loss": 0.03207075335085392, + "weighted_rkd_loss": 0.03851975277066231 + }, + { + "epoch": 2.685481524585394, + "flow_matching_loss": 0.03157002180814743, + "grad_norm": 0.3834008276462555, + "learning_rate": 2.912465739547134e-05, + "loss": 0.066, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014772657956928014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006992774736136198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03653809130191803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.942802369594574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9230, + "weighted_flow_matching_loss": 0.03157002180814743, + "weighted_rkd_loss": 0.03444825708866119 + }, + { + "epoch": 2.6883910386965377, + "flow_matching_loss": 0.031744104251265524, + "grad_norm": 0.3662964105606079, + "learning_rate": 2.9121872370368988e-05, + "loss": 0.0678, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015554120950400829, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00716682942584157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03827507123351097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9426807403564453, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9240, + "weighted_flow_matching_loss": 0.031744104251265524, + "weighted_rkd_loss": 0.036081216856837274 + }, + { + "epoch": 2.691300552807681, + "flow_matching_loss": 0.031910629384219645, + "grad_norm": 0.3469723165035248, + "learning_rate": 2.911908305541467e-05, + "loss": 0.0672, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01508536422625184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007224446255713701, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03739517517387867, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9425589621067048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9250, + "weighted_flow_matching_loss": 0.031910629384219645, + "weighted_rkd_loss": 0.03524711132049561 + }, + { + "epoch": 2.6942100669188247, + "flow_matching_loss": 0.029885956645011903, + "grad_norm": 0.341014564037323, + "learning_rate": 2.9116289451455717e-05, + "loss": 0.0647, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015028534457087517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00693254922516644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03698961846530437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9424370646476745, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9260, + "weighted_flow_matching_loss": 0.029885956645011903, + "weighted_rkd_loss": 0.034860354475677015 + }, + { + "epoch": 2.697119581029968, + "flow_matching_loss": 0.030749919638037682, + "grad_norm": 0.400121808052063, + "learning_rate": 2.911349155934075e-05, + "loss": 0.0667, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015378513559699058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007426677038893104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038183704018592834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9423150479793548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9270, + "weighted_flow_matching_loss": 0.030749919638037682, + "weighted_rkd_loss": 0.03598114848136902 + }, + { + "epoch": 2.7000290951411117, + "flow_matching_loss": 0.02976689115166664, + "grad_norm": 0.28326717019081116, + "learning_rate": 2.9110689379919687e-05, + "loss": 0.0679, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016435099486261606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007600865745916963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04047106448560953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9421929180622101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9280, + "weighted_flow_matching_loss": 0.02976689115166664, + "weighted_rkd_loss": 0.03813157007098198 + }, + { + "epoch": 2.702938609252255, + "flow_matching_loss": 0.029293394461274148, + "grad_norm": 0.29898801445961, + "learning_rate": 2.9107882914043767e-05, + "loss": 0.0635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014825899340212344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00670571462251246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03635751344263553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9420706570148468, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9290, + "weighted_flow_matching_loss": 0.029293394461274148, + "weighted_rkd_loss": 0.03425145372748375 + }, + { + "epoch": 2.705848123363398, + "flow_matching_loss": 0.029669805616140365, + "grad_norm": 0.373801052570343, + "learning_rate": 2.910507216256551e-05, + "loss": 0.0652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015227878093719482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007311964919790625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037767721153795716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9419482767581939, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9300, + "weighted_flow_matching_loss": 0.029669805616140365, + "weighted_rkd_loss": 0.03557521980255842 + }, + { + "epoch": 2.708757637474542, + "flow_matching_loss": 0.032246177271008494, + "grad_norm": 0.36282047629356384, + "learning_rate": 2.910225712633876e-05, + "loss": 0.0672, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014960002433508635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007153770420700312, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037073774822056295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9418257713317871, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9310, + "weighted_flow_matching_loss": 0.032246177271008494, + "weighted_rkd_loss": 0.034916935861110686 + }, + { + "epoch": 2.711667151585685, + "flow_matching_loss": 0.030143758095800877, + "grad_norm": 0.32500892877578735, + "learning_rate": 2.9099437806218637e-05, + "loss": 0.0669, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015762132499366997, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007494007796049118, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0390182726085186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9417031466960907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9320, + "weighted_flow_matching_loss": 0.030143758095800877, + "weighted_rkd_loss": 0.036743636801838876 + }, + { + "epoch": 2.7145766656968284, + "flow_matching_loss": 0.029041655361652374, + "grad_norm": 0.39328625798225403, + "learning_rate": 2.9096614203061584e-05, + "loss": 0.0622, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014350051991641521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00647445316426456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03517455756664276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9415804147720337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9330, + "weighted_flow_matching_loss": 0.029041655361652374, + "weighted_rkd_loss": 0.033119716495275495 + }, + { + "epoch": 2.717486179807972, + "flow_matching_loss": 0.032893828302621844, + "grad_norm": 0.3721657991409302, + "learning_rate": 2.9093786317725336e-05, + "loss": 0.0659, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014141947310417891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006730757560580969, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03501465227454901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9414575338363648, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9340, + "weighted_flow_matching_loss": 0.032893828302621844, + "weighted_rkd_loss": 0.03296483065932989 + }, + { + "epoch": 2.7203956939191154, + "flow_matching_loss": 0.030290868878364564, + "grad_norm": 0.32276174426078796, + "learning_rate": 2.909095415106893e-05, + "loss": 0.0675, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01598325576633215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007610092498362064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03957660347223282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9413345575332641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9350, + "weighted_flow_matching_loss": 0.030290868878364564, + "weighted_rkd_loss": 0.03725476861000061 + }, + { + "epoch": 2.723305208030259, + "flow_matching_loss": 0.03026512321084738, + "grad_norm": 0.2682429552078247, + "learning_rate": 2.9088117703952698e-05, + "loss": 0.0688, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01639270978048444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.008129267347976565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.040914686396718025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9412114381790161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9360, + "weighted_flow_matching_loss": 0.03026512321084738, + "weighted_rkd_loss": 0.03850948940962553 + }, + { + "epoch": 2.7262147221414024, + "flow_matching_loss": 0.03122942801564932, + "grad_norm": 0.3471699357032776, + "learning_rate": 2.908527697723829e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0148119505494833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006662767473608255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0362866684794426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9410882115364074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9370, + "weighted_flow_matching_loss": 0.03122942801564932, + "weighted_rkd_loss": 0.034149041026830675 + }, + { + "epoch": 2.729124236252546, + "flow_matching_loss": 0.030575170740485193, + "grad_norm": 0.3039606809616089, + "learning_rate": 2.9082431971788635e-05, + "loss": 0.0649, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014918408077210187, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00667434991337359, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03651116564869881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9409648478031158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9380, + "weighted_flow_matching_loss": 0.030575170740485193, + "weighted_rkd_loss": 0.03435573969036341 + }, + { + "epoch": 2.7320337503636893, + "flow_matching_loss": 0.03071348648518324, + "grad_norm": 0.33582058548927307, + "learning_rate": 2.9079582688467972e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0149864356033504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006873585982248187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03684645742177963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.940841406583786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9390, + "weighted_flow_matching_loss": 0.03071348648518324, + "weighted_rkd_loss": 0.03466672189533711 + }, + { + "epoch": 2.7349432644748326, + "flow_matching_loss": 0.032586092688143255, + "grad_norm": 0.4190919101238251, + "learning_rate": 2.9076729128141833e-05, + "loss": 0.0686, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015503214206546545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007232724642381072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03823915310204029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9407178103923798, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9400, + "weighted_flow_matching_loss": 0.032586092688143255, + "weighted_rkd_loss": 0.03597216252237558 + }, + { + "epoch": 2.7378527785859763, + "flow_matching_loss": 0.027960641868412496, + "grad_norm": 0.2673741579055786, + "learning_rate": 2.9073871291677063e-05, + "loss": 0.0606, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01415319349616766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006416462501510978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034722849167883395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9405941069126129, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9410, + "weighted_flow_matching_loss": 0.027960641868412496, + "weighted_rkd_loss": 0.032660195231437684 + }, + { + "epoch": 2.7407622926971196, + "flow_matching_loss": 0.028779087774455547, + "grad_norm": 0.3511348068714142, + "learning_rate": 2.9071009179941797e-05, + "loss": 0.0629, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014844970125705004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006629444193094969, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03631938397884369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9404702603816986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9420, + "weighted_flow_matching_loss": 0.028779087774455547, + "weighted_rkd_loss": 0.03415738213807344 + }, + { + "epoch": 2.743671806808263, + "flow_matching_loss": 0.030129502341151238, + "grad_norm": 0.3409799635410309, + "learning_rate": 2.906814279380546e-05, + "loss": 0.0641, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014547939691692591, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006984902825206518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036080782487988475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9403463006019592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9430, + "weighted_flow_matching_loss": 0.030129502341151238, + "weighted_rkd_loss": 0.0339283624663949 + }, + { + "epoch": 2.7465813209194065, + "flow_matching_loss": 0.027216020040214062, + "grad_norm": 0.32276609539985657, + "learning_rate": 2.9065272134138795e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014348087087273597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006499780435115099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035195954516530036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9402222394943237, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9440, + "weighted_flow_matching_loss": 0.027216020040214062, + "weighted_rkd_loss": 0.03309203069657087 + }, + { + "epoch": 2.74949083503055, + "flow_matching_loss": 0.030031227692961692, + "grad_norm": 0.3133694529533386, + "learning_rate": 2.9062397201813823e-05, + "loss": 0.062, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013839288987219334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006346400268375874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034024978056550026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9400980472564697, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9450, + "weighted_flow_matching_loss": 0.030031227692961692, + "weighted_rkd_loss": 0.031986748799681664 + }, + { + "epoch": 2.752400349141693, + "flow_matching_loss": 0.030541598238050936, + "grad_norm": 0.34316909313201904, + "learning_rate": 2.905951799770388e-05, + "loss": 0.0651, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014982702862471341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006825071386992931, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03679047673940659, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9399737298488617, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9460, + "weighted_flow_matching_loss": 0.030541598238050936, + "weighted_rkd_loss": 0.03458205331116915 + }, + { + "epoch": 2.7553098632528368, + "flow_matching_loss": 0.03175279926508665, + "grad_norm": 0.2961534261703491, + "learning_rate": 2.9056634522683586e-05, + "loss": 0.0699, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.016443268582224845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007731553073972464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.04061809033155441, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9398492932319641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9470, + "weighted_flow_matching_loss": 0.03175279926508665, + "weighted_rkd_loss": 0.03817494213581085 + }, + { + "epoch": 2.7582193773639805, + "flow_matching_loss": 0.031183403544127943, + "grad_norm": 0.2947844862937927, + "learning_rate": 2.9053746777628872e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014779418706893921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00681997868232429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03637881614267826, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9397247433662415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9480, + "weighted_flow_matching_loss": 0.031183403544127943, + "weighted_rkd_loss": 0.03418610598891973 + }, + { + "epoch": 2.7611288914751237, + "flow_matching_loss": 0.03230531457811594, + "grad_norm": 0.3275122344493866, + "learning_rate": 2.9050854763416952e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014199087768793106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006851139990612864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03524931594729423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9396000623703002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9490, + "weighted_flow_matching_loss": 0.03230531457811594, + "weighted_rkd_loss": 0.03312033507972956 + }, + { + "epoch": 2.764038405586267, + "flow_matching_loss": 0.033082169480621815, + "grad_norm": 0.35095158219337463, + "learning_rate": 2.9047958480926346e-05, + "loss": 0.0681, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015002227388322353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007264521438628435, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03726897612214088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9394752681255341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9500, + "weighted_flow_matching_loss": 0.033082169480621815, + "weighted_rkd_loss": 0.03501324579119682 + }, + { + "epoch": 2.7669479196974107, + "flow_matching_loss": 0.029381787218153477, + "grad_norm": 0.29321616888046265, + "learning_rate": 2.9045057931036874e-05, + "loss": 0.0634, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014717277232557535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0068059480283409355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036240502074360845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9393503427505493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9510, + "weighted_flow_matching_loss": 0.029381787218153477, + "weighted_rkd_loss": 0.034042591042816636 + }, + { + "epoch": 2.769857433808554, + "flow_matching_loss": 0.03427652046084404, + "grad_norm": 0.3344965875148773, + "learning_rate": 2.9042153114629636e-05, + "loss": 0.0679, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014525867532938719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006793045811355114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03584478087723255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9392253041267395, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9520, + "weighted_flow_matching_loss": 0.03427652046084404, + "weighted_rkd_loss": 0.03366635888814926 + }, + { + "epoch": 2.7727669479196972, + "flow_matching_loss": 0.03170732092112303, + "grad_norm": 0.39330005645751953, + "learning_rate": 2.9039244032587043e-05, + "loss": 0.0668, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015040042251348496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0072375483810901645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037317633256316186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9391001462936401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9530, + "weighted_flow_matching_loss": 0.03170732092112303, + "weighted_rkd_loss": 0.035044912993907926 + }, + { + "epoch": 2.775676462030841, + "flow_matching_loss": 0.03292042687535286, + "grad_norm": 0.3558215796947479, + "learning_rate": 2.90363306857928e-05, + "loss": 0.0671, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014703556243330241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006947853090241552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03635496590286493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9389748692512512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9540, + "weighted_flow_matching_loss": 0.03292042687535286, + "weighted_rkd_loss": 0.03413640838116407 + }, + { + "epoch": 2.778585976141984, + "flow_matching_loss": 0.028976745903491974, + "grad_norm": 0.32313886284828186, + "learning_rate": 2.90334130751319e-05, + "loss": 0.065, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015673670638352632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007053341995924711, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03840068355202675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9388494729995728, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9550, + "weighted_flow_matching_loss": 0.028976745903491974, + "weighted_rkd_loss": 0.03605242036283016 + }, + { + "epoch": 2.7814954902531275, + "flow_matching_loss": 0.02980579826980829, + "grad_norm": 0.30328795313835144, + "learning_rate": 2.903049120149064e-05, + "loss": 0.0634, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014432849083095788, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006925751175731421, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03579144962131977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9387239515781403, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9560, + "weighted_flow_matching_loss": 0.02980579826980829, + "weighted_rkd_loss": 0.03359828852117062 + }, + { + "epoch": 2.784405004364271, + "flow_matching_loss": 0.031885413080453874, + "grad_norm": 0.3498384654521942, + "learning_rate": 2.9027565065756608e-05, + "loss": 0.0652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0143937555141747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006656551081687212, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03544406183063984, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9385982990264893, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9570, + "weighted_flow_matching_loss": 0.031885413080453874, + "weighted_rkd_loss": 0.03326768893748522 + }, + { + "epoch": 2.7873145184754144, + "flow_matching_loss": 0.03194235451519489, + "grad_norm": 0.3375087380409241, + "learning_rate": 2.9024634668818682e-05, + "loss": 0.0669, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015051598660647869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007131310552358628, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03723450712859631, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9384725391864777, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9580, + "weighted_flow_matching_loss": 0.03194235451519489, + "weighted_rkd_loss": 0.034943592734634876 + }, + { + "epoch": 2.790224032586558, + "flow_matching_loss": 0.031143015809357168, + "grad_norm": 0.2729763984680176, + "learning_rate": 2.9021700011567045e-05, + "loss": 0.0659, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015025673806667328, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070119536016136404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03706330135464668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.938346666097641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9590, + "weighted_flow_matching_loss": 0.031143015809357168, + "weighted_rkd_loss": 0.034778218157589434 + }, + { + "epoch": 2.7931335466977014, + "flow_matching_loss": 0.03091367706656456, + "grad_norm": 0.3442994952201843, + "learning_rate": 2.9018761094893166e-05, + "loss": 0.0638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014229299779981375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006621350068598986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035079949721693995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9382206737995148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9600, + "weighted_flow_matching_loss": 0.03091367706656456, + "weighted_rkd_loss": 0.032912701927125455 + }, + { + "epoch": 2.796043060808845, + "flow_matching_loss": 0.03213137947022915, + "grad_norm": 0.38379043340682983, + "learning_rate": 2.9015817919689808e-05, + "loss": 0.0684, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015596373099833727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007468658685684204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038661404699087146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9380945444107056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9610, + "weighted_flow_matching_loss": 0.03213137947022915, + "weighted_rkd_loss": 0.03626802507787943 + }, + { + "epoch": 2.7989525749199884, + "flow_matching_loss": 0.030353519693017007, + "grad_norm": 0.3146935701370239, + "learning_rate": 2.9012870486851034e-05, + "loss": 0.0618, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013596323598176242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063386211637407545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033531267754733564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9379683017730713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9620, + "weighted_flow_matching_loss": 0.030353519693017007, + "weighted_rkd_loss": 0.03145128730684519 + }, + { + "epoch": 2.8018620890311317, + "flow_matching_loss": 0.031077305600047112, + "grad_norm": 0.3391432464122772, + "learning_rate": 2.9009918797272195e-05, + "loss": 0.067, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015591415297240018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007100204937160015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038283036276698115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.937841945886612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9630, + "weighted_flow_matching_loss": 0.031077305600047112, + "weighted_rkd_loss": 0.035903374664485455 + }, + { + "epoch": 2.8047716031422754, + "flow_matching_loss": 0.03120327219367027, + "grad_norm": 0.3191744089126587, + "learning_rate": 2.9006962851849933e-05, + "loss": 0.065, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014664625003933906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0067581093404442075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03608735930174589, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9377154648303986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9640, + "weighted_flow_matching_loss": 0.03120327219367027, + "weighted_rkd_loss": 0.03383971955627203 + }, + { + "epoch": 2.8076811172534186, + "flow_matching_loss": 0.030953659489750863, + "grad_norm": 0.3598615229129791, + "learning_rate": 2.9004002651482192e-05, + "loss": 0.064, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014208583533763886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006787851685658097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035205019265413286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9375888705253601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9650, + "weighted_flow_matching_loss": 0.030953659489750863, + "weighted_rkd_loss": 0.03300777208060026 + }, + { + "epoch": 2.810590631364562, + "flow_matching_loss": 0.029939929209649562, + "grad_norm": 0.37613391876220703, + "learning_rate": 2.9001038197068198e-05, + "loss": 0.0638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014507520664483308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070751247927546505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.036090165935456754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9374621450901032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9660, + "weighted_flow_matching_loss": 0.029939929209649562, + "weighted_rkd_loss": 0.03383314311504364 + }, + { + "epoch": 2.8135001454757056, + "flow_matching_loss": 0.03143588993698358, + "grad_norm": 0.3335900902748108, + "learning_rate": 2.899806948950848e-05, + "loss": 0.0621, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01325863292440772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00618282500654459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03270009122788906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9373353064060211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9670, + "weighted_flow_matching_loss": 0.03143588993698358, + "weighted_rkd_loss": 0.030650906078517438 + }, + { + "epoch": 2.816409659586849, + "flow_matching_loss": 0.03127127755433321, + "grad_norm": 0.29691991209983826, + "learning_rate": 2.8995096529704846e-05, + "loss": 0.0652, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01470796661451459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00681060766801238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0362265408039093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9372083425521851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9680, + "weighted_flow_matching_loss": 0.03127127755433321, + "weighted_rkd_loss": 0.03395186811685562 + }, + { + "epoch": 2.8193191736979926, + "flow_matching_loss": 0.030977802723646163, + "grad_norm": 0.34022462368011475, + "learning_rate": 2.8992119318560403e-05, + "loss": 0.0631, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013873135577887297, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006518348213285207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03426461946219206, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.937081253528595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9690, + "weighted_flow_matching_loss": 0.030977802723646163, + "weighted_rkd_loss": 0.032108734361827374 + }, + { + "epoch": 2.822228687809136, + "flow_matching_loss": 0.02948165833950043, + "grad_norm": 0.31871700286865234, + "learning_rate": 2.8989137856979555e-05, + "loss": 0.0613, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014020580239593983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005927313584834337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03396847397089005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9369540512561798, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9700, + "weighted_flow_matching_loss": 0.02948165833950043, + "weighted_rkd_loss": 0.03182704206556082 + }, + { + "epoch": 2.8251382019202795, + "flow_matching_loss": 0.029100598767399788, + "grad_norm": 0.29269757866859436, + "learning_rate": 2.8986152145867983e-05, + "loss": 0.0607, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013796185608953238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006142721138894558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033735092543065545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9368267357349396, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9710, + "weighted_flow_matching_loss": 0.029100598767399788, + "weighted_rkd_loss": 0.031603970751166346 + }, + { + "epoch": 2.828047716031423, + "flow_matching_loss": 0.03137562647461891, + "grad_norm": 0.29706695675849915, + "learning_rate": 2.8983162186132678e-05, + "loss": 0.0641, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014218100160360337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006511111417785287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034947311878204344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9366993069648742, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9720, + "weighted_flow_matching_loss": 0.03137562647461891, + "weighted_rkd_loss": 0.032735191844403745 + }, + { + "epoch": 2.830957230142566, + "flow_matching_loss": 0.030287200398743153, + "grad_norm": 0.3084027171134949, + "learning_rate": 2.8980167978681905e-05, + "loss": 0.0616, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01368182161822915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006100681098178029, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0334643242880702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9365717470645905, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9730, + "weighted_flow_matching_loss": 0.030287200398743153, + "weighted_rkd_loss": 0.03134180158376694 + }, + { + "epoch": 2.8338667442537098, + "flow_matching_loss": 0.031186359561979772, + "grad_norm": 0.3741031885147095, + "learning_rate": 2.897716952442522e-05, + "loss": 0.0653, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014573494251817466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00725369956344366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03640068843960762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9364440619945527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9740, + "weighted_flow_matching_loss": 0.031186359561979772, + "weighted_rkd_loss": 0.03408723883330822 + }, + { + "epoch": 2.836776258364853, + "flow_matching_loss": 0.031442526169121264, + "grad_norm": 0.3398495316505432, + "learning_rate": 2.8974166824273482e-05, + "loss": 0.0632, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013706333748996258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006509864702820778, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033922532387077806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9363162517547607, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9750, + "weighted_flow_matching_loss": 0.031442526169121264, + "weighted_rkd_loss": 0.03176223021000624 + }, + { + "epoch": 2.8396857724759963, + "flow_matching_loss": 0.029156644456088544, + "grad_norm": 0.36401376128196716, + "learning_rate": 2.897115987913883e-05, + "loss": 0.0621, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01439896272495389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0064196270424872635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035217552073299886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9361883342266083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9760, + "weighted_flow_matching_loss": 0.029156644456088544, + "weighted_rkd_loss": 0.032970307022333147 + }, + { + "epoch": 2.84259528658714, + "flow_matching_loss": 0.0299494668841362, + "grad_norm": 0.3316827118396759, + "learning_rate": 2.896814868993469e-05, + "loss": 0.0635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014449190907180309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0069521617144346235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03585054352879524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9360603094100952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9770, + "weighted_flow_matching_loss": 0.0299494668841362, + "weighted_rkd_loss": 0.03355834633111954 + }, + { + "epoch": 2.8455048006982833, + "flow_matching_loss": 0.03157884180545807, + "grad_norm": 0.3527454733848572, + "learning_rate": 2.896513325757579e-05, + "loss": 0.0663, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015170193649828434, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006739882891997695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03708027042448521, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9359321475028992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9780, + "weighted_flow_matching_loss": 0.03157884180545807, + "weighted_rkd_loss": 0.0347045935690403 + }, + { + "epoch": 2.848414314809427, + "flow_matching_loss": 0.029986545071005823, + "grad_norm": 0.2850632965564728, + "learning_rate": 2.896211358297813e-05, + "loss": 0.0627, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014168714825063943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0065884308889508246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034925860725343226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9358038663864136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9790, + "weighted_flow_matching_loss": 0.029986545071005823, + "weighted_rkd_loss": 0.03268384076654911 + }, + { + "epoch": 2.8513238289205702, + "flow_matching_loss": 0.033227456733584404, + "grad_norm": 0.34513580799102783, + "learning_rate": 2.8959089667059014e-05, + "loss": 0.0672, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014758455473929644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0067651832941919565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03628209438174963, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9356754720211029, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9800, + "weighted_flow_matching_loss": 0.033227456733584404, + "weighted_rkd_loss": 0.033948222361505034 + }, + { + "epoch": 2.854233343031714, + "flow_matching_loss": 0.02857920378446579, + "grad_norm": 0.34926992654800415, + "learning_rate": 2.8956061510737027e-05, + "loss": 0.0607, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013828062266111375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006650243559852243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03430636841803789, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9355469524860383, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9810, + "weighted_flow_matching_loss": 0.02857920378446579, + "weighted_rkd_loss": 0.03209517113864422 + }, + { + "epoch": 2.857142857142857, + "flow_matching_loss": 0.028911647759377955, + "grad_norm": 0.38739874958992004, + "learning_rate": 2.8953029114932042e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014377212524414063, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00711235161870718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03586677685379982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9354183077812195, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9820, + "weighted_flow_matching_loss": 0.028911647759377955, + "weighted_rkd_loss": 0.03355036061257124 + }, + { + "epoch": 2.8600523712540005, + "flow_matching_loss": 0.030493303202092648, + "grad_norm": 0.29548007249832153, + "learning_rate": 2.8949992480565214e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013719443790614605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006531931040808558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03397081885486841, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9352895557880402, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9830, + "weighted_flow_matching_loss": 0.030493303202092648, + "weighted_rkd_loss": 0.03177252635359764 + }, + { + "epoch": 2.862961885365144, + "flow_matching_loss": 0.030047582089900972, + "grad_norm": 0.2738937437534332, + "learning_rate": 2.894695160855901e-05, + "loss": 0.0622, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014127315208315849, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006093625817447901, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03434825576841831, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9351606845855713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9840, + "weighted_flow_matching_loss": 0.030047582089900972, + "weighted_rkd_loss": 0.03212121035903692 + }, + { + "epoch": 2.8658713994762874, + "flow_matching_loss": 0.030355922318994998, + "grad_norm": 0.3166438937187195, + "learning_rate": 2.8943906499837146e-05, + "loss": 0.0626, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013841599877923726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00676256357692182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03444576282054186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9350316941738128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9850, + "weighted_flow_matching_loss": 0.030355922318994998, + "weighted_rkd_loss": 0.03220789153128863 + }, + { + "epoch": 2.8687809135874307, + "flow_matching_loss": 0.029806940257549285, + "grad_norm": 0.27869468927383423, + "learning_rate": 2.8940857155324655e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013951990101486445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006372705474495888, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03427668549120426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9349025845527649, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9860, + "weighted_flow_matching_loss": 0.029806940257549285, + "weighted_rkd_loss": 0.03204540964215994 + }, + { + "epoch": 2.8716904276985744, + "flow_matching_loss": 0.02933635748922825, + "grad_norm": 0.3159496486186981, + "learning_rate": 2.893780357594785e-05, + "loss": 0.062, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014056778978556395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0067934301681816574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03490698821842671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9347733497619629, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9870, + "weighted_flow_matching_loss": 0.02933635748922825, + "weighted_rkd_loss": 0.03263013046234846 + }, + { + "epoch": 2.8745999418097177, + "flow_matching_loss": 0.030125528946518897, + "grad_norm": 0.4302469491958618, + "learning_rate": 2.8934745762634326e-05, + "loss": 0.0657, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015636883955448867, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006795593211427331, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0380693607032299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9346439957618713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9880, + "weighted_flow_matching_loss": 0.030125528946518897, + "weighted_rkd_loss": 0.035581179708242414 + }, + { + "epoch": 2.8775094559208614, + "flow_matching_loss": 0.029635535553097725, + "grad_norm": 0.3903559446334839, + "learning_rate": 2.8931683716312964e-05, + "loss": 0.0646, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.015142095647752285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007145419018343091, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.037429610081017015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9345145225524902, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9890, + "weighted_flow_matching_loss": 0.029635535553097725, + "weighted_rkd_loss": 0.03497854620218277 + }, + { + "epoch": 2.8804189700320046, + "flow_matching_loss": 0.03036311101168394, + "grad_norm": 0.2971019148826599, + "learning_rate": 2.8928617437913932e-05, + "loss": 0.062, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013749805931001901, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006387786474078894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033887398429214954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9343849420547485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9900, + "weighted_flow_matching_loss": 0.03036311101168394, + "weighted_rkd_loss": 0.03166383560746908 + }, + { + "epoch": 2.8833284841431484, + "flow_matching_loss": 0.029653786681592464, + "grad_norm": 0.36988991498947144, + "learning_rate": 2.8925546928368688e-05, + "loss": 0.0628, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01436868291348219, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0067525338381528854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035489899665117265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9342552244663238, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9910, + "weighted_flow_matching_loss": 0.029653786681592464, + "weighted_rkd_loss": 0.03315665666013956 + }, + { + "epoch": 2.8862379982542916, + "flow_matching_loss": 0.031200047582387924, + "grad_norm": 0.3076449930667877, + "learning_rate": 2.8922472188609968e-05, + "loss": 0.0629, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013771192450076342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0063686999026685955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03391108438372612, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.934125405550003, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9920, + "weighted_flow_matching_loss": 0.031200047582387924, + "weighted_rkd_loss": 0.031677281484007835 + }, + { + "epoch": 2.889147512365435, + "flow_matching_loss": 0.03201572224497795, + "grad_norm": 0.2768929600715637, + "learning_rate": 2.8919393219571805e-05, + "loss": 0.0638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013782161939889192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0064212210476398465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03398554492741823, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9339954555034637, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9930, + "weighted_flow_matching_loss": 0.03201572224497795, + "weighted_rkd_loss": 0.03174249324947596 + }, + { + "epoch": 2.8920570264765786, + "flow_matching_loss": 0.029501396976411343, + "grad_norm": 0.321081280708313, + "learning_rate": 2.8916310022189495e-05, + "loss": 0.0622, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014213428739458323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006582892825827002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03500975016504526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9338653922080994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9940, + "weighted_flow_matching_loss": 0.029501396976411343, + "weighted_rkd_loss": 0.03269441742449999 + }, + { + "epoch": 2.894966540587722, + "flow_matching_loss": 0.031375405937433244, + "grad_norm": 0.3455294668674469, + "learning_rate": 2.891322259739964e-05, + "loss": 0.0626, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013618169818073511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006225717859342695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03346205782145262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9337351977825165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9950, + "weighted_flow_matching_loss": 0.031375405937433244, + "weighted_rkd_loss": 0.031244680657982825 + }, + { + "epoch": 2.897876054698865, + "flow_matching_loss": 0.029767658561468124, + "grad_norm": 0.329014390707016, + "learning_rate": 2.891013094614012e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014136698842048646, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006543153198435903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03481655139476061, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.933604896068573, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9960, + "weighted_flow_matching_loss": 0.029767658561468124, + "weighted_rkd_loss": 0.032504852302372454 + }, + { + "epoch": 2.900785568810009, + "flow_matching_loss": 0.02950260378420353, + "grad_norm": 0.3000146448612213, + "learning_rate": 2.8907035069350093e-05, + "loss": 0.0614, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013743948843330144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006659215409308672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0341471130028367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9334744811058044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9970, + "weighted_flow_matching_loss": 0.02950260378420353, + "weighted_rkd_loss": 0.031875472888350485 + }, + { + "epoch": 2.903695082921152, + "flow_matching_loss": 0.02947846744209528, + "grad_norm": 0.3242355287075043, + "learning_rate": 2.8903934967970007e-05, + "loss": 0.0594, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012986192107200622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006093750800937414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03206613473594189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9333439409732819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9980, + "weighted_flow_matching_loss": 0.02947846744209528, + "weighted_rkd_loss": 0.029928750917315484 + }, + { + "epoch": 2.9066045970322953, + "flow_matching_loss": 0.02851237077265978, + "grad_norm": 0.3541213572025299, + "learning_rate": 2.8900830642941583e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014778311643749476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006905201263725758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03646182455122471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9332132697105407, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 9990, + "weighted_flow_matching_loss": 0.02851237077265978, + "weighted_rkd_loss": 0.034026547148823735 + }, + { + "epoch": 2.909514111143439, + "flow_matching_loss": 0.028252163901925087, + "grad_norm": 0.2814348638057709, + "learning_rate": 2.889772209520785e-05, + "loss": 0.06, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01398411113768816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061094183940440415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034077640436589715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9330824911594391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10000, + "weighted_flow_matching_loss": 0.028252163901925087, + "weighted_rkd_loss": 0.0317973293364048 + }, + { + "epoch": 2.9124236252545828, + "flow_matching_loss": 0.02957969345152378, + "grad_norm": 0.3816821873188019, + "learning_rate": 2.8894609325713087e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014097931329160929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00685209077782929, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.035047953575849534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9329515933990479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10010, + "weighted_flow_matching_loss": 0.02957969345152378, + "weighted_rkd_loss": 0.032697971910238266 + }, + { + "epoch": 2.915333139365726, + "flow_matching_loss": 0.027569988183677195, + "grad_norm": 0.2953154444694519, + "learning_rate": 2.889149233540288e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014238129649311305, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006647026957944035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03512328639626503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9328205823898316, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10020, + "weighted_flow_matching_loss": 0.027569988183677195, + "weighted_rkd_loss": 0.032763649523258206 + }, + { + "epoch": 2.9182426534768693, + "flow_matching_loss": 0.03183102551847696, + "grad_norm": 0.31275254487991333, + "learning_rate": 2.8888371125224084e-05, + "loss": 0.0636, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01373939160257578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0066121034789830445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03409088663756847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9326894462108613, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10030, + "weighted_flow_matching_loss": 0.03183102551847696, + "weighted_rkd_loss": 0.03179629277437925 + }, + { + "epoch": 2.921152167588013, + "flow_matching_loss": 0.028377184830605983, + "grad_norm": 0.35939329862594604, + "learning_rate": 2.8885245696124835e-05, + "loss": 0.0616, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01453371588140726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0065702382940799, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03563767019659281, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9325581848621368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10040, + "weighted_flow_matching_loss": 0.028377184830605983, + "weighted_rkd_loss": 0.03323418237268925 + }, + { + "epoch": 2.9240616816991563, + "flow_matching_loss": 0.031529971212148664, + "grad_norm": 0.32841771841049194, + "learning_rate": 2.888211604905457e-05, + "loss": 0.0649, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014488313347101212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006821727426722646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03579835407435894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9324268102645874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10050, + "weighted_flow_matching_loss": 0.031529971212148664, + "weighted_rkd_loss": 0.03337934520095587 + }, + { + "epoch": 2.9269711958102995, + "flow_matching_loss": 0.02868347577750683, + "grad_norm": 0.29558855295181274, + "learning_rate": 2.8878982184963986e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013155072741210461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006296745548024773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03260689098387957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9322953343391418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10060, + "weighted_flow_matching_loss": 0.02868347577750683, + "weighted_rkd_loss": 0.030399347841739654 + }, + { + "epoch": 2.9298807099214432, + "flow_matching_loss": 0.02850923631340265, + "grad_norm": 0.32933560013771057, + "learning_rate": 2.8875844104805066e-05, + "loss": 0.0583, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012964520044624806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005985452746972442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031914493255317214, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9321637213230133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10070, + "weighted_flow_matching_loss": 0.02850923631340265, + "weighted_rkd_loss": 0.029749614000320435 + }, + { + "epoch": 2.9327902240325865, + "flow_matching_loss": 0.03250644281506539, + "grad_norm": 0.32042092084884644, + "learning_rate": 2.887270180953107e-05, + "loss": 0.0635, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013522578030824661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006176735181361437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03322189152240753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9320319950580597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10080, + "weighted_flow_matching_loss": 0.03250644281506539, + "weighted_rkd_loss": 0.03096395917236805 + }, + { + "epoch": 2.9356997381437298, + "flow_matching_loss": 0.0278877604752779, + "grad_norm": 0.3626551926136017, + "learning_rate": 2.886955530009656e-05, + "loss": 0.0597, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013780476618558168, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006549378577619791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034110331907868385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.931900143623352, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10090, + "weighted_flow_matching_loss": 0.0278877604752779, + "weighted_rkd_loss": 0.03178737908601761 + }, + { + "epoch": 2.9386092522548735, + "flow_matching_loss": 0.03106317427009344, + "grad_norm": 0.3051478862762451, + "learning_rate": 2.8866404577457352e-05, + "loss": 0.0646, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014549005217850209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00691971043124795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03601772077381611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9317681789398193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10100, + "weighted_flow_matching_loss": 0.03106317427009344, + "weighted_rkd_loss": 0.03356005512177944 + }, + { + "epoch": 2.9415187663660167, + "flow_matching_loss": 0.028400237672030926, + "grad_norm": 0.2899110019207001, + "learning_rate": 2.8863249642570557e-05, + "loss": 0.0627, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014915261417627335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0069670602679252625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03679758328944445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9316361010074615, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10110, + "weighted_flow_matching_loss": 0.028400237672030926, + "weighted_rkd_loss": 0.034282095916569236 + }, + { + "epoch": 2.9444282804771604, + "flow_matching_loss": 0.03088349476456642, + "grad_norm": 0.3157297670841217, + "learning_rate": 2.8860090496394552e-05, + "loss": 0.0618, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01345409257337451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006268508452922106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03317669350653887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9315038859844208, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10120, + "weighted_flow_matching_loss": 0.03088349476456642, + "weighted_rkd_loss": 0.030904238112270832 + }, + { + "epoch": 2.9473377945883037, + "flow_matching_loss": 0.028973296284675598, + "grad_norm": 0.2984567880630493, + "learning_rate": 2.8856927139889015e-05, + "loss": 0.0618, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014377867896109819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006439428869634866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03519516438245773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9313715696334839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10130, + "weighted_flow_matching_loss": 0.028973296284675598, + "weighted_rkd_loss": 0.03277973774820566 + }, + { + "epoch": 2.9502473086994474, + "flow_matching_loss": 0.02864384315907955, + "grad_norm": 0.3172033727169037, + "learning_rate": 2.8853759574014878e-05, + "loss": 0.0612, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014287353120744228, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006349544413387776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034924250841140744, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9312391340732574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10140, + "weighted_flow_matching_loss": 0.02864384315907955, + "weighted_rkd_loss": 0.03252273909747601 + }, + { + "epoch": 2.9531568228105907, + "flow_matching_loss": 0.029388042539358138, + "grad_norm": 0.3317939043045044, + "learning_rate": 2.8850587799734372e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014039418008178473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00686241677030921, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034941252507269385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9311065793037414, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10150, + "weighted_flow_matching_loss": 0.029388042539358138, + "weighted_rkd_loss": 0.03253397103399038 + }, + { + "epoch": 2.956066336921734, + "flow_matching_loss": 0.03141817655414343, + "grad_norm": 0.3097609877586365, + "learning_rate": 2.8847411818011e-05, + "loss": 0.0638, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014158354699611663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006450563156977296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03476727213710547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.930973905324936, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10160, + "weighted_flow_matching_loss": 0.03141817655414343, + "weighted_rkd_loss": 0.03236747402697802 + }, + { + "epoch": 2.9589758510328776, + "flow_matching_loss": 0.029155940189957617, + "grad_norm": 0.302959144115448, + "learning_rate": 2.8844231629809534e-05, + "loss": 0.0598, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01337084760889411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006214338447898626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0329560337588191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9308411061763764, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10170, + "weighted_flow_matching_loss": 0.029155940189957617, + "weighted_rkd_loss": 0.03067676741629839 + }, + { + "epoch": 2.961885365144021, + "flow_matching_loss": 0.027355018071830273, + "grad_norm": 0.34403640031814575, + "learning_rate": 2.8841047236096037e-05, + "loss": 0.0578, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013245831709355116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006200742069631815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03269240539520979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9307081997394562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10180, + "weighted_flow_matching_loss": 0.027355018071830273, + "weighted_rkd_loss": 0.03042709082365036 + }, + { + "epoch": 2.964794879255164, + "flow_matching_loss": 0.02956055514514446, + "grad_norm": 0.3090493977069855, + "learning_rate": 2.8837858637837838e-05, + "loss": 0.0629, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014511201996356249, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006815976789221167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03583838120102882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.930575168132782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10190, + "weighted_flow_matching_loss": 0.02956055514514446, + "weighted_rkd_loss": 0.03335043918341398 + }, + { + "epoch": 2.967704393366308, + "flow_matching_loss": 0.02783243767917156, + "grad_norm": 0.31382304430007935, + "learning_rate": 2.8834665836003556e-05, + "loss": 0.0613, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014440871682018042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0070560346357524395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03593777772039175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9304420292377472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10200, + "weighted_flow_matching_loss": 0.02783243767917156, + "weighted_rkd_loss": 0.03343800511211157 + }, + { + "epoch": 2.970613907477451, + "flow_matching_loss": 0.0296140693128109, + "grad_norm": 0.34813955426216125, + "learning_rate": 2.8831468831563075e-05, + "loss": 0.0574, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012109944690018892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005675451317802072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029895341023802757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9303087592124939, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10210, + "weighted_flow_matching_loss": 0.0296140693128109, + "weighted_rkd_loss": 0.027811931259930135 + }, + { + "epoch": 2.973523421588595, + "flow_matching_loss": 0.028008808940649034, + "grad_norm": 0.3069765567779541, + "learning_rate": 2.882826762548756e-05, + "loss": 0.0599, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013877651933580638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006570479646325111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0343257836997509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.93017538189888, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10220, + "weighted_flow_matching_loss": 0.028008808940649034, + "weighted_rkd_loss": 0.03192897513508797 + }, + { + "epoch": 2.976432935699738, + "flow_matching_loss": 0.028500430472195148, + "grad_norm": 0.3029159605503082, + "learning_rate": 2.8825062218749456e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013433957193046808, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00604522661305964, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032913141138851645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9300418794155121, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10230, + "weighted_flow_matching_loss": 0.028500430472195148, + "weighted_rkd_loss": 0.030610533244907857 + }, + { + "epoch": 2.979342449810882, + "flow_matching_loss": 0.029327880032360552, + "grad_norm": 0.32428672909736633, + "learning_rate": 2.8821852612322477e-05, + "loss": 0.0609, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013845464214682579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006271486170589924, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033962414041161536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9299082577228546, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10240, + "weighted_flow_matching_loss": 0.029327880032360552, + "weighted_rkd_loss": 0.03158186040818691 + }, + { + "epoch": 2.982251963922025, + "flow_matching_loss": 0.027307946979999543, + "grad_norm": 0.31562501192092896, + "learning_rate": 2.8818638807181622e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012972184829413891, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005857701320201159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031802070885896684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9297745227813721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10250, + "weighted_flow_matching_loss": 0.027307946979999543, + "weighted_rkd_loss": 0.029568685591220854 + }, + { + "epoch": 2.9851614780331683, + "flow_matching_loss": 0.03127262387424708, + "grad_norm": 0.3550442159175873, + "learning_rate": 2.8815420804303154e-05, + "loss": 0.0654, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014832188468426466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00699267634190619, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03665705341845751, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9296406686306, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10260, + "weighted_flow_matching_loss": 0.03127262387424708, + "weighted_rkd_loss": 0.03407796788960695 + }, + { + "epoch": 2.988070992144312, + "flow_matching_loss": 0.029513681307435036, + "grad_norm": 0.26922059059143066, + "learning_rate": 2.8812198604664627e-05, + "loss": 0.0607, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013464488089084625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006635987944900989, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03356496449559927, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9295066893100739, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10270, + "weighted_flow_matching_loss": 0.029513681307435036, + "weighted_rkd_loss": 0.0311988340690732 + }, + { + "epoch": 2.9909805062554553, + "flow_matching_loss": 0.02906907256692648, + "grad_norm": 0.3063719868659973, + "learning_rate": 2.8808972209244848e-05, + "loss": 0.0604, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013769758120179176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006165406526997685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03370492234826088, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9293726086616516, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10280, + "weighted_flow_matching_loss": 0.02906907256692648, + "weighted_rkd_loss": 0.031324368342757224 + }, + { + "epoch": 2.9938900203665986, + "flow_matching_loss": 0.02726986426860094, + "grad_norm": 0.3296976685523987, + "learning_rate": 2.8805741619023922e-05, + "loss": 0.0585, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013602560013532638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006423691241070628, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03362881131470204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9292383909225463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10290, + "weighted_flow_matching_loss": 0.02726986426860094, + "weighted_rkd_loss": 0.03124915212392807 + }, + { + "epoch": 2.9967995344777423, + "flow_matching_loss": 0.027905507199466228, + "grad_norm": 0.3232160806655884, + "learning_rate": 2.880250683498321e-05, + "loss": 0.0581, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01317749936133623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006106317695230246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03246131632477045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.929104071855545, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10300, + "weighted_flow_matching_loss": 0.027905507199466228, + "weighted_rkd_loss": 0.03015987705439329 + }, + { + "epoch": 2.9997090485888855, + "flow_matching_loss": 0.030574833415448664, + "grad_norm": 0.2892548739910126, + "learning_rate": 2.879926785810535e-05, + "loss": 0.0617, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01356283389031887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006406490551307797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03353215828537941, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9289696216583252, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10310, + "weighted_flow_matching_loss": 0.030574833415448664, + "weighted_rkd_loss": 0.0311502980068326 + }, + { + "epoch": 3.0026185627000292, + "flow_matching_loss": 0.030907134525477886, + "grad_norm": 0.3176196217536926, + "learning_rate": 2.8796024689374267e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012939737085253, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005776355136185884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03165582977235317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9288350701332092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10320, + "weighted_flow_matching_loss": 0.030907134525477886, + "weighted_rkd_loss": 0.02940293066203594 + }, + { + "epoch": 3.0055280768111725, + "flow_matching_loss": 0.03069208674132824, + "grad_norm": 0.3208400309085846, + "learning_rate": 2.8792777329775148e-05, + "loss": 0.0619, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013682213425636292, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00628138747997582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033645814284682274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9287003874778748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10330, + "weighted_flow_matching_loss": 0.03069208674132824, + "weighted_rkd_loss": 0.03124694600701332 + }, + { + "epoch": 3.0084375909223158, + "flow_matching_loss": 0.029422335885465146, + "grad_norm": 0.3313102424144745, + "learning_rate": 2.878952578029445e-05, + "loss": 0.0611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013757558260113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006545992754399777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03406110890209675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9285655856132508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10340, + "weighted_flow_matching_loss": 0.029422335885465146, + "weighted_rkd_loss": 0.03162800818681717 + }, + { + "epoch": 3.0113471050334595, + "flow_matching_loss": 0.028691714443266392, + "grad_norm": 0.32703521847724915, + "learning_rate": 2.8786270041919915e-05, + "loss": 0.0599, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013621229864656925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006390033103525638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03363249320536852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9284306764602661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10350, + "weighted_flow_matching_loss": 0.028691714443266392, + "weighted_rkd_loss": 0.03122543077915907 + }, + { + "epoch": 3.0142566191446027, + "flow_matching_loss": 0.02944072801619768, + "grad_norm": 0.2983033061027527, + "learning_rate": 2.8783010115640546e-05, + "loss": 0.0592, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013041570689529181, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006003226246684789, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03208636753261089, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9282956480979919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10360, + "weighted_flow_matching_loss": 0.02944072801619768, + "weighted_rkd_loss": 0.029785613343119622 + }, + { + "epoch": 3.0171661332557465, + "flow_matching_loss": 0.02755788341164589, + "grad_norm": 0.29525479674339294, + "learning_rate": 2.877974600244662e-05, + "loss": 0.0566, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012738234363496304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005799530074000359, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03127599842846394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9281604945659637, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10370, + "weighted_flow_matching_loss": 0.02755788341164589, + "weighted_rkd_loss": 0.029029209911823273 + }, + { + "epoch": 3.0200756473668897, + "flow_matching_loss": 0.028408727236092092, + "grad_norm": 0.36577051877975464, + "learning_rate": 2.8776477703329696e-05, + "loss": 0.0598, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013799057248979807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00618004691787064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03377816155552864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9280252158641815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10380, + "weighted_flow_matching_loss": 0.028408727236092092, + "weighted_rkd_loss": 0.03134689275175333 + }, + { + "epoch": 3.022985161478033, + "flow_matching_loss": 0.02835261020809412, + "grad_norm": 0.296121746301651, + "learning_rate": 2.877320521928259e-05, + "loss": 0.0594, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013557309564203025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006337885791435838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03345250487327576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9278898537158966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10390, + "weighted_flow_matching_loss": 0.02835261020809412, + "weighted_rkd_loss": 0.03104022704064846 + }, + { + "epoch": 3.0258946755891767, + "flow_matching_loss": 0.029998121969401836, + "grad_norm": 0.30091774463653564, + "learning_rate": 2.87699285512994e-05, + "loss": 0.06, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01309389267116785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006117218686267733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03230500407516956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9277543485164642, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10400, + "weighted_flow_matching_loss": 0.029998121969401836, + "weighted_rkd_loss": 0.029971151426434518 + }, + { + "epoch": 3.02880418970032, + "flow_matching_loss": 0.028669328056275846, + "grad_norm": 0.28701314330101013, + "learning_rate": 2.876664770037549e-05, + "loss": 0.0601, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013766464684158564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006353308260440826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033886238001286983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9276187241077423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10410, + "weighted_flow_matching_loss": 0.028669328056275846, + "weighted_rkd_loss": 0.03143356367945671 + }, + { + "epoch": 3.0317137038114637, + "flow_matching_loss": 0.0316011929884553, + "grad_norm": 0.3359370529651642, + "learning_rate": 2.8763362667507495e-05, + "loss": 0.0613, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012896427232772113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006191836483776569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03198469057679176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9274829983711242, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10420, + "weighted_flow_matching_loss": 0.0316011929884553, + "weighted_rkd_loss": 0.029665305651724337 + }, + { + "epoch": 3.034623217922607, + "flow_matching_loss": 0.031032660230994224, + "grad_norm": 0.36516207456588745, + "learning_rate": 2.8760073453693322e-05, + "loss": 0.0623, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013684317097067833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006316066952422261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03368470054119825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9273471236228943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10430, + "weighted_flow_matching_loss": 0.031032660230994224, + "weighted_rkd_loss": 0.031237355992197992 + }, + { + "epoch": 3.03753273203375, + "flow_matching_loss": 0.02801685780286789, + "grad_norm": 0.31976544857025146, + "learning_rate": 2.8756780059932146e-05, + "loss": 0.0583, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013297459296882152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00606021136045456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03265512976795435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9272111773490905, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10440, + "weighted_flow_matching_loss": 0.02801685780286789, + "weighted_rkd_loss": 0.0302782092243433 + }, + { + "epoch": 3.040442246144894, + "flow_matching_loss": 0.029049790278077126, + "grad_norm": 0.36033061146736145, + "learning_rate": 2.875348248722442e-05, + "loss": 0.061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01384362941607833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006822542659938336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03450980111956596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9270750880241394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10450, + "weighted_flow_matching_loss": 0.029049790278077126, + "weighted_rkd_loss": 0.03199308533221483 + }, + { + "epoch": 3.043351760256037, + "flow_matching_loss": 0.031119352951645852, + "grad_norm": 0.32937681674957275, + "learning_rate": 2.8750180736571848e-05, + "loss": 0.0632, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01410374566912651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006350952945649624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03455844484269619, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9269388854503632, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10460, + "weighted_flow_matching_loss": 0.031119352951645852, + "weighted_rkd_loss": 0.03203356470912695 + }, + { + "epoch": 3.046261274367181, + "flow_matching_loss": 0.02834844421595335, + "grad_norm": 0.3140304386615753, + "learning_rate": 2.8746874808977424e-05, + "loss": 0.0605, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01411818079650402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006485482212156057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034721843898296356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9268025696277619, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10470, + "weighted_flow_matching_loss": 0.02834844421595335, + "weighted_rkd_loss": 0.03218026757240296 + }, + { + "epoch": 3.049170788478324, + "flow_matching_loss": 0.028931792825460434, + "grad_norm": 0.3049183785915375, + "learning_rate": 2.87435647054454e-05, + "loss": 0.0587, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013135838508605956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005892625451087952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032164302095770834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9266661107540131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10480, + "weighted_flow_matching_loss": 0.028931792825460434, + "weighted_rkd_loss": 0.029805465415120126 + }, + { + "epoch": 3.0520803025894674, + "flow_matching_loss": 0.029588560201227666, + "grad_norm": 0.3038621246814728, + "learning_rate": 2.8740250426981292e-05, + "loss": 0.0612, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013961190264672041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061814588028937575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034103838913142684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9265295743942261, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10490, + "weighted_flow_matching_loss": 0.029588560201227666, + "weighted_rkd_loss": 0.031598209775984286 + }, + { + "epoch": 3.054989816700611, + "flow_matching_loss": 0.027797790244221687, + "grad_norm": 0.37421417236328125, + "learning_rate": 2.8736931974591895e-05, + "loss": 0.0595, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013854448683559895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00652036271058023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03422925993800163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9263929009437561, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10500, + "weighted_flow_matching_loss": 0.027797790244221687, + "weighted_rkd_loss": 0.03170965164899826 + }, + { + "epoch": 3.0578993308117544, + "flow_matching_loss": 0.02897066306322813, + "grad_norm": 0.3406999409198761, + "learning_rate": 2.8733609349285268e-05, + "loss": 0.0598, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013503830041736364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006248524412512779, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03325618449598551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9262561142444611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10510, + "weighted_flow_matching_loss": 0.02897066306322813, + "weighted_rkd_loss": 0.03080368898808956 + }, + { + "epoch": 3.060808844922898, + "flow_matching_loss": 0.028700397349894047, + "grad_norm": 0.34003952145576477, + "learning_rate": 2.8730282552070734e-05, + "loss": 0.0611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014197856932878495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006600422225892544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03499613627791405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9261192083358765, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10520, + "weighted_flow_matching_loss": 0.028700397349894047, + "weighted_rkd_loss": 0.03241061605513096 + }, + { + "epoch": 3.0637183590340413, + "flow_matching_loss": 0.02912224605679512, + "grad_norm": 0.3484744131565094, + "learning_rate": 2.872695158395889e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013647429458796979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006348025286570191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033642884343862534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9259821772575378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10530, + "weighted_flow_matching_loss": 0.02912224605679512, + "weighted_rkd_loss": 0.031152716279029845 + }, + { + "epoch": 3.0666278731451846, + "flow_matching_loss": 0.029063655994832516, + "grad_norm": 0.33297064900398254, + "learning_rate": 2.872361644596159e-05, + "loss": 0.0583, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012885281257331372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005818867729976774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0315894303843379, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9258450508117676, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10540, + "weighted_flow_matching_loss": 0.029063655994832516, + "weighted_rkd_loss": 0.029246947169303893 + }, + { + "epoch": 3.0695373872563283, + "flow_matching_loss": 0.0276536813005805, + "grad_norm": 0.33991390466690063, + "learning_rate": 2.8720277139091972e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012799920607358217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006023520231246948, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0316233616322279, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9257077932357788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10550, + "weighted_flow_matching_loss": 0.0276536813005805, + "weighted_rkd_loss": 0.02927402816712856 + }, + { + "epoch": 3.0724469013674716, + "flow_matching_loss": 0.029099046997725964, + "grad_norm": 0.32802245020866394, + "learning_rate": 2.8716933664364417e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013144699390977621, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005900626117363572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032190025225281714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9255704164505005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10560, + "weighted_flow_matching_loss": 0.029099046997725964, + "weighted_rkd_loss": 0.02979414090514183 + }, + { + "epoch": 3.0753564154786153, + "flow_matching_loss": 0.02880549319088459, + "grad_norm": 0.3011145293712616, + "learning_rate": 2.871358602279459e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013365133386105298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006019118474796414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032749385572969913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9254329323768615, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10570, + "weighted_flow_matching_loss": 0.02880549319088459, + "weighted_rkd_loss": 0.030307378247380258 + }, + { + "epoch": 3.0782659295897585, + "flow_matching_loss": 0.028179449401795863, + "grad_norm": 0.28139570355415344, + "learning_rate": 2.871023421539942e-05, + "loss": 0.059, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013643726706504822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006016458384692669, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033303911611437795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9252953231334686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10580, + "weighted_flow_matching_loss": 0.028179449401795863, + "weighted_rkd_loss": 0.030815964564681053 + }, + { + "epoch": 3.081175443700902, + "flow_matching_loss": 0.02895564418286085, + "grad_norm": 0.4620151221752167, + "learning_rate": 2.8706878243197085e-05, + "loss": 0.0598, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01357204569503665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00616939072497189, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033313481323421004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9251576066017151, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10590, + "weighted_flow_matching_loss": 0.02895564418286085, + "weighted_rkd_loss": 0.03082014210522175 + }, + { + "epoch": 3.0840849578120455, + "flow_matching_loss": 0.027557658217847347, + "grad_norm": 0.2746884524822235, + "learning_rate": 2.8703518107207052e-05, + "loss": 0.0573, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013088294211775065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006002820935100317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03217940926551819, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9250197589397431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10600, + "weighted_flow_matching_loss": 0.027557658217847347, + "weighted_rkd_loss": 0.029766695946455 + }, + { + "epoch": 3.0869944719231888, + "flow_matching_loss": 0.028186590783298014, + "grad_norm": 0.3324798345565796, + "learning_rate": 2.870015380845004e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01344471275806427, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006356616923585534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03324604257941246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9248818099498749, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10610, + "weighted_flow_matching_loss": 0.028186590783298014, + "weighted_rkd_loss": 0.030748629197478294 + }, + { + "epoch": 3.0899039860343325, + "flow_matching_loss": 0.02889210395514965, + "grad_norm": 0.33932599425315857, + "learning_rate": 2.8696785347948035e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01468050917610526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007022034469991922, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03638305235654116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9247437357902527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10620, + "weighted_flow_matching_loss": 0.02889210395514965, + "weighted_rkd_loss": 0.03364506624639034 + }, + { + "epoch": 3.0928135001454757, + "flow_matching_loss": 0.028623178787529468, + "grad_norm": 0.30215075612068176, + "learning_rate": 2.8693412726724282e-05, + "loss": 0.0608, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01405872730538249, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006629849411547184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03474730364978314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9246055483818054, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10630, + "weighted_flow_matching_loss": 0.028623178787529468, + "weighted_rkd_loss": 0.032127547450363635 + }, + { + "epoch": 3.095723014256619, + "flow_matching_loss": 0.030673382245004176, + "grad_norm": 0.32915082573890686, + "learning_rate": 2.869003594580329e-05, + "loss": 0.0625, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014005404710769654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0064148666802793745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034425676241517066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.924467247724533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10640, + "weighted_flow_matching_loss": 0.030673382245004176, + "weighted_rkd_loss": 0.031825358606874944 + }, + { + "epoch": 3.0986325283677627, + "flow_matching_loss": 0.029474103078246116, + "grad_norm": 0.32338520884513855, + "learning_rate": 2.868665500621085e-05, + "loss": 0.0582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01273848870769143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056471772026270625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031124155037105083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9243288278579712, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10650, + "weighted_flow_matching_loss": 0.029474103078246116, + "weighted_rkd_loss": 0.028768960386514664 + }, + { + "epoch": 3.101542042478906, + "flow_matching_loss": 0.027551540918648244, + "grad_norm": 0.3764817714691162, + "learning_rate": 2.8683269908973992e-05, + "loss": 0.0592, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013913547061383725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006438111094757915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03426520526409149, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9241902828216553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10660, + "weighted_flow_matching_loss": 0.027551540918648244, + "weighted_rkd_loss": 0.031667581759393215 + }, + { + "epoch": 3.1044515565900497, + "flow_matching_loss": 0.029681161791086198, + "grad_norm": 0.28632286190986633, + "learning_rate": 2.867988065512102e-05, + "loss": 0.0611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013750328123569489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006505241896957159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03400589805096388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9240516304969788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10670, + "weighted_flow_matching_loss": 0.029681161791086198, + "weighted_rkd_loss": 0.031423288583755496 + }, + { + "epoch": 3.107361070701193, + "flow_matching_loss": 0.02960311658680439, + "grad_norm": 0.30262187123298645, + "learning_rate": 2.8676487245681507e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012896949518471956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005886293016374111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03168019223958254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9239128530025482, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10680, + "weighted_flow_matching_loss": 0.02960311658680439, + "weighted_rkd_loss": 0.029269797541201115 + }, + { + "epoch": 3.110270584812336, + "flow_matching_loss": 0.027990597113966943, + "grad_norm": 0.2874208390712738, + "learning_rate": 2.867308968168627e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012780906539410353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061069847084581855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03166879806667566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9237739682197571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10690, + "weighted_flow_matching_loss": 0.027990597113966943, + "weighted_rkd_loss": 0.02925490476191044 + }, + { + "epoch": 3.11318009892348, + "flow_matching_loss": 0.029735643602907657, + "grad_norm": 0.35570836067199707, + "learning_rate": 2.8669687964167404e-05, + "loss": 0.0598, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013307060394436122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005922046909108758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03253616783767939, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9236349821090698, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10700, + "weighted_flow_matching_loss": 0.029735643602907657, + "weighted_rkd_loss": 0.03005142752081156 + }, + { + "epoch": 3.116089613034623, + "flow_matching_loss": 0.03064459003508091, + "grad_norm": 0.3352021276950836, + "learning_rate": 2.8666282094158264e-05, + "loss": 0.0658, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01544206840917468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007207584939897061, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.038091722130775454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9234958589076996, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10710, + "weighted_flow_matching_loss": 0.03064459003508091, + "weighted_rkd_loss": 0.03517750725150108 + }, + { + "epoch": 3.1189991271457664, + "flow_matching_loss": 0.028104372136294842, + "grad_norm": 0.2885538935661316, + "learning_rate": 2.866287207269346e-05, + "loss": 0.0562, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012250771559774876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005871922289952636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030373465083539487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9233566224575043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10720, + "weighted_flow_matching_loss": 0.028104372136294842, + "weighted_rkd_loss": 0.028045633062720298 + }, + { + "epoch": 3.12190864125691, + "flow_matching_loss": 0.029037977755069732, + "grad_norm": 0.31379204988479614, + "learning_rate": 2.8659457900808868e-05, + "loss": 0.0586, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012987008970230817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00603494206443429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03200895972549915, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9232172727584839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10730, + "weighted_flow_matching_loss": 0.029037977755069732, + "weighted_rkd_loss": 0.02955122720450163 + }, + { + "epoch": 3.1248181553680534, + "flow_matching_loss": 0.030108313262462615, + "grad_norm": 0.30884549021720886, + "learning_rate": 2.8656039579541628e-05, + "loss": 0.0617, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013822904601693154, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0065263395197689535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03417214825749397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9230778098106385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10740, + "weighted_flow_matching_loss": 0.030108313262462615, + "weighted_rkd_loss": 0.03154363185167312 + }, + { + "epoch": 3.127727669479197, + "flow_matching_loss": 0.028065699525177477, + "grad_norm": 0.38046950101852417, + "learning_rate": 2.865261710993012e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012667441368103027, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00592795847915113, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031262841261923316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9229382276535034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10750, + "weighted_flow_matching_loss": 0.028065699525177477, + "weighted_rkd_loss": 0.02885360810905695 + }, + { + "epoch": 3.1306371835903404, + "flow_matching_loss": 0.029133317805826665, + "grad_norm": 0.36581432819366455, + "learning_rate": 2.864919049301402e-05, + "loss": 0.0568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012105876300483941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005823252629488707, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030035004764795304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9227985203266144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10760, + "weighted_flow_matching_loss": 0.029133317805826665, + "weighted_rkd_loss": 0.027716160006821155 + }, + { + "epoch": 3.133546697701484, + "flow_matching_loss": 0.026368698477745055, + "grad_norm": 0.2851024866104126, + "learning_rate": 2.8645759729834227e-05, + "loss": 0.0554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012874494213610888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005662711011245847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03141169939190149, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9226587116718292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10770, + "weighted_flow_matching_loss": 0.026368698477745055, + "weighted_rkd_loss": 0.028982340171933175 + }, + { + "epoch": 3.1364562118126273, + "flow_matching_loss": 0.02926948368549347, + "grad_norm": 0.3095395863056183, + "learning_rate": 2.864232482143293e-05, + "loss": 0.0597, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013407961651682853, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061988580040633675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033014781400561334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9225187838077545, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10780, + "weighted_flow_matching_loss": 0.02926948368549347, + "weighted_rkd_loss": 0.03045670073479414 + }, + { + "epoch": 3.1393657259237706, + "flow_matching_loss": 0.02828533612191677, + "grad_norm": 0.3066912591457367, + "learning_rate": 2.8638885768853548e-05, + "loss": 0.0564, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012341807503253222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005808532377704978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030492147617042064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9223787426948548, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10790, + "weighted_flow_matching_loss": 0.02828533612191677, + "weighted_rkd_loss": 0.02812530193477869 + }, + { + "epoch": 3.1422752400349143, + "flow_matching_loss": 0.027667996659874915, + "grad_norm": 0.33519086241722107, + "learning_rate": 2.8635442573140793e-05, + "loss": 0.06, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014142581913620233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006733354227617383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03501851800829172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9222385704517364, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10800, + "weighted_flow_matching_loss": 0.027667996659874915, + "weighted_rkd_loss": 0.03229540307074785 + }, + { + "epoch": 3.1451847541460576, + "flow_matching_loss": 0.027784271724522115, + "grad_norm": 0.29035112261772156, + "learning_rate": 2.8631995235340606e-05, + "loss": 0.0571, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013037392403930426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005765476357191801, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03184026144444942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9220982968807221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10810, + "weighted_flow_matching_loss": 0.027784271724522115, + "weighted_rkd_loss": 0.029359891079366207 + }, + { + "epoch": 3.148094268257201, + "flow_matching_loss": 0.029069545120000838, + "grad_norm": 0.31808289885520935, + "learning_rate": 2.8628543756500192e-05, + "loss": 0.0586, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012980912439525127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061107433401048185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03207256849855185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9219578981399537, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10820, + "weighted_flow_matching_loss": 0.029069545120000838, + "weighted_rkd_loss": 0.029569558426737787 + }, + { + "epoch": 3.1510037823683446, + "flow_matching_loss": 0.030481996200978755, + "grad_norm": 0.315701961517334, + "learning_rate": 2.862508813766803e-05, + "loss": 0.0603, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013189784344285726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006006090063601732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03238565884530544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9218173921108246, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10830, + "weighted_flow_matching_loss": 0.030481996200978755, + "weighted_rkd_loss": 0.02985369451344013 + }, + { + "epoch": 3.153913296479488, + "flow_matching_loss": 0.031188753060996534, + "grad_norm": 0.3007446825504303, + "learning_rate": 2.8621628379893837e-05, + "loss": 0.0624, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013757612649351358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006369772320613265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03388499766588211, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.921676754951477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10840, + "weighted_flow_matching_loss": 0.031188753060996534, + "weighted_rkd_loss": 0.031231031753122807 + }, + { + "epoch": 3.1568228105906315, + "flow_matching_loss": 0.028448774106800558, + "grad_norm": 0.31401148438453674, + "learning_rate": 2.861816448422861e-05, + "loss": 0.0582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01309579899534583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006069568917155266, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03226116672158241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9215360283851624, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10850, + "weighted_flow_matching_loss": 0.028448774106800558, + "weighted_rkd_loss": 0.02972985841333866 + }, + { + "epoch": 3.159732324701775, + "flow_matching_loss": 0.025825664587318897, + "grad_norm": 0.2788149416446686, + "learning_rate": 2.8614696451724573e-05, + "loss": 0.0559, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013285446912050247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0060465613845735785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03261745516210794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9213951706886292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10860, + "weighted_flow_matching_loss": 0.025825664587318897, + "weighted_rkd_loss": 0.030053513869643212 + }, + { + "epoch": 3.162641838812918, + "flow_matching_loss": 0.028511478379368783, + "grad_norm": 0.2817082107067108, + "learning_rate": 2.861122428343523e-05, + "loss": 0.0586, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013315826375037431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005993232782930136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03262488581240177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9212541997432708, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10870, + "weighted_flow_matching_loss": 0.028511478379368783, + "weighted_rkd_loss": 0.030055935867130755 + }, + { + "epoch": 3.1655513529240618, + "flow_matching_loss": 0.02769482769072056, + "grad_norm": 0.2723005712032318, + "learning_rate": 2.8607747980415333e-05, + "loss": 0.0583, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013626347668468953, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005999041814357042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033251737244427204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.921113109588623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10880, + "weighted_flow_matching_loss": 0.02769482769072056, + "weighted_rkd_loss": 0.030628593638539314 + }, + { + "epoch": 3.168460867035205, + "flow_matching_loss": 0.028241120092570782, + "grad_norm": 0.36760884523391724, + "learning_rate": 2.860426754372089e-05, + "loss": 0.059, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013604835141450167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006183942034840584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0333936121314764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9209719002246857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10890, + "weighted_flow_matching_loss": 0.028241120092570782, + "weighted_rkd_loss": 0.03075452446937561 + }, + { + "epoch": 3.1713703811463487, + "flow_matching_loss": 0.029738193564116954, + "grad_norm": 0.3111676275730133, + "learning_rate": 2.860078297440917e-05, + "loss": 0.0593, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013039049692451954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005988912843167782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0320670124143362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9208305954933167, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10900, + "weighted_flow_matching_loss": 0.029738193564116954, + "weighted_rkd_loss": 0.029528307914733886 + }, + { + "epoch": 3.174279895257492, + "flow_matching_loss": 0.0264543442055583, + "grad_norm": 0.2911560535430908, + "learning_rate": 2.8597294273538687e-05, + "loss": 0.0549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012532047182321548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005883086891844869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030947180837392806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9206891596317291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10910, + "weighted_flow_matching_loss": 0.0264543442055583, + "weighted_rkd_loss": 0.02849283255636692 + }, + { + "epoch": 3.1771894093686353, + "flow_matching_loss": 0.0285931583493948, + "grad_norm": 0.26268091797828674, + "learning_rate": 2.8593801442169223e-05, + "loss": 0.0561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01218665884807706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00550015764310956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02987347524613142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9205476105213165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10920, + "weighted_flow_matching_loss": 0.0285931583493948, + "weighted_rkd_loss": 0.027499960176646708 + }, + { + "epoch": 3.180098923479779, + "flow_matching_loss": 0.03079189620912075, + "grad_norm": 0.3133189082145691, + "learning_rate": 2.85903044813618e-05, + "loss": 0.061, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013209186680614948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006360257277265191, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03277863021939993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9204059481620789, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10930, + "weighted_flow_matching_loss": 0.03079189620912075, + "weighted_rkd_loss": 0.03016947116702795 + }, + { + "epoch": 3.1830084375909222, + "flow_matching_loss": 0.027407003007829188, + "grad_norm": 0.31667542457580566, + "learning_rate": 2.8586803392178715e-05, + "loss": 0.0593, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013846631441265345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007007891219109297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03470115400850773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9202641785144806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10940, + "weighted_flow_matching_loss": 0.027407003007829188, + "weighted_rkd_loss": 0.03193417880684137 + }, + { + "epoch": 3.185917951702066, + "flow_matching_loss": 0.027882028371095657, + "grad_norm": 0.37811192870140076, + "learning_rate": 2.858329817568349e-05, + "loss": 0.0578, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013105992134660483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006313181575387716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03252516575157642, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9201222777366638, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10950, + "weighted_flow_matching_loss": 0.027882028371095657, + "weighted_rkd_loss": 0.029926971532404423 + }, + { + "epoch": 3.188827465813209, + "flow_matching_loss": 0.027362870424985884, + "grad_norm": 0.33420059084892273, + "learning_rate": 2.8579788832940925e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.014015763998031616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062047038692981, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034236231818795204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9199802577495575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10960, + "weighted_flow_matching_loss": 0.027362870424985884, + "weighted_rkd_loss": 0.031496606022119525 + }, + { + "epoch": 3.1917369799243525, + "flow_matching_loss": 0.02673804759979248, + "grad_norm": 0.2940081059932709, + "learning_rate": 2.8576275365017073e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01336741866543889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006086841877549887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03282167911529541, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9198381304740906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10970, + "weighted_flow_matching_loss": 0.02673804759979248, + "weighted_rkd_loss": 0.030190536566078664 + }, + { + "epoch": 3.194646494035496, + "flow_matching_loss": 0.02921083550900221, + "grad_norm": 0.3845598101615906, + "learning_rate": 2.857275777297922e-05, + "loss": 0.0585, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013063483778387308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057167696300894025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03184373676776886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.919695895910263, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10980, + "weighted_flow_matching_loss": 0.02921083550900221, + "weighted_rkd_loss": 0.02928653545677662 + }, + { + "epoch": 3.1975560081466394, + "flow_matching_loss": 0.029036371037364006, + "grad_norm": 0.34892934560775757, + "learning_rate": 2.8569236057895924e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012429072149097919, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005768215330317616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03062635976821184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.919553542137146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 10990, + "weighted_flow_matching_loss": 0.029036371037364006, + "weighted_rkd_loss": 0.028162558935582636 + }, + { + "epoch": 3.200465522257783, + "flow_matching_loss": 0.026659191772341728, + "grad_norm": 0.3122139573097229, + "learning_rate": 2.8565710220836994e-05, + "loss": 0.0566, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013224540185183287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006107493862509727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0325565742328763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9194110751152038, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11000, + "weighted_flow_matching_loss": 0.026659191772341728, + "weighted_rkd_loss": 0.029932809248566627 + }, + { + "epoch": 3.2033750363689264, + "flow_matching_loss": 0.02782816607505083, + "grad_norm": 0.2680058777332306, + "learning_rate": 2.8562180262873474e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012330553494393826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005525523517280817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03018663004040718, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9192684888839722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11010, + "weighted_flow_matching_loss": 0.02782816607505083, + "weighted_rkd_loss": 0.027749579027295114 + }, + { + "epoch": 3.2062845504800697, + "flow_matching_loss": 0.02799332235008478, + "grad_norm": 0.32515379786491394, + "learning_rate": 2.8558646185077682e-05, + "loss": 0.0567, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012765258271247148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005685030575841665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031215547397732735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9191257953643799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11020, + "weighted_flow_matching_loss": 0.02799332235008478, + "weighted_rkd_loss": 0.028690976463258265 + }, + { + "epoch": 3.2091940645912134, + "flow_matching_loss": 0.026621663570404054, + "grad_norm": 0.26258978247642517, + "learning_rate": 2.8555107988523177e-05, + "loss": 0.055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012548331636935472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005806743307039142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0309034064412117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9189829707145691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11030, + "weighted_flow_matching_loss": 0.026621663570404054, + "weighted_rkd_loss": 0.02839970085769892 + }, + { + "epoch": 3.2121035787023566, + "flow_matching_loss": 0.02776345517486334, + "grad_norm": 0.3128056824207306, + "learning_rate": 2.8551565674284762e-05, + "loss": 0.0571, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012959585059434175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005986994737759233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031906164810061456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9188400387763977, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11040, + "weighted_flow_matching_loss": 0.02776345517486334, + "weighted_rkd_loss": 0.029316695034503938 + }, + { + "epoch": 3.2150130928135003, + "flow_matching_loss": 0.029243888705968855, + "grad_norm": 0.3215581476688385, + "learning_rate": 2.8548019243438503e-05, + "loss": 0.0593, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013254301249980926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0062105465214699505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03271914888173342, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9186969935894013, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11050, + "weighted_flow_matching_loss": 0.029243888705968855, + "weighted_rkd_loss": 0.03005900848656893 + }, + { + "epoch": 3.2179226069246436, + "flow_matching_loss": 0.028890020586550234, + "grad_norm": 0.3223981559276581, + "learning_rate": 2.8544468697061715e-05, + "loss": 0.0588, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01329000536352396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006017512083053589, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03259752299636602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9185538411140441, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11060, + "weighted_flow_matching_loss": 0.028890020586550234, + "weighted_rkd_loss": 0.029942659288644792 + }, + { + "epoch": 3.220832121035787, + "flow_matching_loss": 0.02862502299249172, + "grad_norm": 0.2898956835269928, + "learning_rate": 2.854091403623296e-05, + "loss": 0.0581, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01316502969712019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005816433159634471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03214649297297001, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9184105634689331, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11070, + "weighted_flow_matching_loss": 0.02862502299249172, + "weighted_rkd_loss": 0.029523693583905698 + }, + { + "epoch": 3.2237416351469306, + "flow_matching_loss": 0.028386818431317806, + "grad_norm": 0.3253194987773895, + "learning_rate": 2.8537355262032047e-05, + "loss": 0.0563, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012364673707634211, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056679644621908665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030397311598062516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9182671785354615, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11080, + "weighted_flow_matching_loss": 0.028386818431317806, + "weighted_rkd_loss": 0.027912742272019387 + }, + { + "epoch": 3.226651149258074, + "flow_matching_loss": 0.0272863183170557, + "grad_norm": 0.2691970765590668, + "learning_rate": 2.853379237554004e-05, + "loss": 0.0552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012280703708529472, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005872947722673416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030434355325996875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9181236743927002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11090, + "weighted_flow_matching_loss": 0.0272863183170557, + "weighted_rkd_loss": 0.027942445129156113 + }, + { + "epoch": 3.2295606633692175, + "flow_matching_loss": 0.028909244760870933, + "grad_norm": 0.34240928292274475, + "learning_rate": 2.853022537783924e-05, + "loss": 0.0553, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011741670500487088, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005312994192354381, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028796335123479365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9179800450801849, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11100, + "weighted_flow_matching_loss": 0.028909244760870933, + "weighted_rkd_loss": 0.02643446736037731 + }, + { + "epoch": 3.232470177480361, + "flow_matching_loss": 0.028184304386377333, + "grad_norm": 0.34256258606910706, + "learning_rate": 2.8526654270013223e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01285923197865486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005567289469763636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03128575310111046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9178363263607026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11110, + "weighted_flow_matching_loss": 0.028184304386377333, + "weighted_rkd_loss": 0.02871514968574047 + }, + { + "epoch": 3.235379691591504, + "flow_matching_loss": 0.02734358124434948, + "grad_norm": 0.2824041545391083, + "learning_rate": 2.8523079053146785e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013969642762094735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0066590358968824145, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03459832146763801, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9176924705505372, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11120, + "weighted_flow_matching_loss": 0.02734358124434948, + "weighted_rkd_loss": 0.03175055794417858 + }, + { + "epoch": 3.2382892057026478, + "flow_matching_loss": 0.027742472477257252, + "grad_norm": 0.2630932927131653, + "learning_rate": 2.851949972832599e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011845776345580816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005426462786272168, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029118015617132186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.91754851937294, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11130, + "weighted_flow_matching_loss": 0.027742472477257252, + "weighted_rkd_loss": 0.02671718206256628 + }, + { + "epoch": 3.241198719813791, + "flow_matching_loss": 0.027811986580491065, + "grad_norm": 0.2996232211589813, + "learning_rate": 2.851591629663814e-05, + "loss": 0.0595, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013963684812188149, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006591816665604711, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03451918605715036, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9174044370651245, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11140, + "weighted_flow_matching_loss": 0.027811986580491065, + "weighted_rkd_loss": 0.03166804276406765 + }, + { + "epoch": 3.2441082339249343, + "flow_matching_loss": 0.029179626144468784, + "grad_norm": 0.2887454330921173, + "learning_rate": 2.8512328759171783e-05, + "loss": 0.0589, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013206546660512686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0059432975016534325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03235639072954655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9172602295875549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11150, + "weighted_flow_matching_loss": 0.029179626144468784, + "weighted_rkd_loss": 0.029679288156330585 + }, + { + "epoch": 3.247017748036078, + "flow_matching_loss": 0.028380247205495833, + "grad_norm": 0.29742151498794556, + "learning_rate": 2.8508737117016722e-05, + "loss": 0.057, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012612782139331102, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006008834205567836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031234398111701013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9171159446239472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11160, + "weighted_flow_matching_loss": 0.028380247205495833, + "weighted_rkd_loss": 0.028645585291087627 + }, + { + "epoch": 3.2499272621472213, + "flow_matching_loss": 0.029329307563602924, + "grad_norm": 0.3142494857311249, + "learning_rate": 2.8505141371264005e-05, + "loss": 0.0597, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013430589996278285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006250227615237236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03311140742152929, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9169715225696564, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11170, + "weighted_flow_matching_loss": 0.029329307563602924, + "weighted_rkd_loss": 0.030362148582935334 + }, + { + "epoch": 3.252836776258365, + "flow_matching_loss": 0.031187831051647665, + "grad_norm": 0.3060455322265625, + "learning_rate": 2.8501541523005918e-05, + "loss": 0.0611, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013252208381891251, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006077384483069182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032581801526248455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9168269872665405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11180, + "weighted_flow_matching_loss": 0.031187831051647665, + "weighted_rkd_loss": 0.029871900752186775 + }, + { + "epoch": 3.2557462903695082, + "flow_matching_loss": 0.029469571262598037, + "grad_norm": 0.2997782528400421, + "learning_rate": 2.8497937573336007e-05, + "loss": 0.0584, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012843050714582204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005897093843668699, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03158319517970085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.916682344675064, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11190, + "weighted_flow_matching_loss": 0.029469571262598037, + "weighted_rkd_loss": 0.028951757960021494 + }, + { + "epoch": 3.258655804480652, + "flow_matching_loss": 0.028273982182145118, + "grad_norm": 0.31377455592155457, + "learning_rate": 2.8494329523349053e-05, + "loss": 0.0564, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012542302906513213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005590456491336227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030675062164664268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9165375769138336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11200, + "weighted_flow_matching_loss": 0.028273982182145118, + "weighted_rkd_loss": 0.02811485603451729 + }, + { + "epoch": 3.261565318591795, + "flow_matching_loss": 0.026137762889266013, + "grad_norm": 0.33859798312187195, + "learning_rate": 2.8490717374141088e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01239336170256138, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005715853907167912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030502577498555183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.916392695903778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11210, + "weighted_flow_matching_loss": 0.026137762889266013, + "weighted_rkd_loss": 0.02795219961553812 + }, + { + "epoch": 3.2644748327029385, + "flow_matching_loss": 0.026956018432974814, + "grad_norm": 0.3456546664237976, + "learning_rate": 2.8487101126809384e-05, + "loss": 0.0564, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01305139074102044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006053818971849978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03215660061687231, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9162477135658265, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11220, + "weighted_flow_matching_loss": 0.026956018432974814, + "weighted_rkd_loss": 0.029463394172489644 + }, + { + "epoch": 3.267384346814082, + "flow_matching_loss": 0.028070475533604623, + "grad_norm": 0.3604898154735565, + "learning_rate": 2.8483480782452463e-05, + "loss": 0.056, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012473696004599332, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00558982181828469, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030537213385105132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9161026060581208, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11230, + "weighted_flow_matching_loss": 0.028070475533604623, + "weighted_rkd_loss": 0.027975231222808362 + }, + { + "epoch": 3.2702938609252254, + "flow_matching_loss": 0.02886912152171135, + "grad_norm": 0.26851704716682434, + "learning_rate": 2.8479856342170097e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01198042556643486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052650175523012875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029225868918001652, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9159573912620544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11240, + "weighted_flow_matching_loss": 0.02886912152171135, + "weighted_rkd_loss": 0.026769622601568697 + }, + { + "epoch": 3.2732033750363687, + "flow_matching_loss": 0.026853700168430805, + "grad_norm": 0.2767212390899658, + "learning_rate": 2.8476227807063283e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01265159184113145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006102263741195202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03140544723719359, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9158120632171631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11250, + "weighted_flow_matching_loss": 0.026853700168430805, + "weighted_rkd_loss": 0.028761524893343448 + }, + { + "epoch": 3.2761128891475124, + "flow_matching_loss": 0.02898221369832754, + "grad_norm": 0.3245231509208679, + "learning_rate": 2.8472595178234284e-05, + "loss": 0.0595, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013510174956172705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006343719968572259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03336406983435154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9156666159629822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11260, + "weighted_flow_matching_loss": 0.02898221369832754, + "weighted_rkd_loss": 0.03055036347359419 + }, + { + "epoch": 3.2790224032586557, + "flow_matching_loss": 0.029112637601792814, + "grad_norm": 0.3056650757789612, + "learning_rate": 2.846895845678659e-05, + "loss": 0.0582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012898645736277104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0059208459686487915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03171813767403364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9155210494995117, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11270, + "weighted_flow_matching_loss": 0.029112637601792814, + "weighted_rkd_loss": 0.029038533754646777 + }, + { + "epoch": 3.2819319173697994, + "flow_matching_loss": 0.0299185823649168, + "grad_norm": 0.34918567538261414, + "learning_rate": 2.8465317643824945e-05, + "loss": 0.06, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013230852782726288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006411016779020428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03287272211164236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9153753757476807, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11280, + "weighted_flow_matching_loss": 0.0299185823649168, + "weighted_rkd_loss": 0.030090846680104734 + }, + { + "epoch": 3.2848414314809427, + "flow_matching_loss": 0.026890826970338823, + "grad_norm": 0.27440589666366577, + "learning_rate": 2.8461672740455334e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011902055144309998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005413909209892154, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02921801954507828, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9152295827865601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11290, + "weighted_flow_matching_loss": 0.026890826970338823, + "weighted_rkd_loss": 0.026741199381649493 + }, + { + "epoch": 3.2877509455920864, + "flow_matching_loss": 0.031174234673380853, + "grad_norm": 0.28008559346199036, + "learning_rate": 2.8458023747784976e-05, + "loss": 0.0622, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013522648997604848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006832657335326076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03387795574963093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9150836765766144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11300, + "weighted_flow_matching_loss": 0.031174234673380853, + "weighted_rkd_loss": 0.031001192145049573 + }, + { + "epoch": 3.2906604597032296, + "flow_matching_loss": 0.02833523936569691, + "grad_norm": 0.2793373763561249, + "learning_rate": 2.8454370666922343e-05, + "loss": 0.0563, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012269904837012291, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006006247503682971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03054605722427368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9149376690387726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11310, + "weighted_flow_matching_loss": 0.02833523936569691, + "weighted_rkd_loss": 0.027947669103741645 + }, + { + "epoch": 3.293569973814373, + "flow_matching_loss": 0.030545560270547868, + "grad_norm": 0.3240315616130829, + "learning_rate": 2.8450713498977145e-05, + "loss": 0.0629, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01420006500557065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.007016333751380444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03541646357625723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9147915363311767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11320, + "weighted_flow_matching_loss": 0.030545560270547868, + "weighted_rkd_loss": 0.032398566417396066 + }, + { + "epoch": 3.2964794879255166, + "flow_matching_loss": 0.02790763471275568, + "grad_norm": 0.3202802836894989, + "learning_rate": 2.844705224506033e-05, + "loss": 0.0577, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013249489013105631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006020824564620853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03251980226486921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9146452963352203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11330, + "weighted_flow_matching_loss": 0.02790763471275568, + "weighted_rkd_loss": 0.029744167067110538 + }, + { + "epoch": 3.29938900203666, + "flow_matching_loss": 0.028185935318470003, + "grad_norm": 0.27170810103416443, + "learning_rate": 2.844338690628409e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012829544581472873, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006114604044705629, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03177369367331266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9144989311695099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11340, + "weighted_flow_matching_loss": 0.028185935318470003, + "weighted_rkd_loss": 0.029056998901069164 + }, + { + "epoch": 3.302298516147803, + "flow_matching_loss": 0.028166859224438668, + "grad_norm": 0.28395015001296997, + "learning_rate": 2.8439717483761863e-05, + "loss": 0.0582, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013348807767033576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061411958187818525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03283881191164255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9143524587154388, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11350, + "weighted_flow_matching_loss": 0.028166859224438668, + "weighted_rkd_loss": 0.030026447400450705 + }, + { + "epoch": 3.305208030258947, + "flow_matching_loss": 0.028645340539515018, + "grad_norm": 0.30673396587371826, + "learning_rate": 2.843604397860832e-05, + "loss": 0.0579, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01285367077216506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006294379429891706, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.032001720555126664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9142058908939361, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11360, + "weighted_flow_matching_loss": 0.028645340539515018, + "weighted_rkd_loss": 0.029256148263812065 + }, + { + "epoch": 3.30811754437009, + "flow_matching_loss": 0.027141240984201433, + "grad_norm": 0.3272933065891266, + "learning_rate": 2.8432366391939373e-05, + "loss": 0.0543, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012001734506338835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005735741881653667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029739210940897466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9140591859817505, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11370, + "weighted_flow_matching_loss": 0.027141240984201433, + "weighted_rkd_loss": 0.027183376252651215 + }, + { + "epoch": 3.311027058481234, + "flow_matching_loss": 0.02548214979469776, + "grad_norm": 0.28963038325309753, + "learning_rate": 2.8428684724872176e-05, + "loss": 0.0524, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01204322287812829, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005403682496398688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029490128345787524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9139123737812043, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11380, + "weighted_flow_matching_loss": 0.02548214979469776, + "weighted_rkd_loss": 0.026951442658901214 + }, + { + "epoch": 3.313936572592377, + "flow_matching_loss": 0.025088816322386263, + "grad_norm": 0.31052646040916443, + "learning_rate": 2.8424998978525123e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01262721512466669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005789409345015884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031043839640915394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9137654423713684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11390, + "weighted_flow_matching_loss": 0.025088816322386263, + "weighted_rkd_loss": 0.028366785123944284 + }, + { + "epoch": 3.3168460867035203, + "flow_matching_loss": 0.02753678560256958, + "grad_norm": 0.30102628469467163, + "learning_rate": 2.842130915401785e-05, + "loss": 0.0564, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012803856004029512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00596246593631804, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03157017733901739, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9136184096336365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11400, + "weighted_flow_matching_loss": 0.02753678560256958, + "weighted_rkd_loss": 0.02884309273213148 + }, + { + "epoch": 3.319755600814664, + "flow_matching_loss": 0.02919732090085745, + "grad_norm": 0.3264353275299072, + "learning_rate": 2.8417615252471226e-05, + "loss": 0.0594, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013352459575980902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006307415571063757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03301233444362879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9134712517261505, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11410, + "weighted_flow_matching_loss": 0.02919732090085745, + "weighted_rkd_loss": 0.030155800469219684 + }, + { + "epoch": 3.3226651149258073, + "flow_matching_loss": 0.02705060914158821, + "grad_norm": 0.28665030002593994, + "learning_rate": 2.841391727500735e-05, + "loss": 0.0559, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012885579653084279, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005808758828788996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03157991822808981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.913323986530304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11420, + "weighted_flow_matching_loss": 0.02705060914158821, + "weighted_rkd_loss": 0.028842703253030778 + }, + { + "epoch": 3.325574629036951, + "flow_matching_loss": 0.028563891537487507, + "grad_norm": 0.2986084222793579, + "learning_rate": 2.8410215222749593e-05, + "loss": 0.0575, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012901555839926004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0058721620589494705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031675273552536964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9131766080856323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11430, + "weighted_flow_matching_loss": 0.028563891537487507, + "weighted_rkd_loss": 0.028925146721303464 + }, + { + "epoch": 3.3284841431480943, + "flow_matching_loss": 0.027770331129431725, + "grad_norm": 0.2738211750984192, + "learning_rate": 2.840650909682252e-05, + "loss": 0.0557, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012449880223721265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005734881293028593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030634641647338867, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9130291223526001, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11440, + "weighted_flow_matching_loss": 0.027770331129431725, + "weighted_rkd_loss": 0.027970292419195176 + }, + { + "epoch": 3.3313936572592375, + "flow_matching_loss": 0.028503617085516453, + "grad_norm": 0.25345733761787415, + "learning_rate": 2.8402798898351963e-05, + "loss": 0.0559, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012321968283504247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005391040677204728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03003497701138258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9128815114498139, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11450, + "weighted_flow_matching_loss": 0.028503617085516453, + "weighted_rkd_loss": 0.02741843033581972 + }, + { + "epoch": 3.3343031713703812, + "flow_matching_loss": 0.028090248070657253, + "grad_norm": 0.33218058943748474, + "learning_rate": 2.8399084628464986e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011367672123014927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00507208991330117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02780743408948183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.912733793258667, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11460, + "weighted_flow_matching_loss": 0.028090248070657253, + "weighted_rkd_loss": 0.025380735099315644 + }, + { + "epoch": 3.3372126854815245, + "flow_matching_loss": 0.02734260056167841, + "grad_norm": 0.2606133222579956, + "learning_rate": 2.8395366288289876e-05, + "loss": 0.055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01226407177746296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005727306846529246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030255449935793877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9125859618186951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11470, + "weighted_flow_matching_loss": 0.02734260056167841, + "weighted_rkd_loss": 0.02761071901768446 + }, + { + "epoch": 3.340122199592668, + "flow_matching_loss": 0.027665658481419088, + "grad_norm": 0.2833198308944702, + "learning_rate": 2.839164387895617e-05, + "loss": 0.0558, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012648060452193022, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005579680157825351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03087580092251301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.912438017129898, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11480, + "weighted_flow_matching_loss": 0.027665658481419088, + "weighted_rkd_loss": 0.028172264434397222 + }, + { + "epoch": 3.3430317137038115, + "flow_matching_loss": 0.02935933191329241, + "grad_norm": 0.2880445420742035, + "learning_rate": 2.8387917401594644e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012741060741245747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005874093202874064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03135621454566717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9122899651527405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11490, + "weighted_flow_matching_loss": 0.02935933191329241, + "weighted_rkd_loss": 0.02860595230013132 + }, + { + "epoch": 3.3459412278149547, + "flow_matching_loss": 0.026871787011623384, + "grad_norm": 0.32124418020248413, + "learning_rate": 2.8384186857337296e-05, + "loss": 0.0538, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011854746285825968, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00576476133428514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02947425376623869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9121417939662934, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11500, + "weighted_flow_matching_loss": 0.026871787011623384, + "weighted_rkd_loss": 0.026884715259075164 + }, + { + "epoch": 3.3488507419260984, + "flow_matching_loss": 0.027888767793774603, + "grad_norm": 0.32709893584251404, + "learning_rate": 2.838045224731737e-05, + "loss": 0.058, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013472494389861821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0061141389422118666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.033059127442538735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9119935095310211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11510, + "weighted_flow_matching_loss": 0.027888767793774603, + "weighted_rkd_loss": 0.030149726942181588 + }, + { + "epoch": 3.3517602560372417, + "flow_matching_loss": 0.029383692890405655, + "grad_norm": 0.32094675302505493, + "learning_rate": 2.8376713572669348e-05, + "loss": 0.0576, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012606707215309144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005702944984659552, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030916359461843967, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9118451118469239, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11520, + "weighted_flow_matching_loss": 0.029383692890405655, + "weighted_rkd_loss": 0.028190921805799008 + }, + { + "epoch": 3.3546697701483854, + "flow_matching_loss": 0.029230945371091367, + "grad_norm": 0.3083152174949646, + "learning_rate": 2.8372970834528924e-05, + "loss": 0.0548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011380910966545343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005319571541622281, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028081393614411355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9116965949535369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11530, + "weighted_flow_matching_loss": 0.029230945371091367, + "weighted_rkd_loss": 0.025601719319820405 + }, + { + "epoch": 3.3575792842595287, + "flow_matching_loss": 0.027969840541481973, + "grad_norm": 0.29782313108444214, + "learning_rate": 2.836922403403306e-05, + "loss": 0.0591, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01376359649002552, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006663421774283052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.034190614521503446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9115479826927185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11540, + "weighted_flow_matching_loss": 0.027969840541481973, + "weighted_rkd_loss": 0.031166432052850725 + }, + { + "epoch": 3.360488798370672, + "flow_matching_loss": 0.029689394868910313, + "grad_norm": 0.3230609595775604, + "learning_rate": 2.8365473172319926e-05, + "loss": 0.0596, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013448258955031633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005942286690697074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0328388050198555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9113992393016815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11550, + "weighted_flow_matching_loss": 0.029689394868910313, + "weighted_rkd_loss": 0.029929180443286896 + }, + { + "epoch": 3.3633983124818156, + "flow_matching_loss": 0.027624006383121014, + "grad_norm": 0.2878817915916443, + "learning_rate": 2.8361718250528936e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012958151660859585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00625086766667664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03216717075556517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9112503886222839, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11560, + "weighted_flow_matching_loss": 0.027624006383121014, + "weighted_rkd_loss": 0.02931229826062918 + }, + { + "epoch": 3.366307826592959, + "flow_matching_loss": 0.02634000573307276, + "grad_norm": 0.3285725712776184, + "learning_rate": 2.835795926980074e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012116946931928396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00573364244773984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029967536218464375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9111014425754547, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11570, + "weighted_flow_matching_loss": 0.02634000573307276, + "weighted_rkd_loss": 0.02730337232351303 + }, + { + "epoch": 3.369217340704102, + "flow_matching_loss": 0.02685701008886099, + "grad_norm": 0.3066326677799225, + "learning_rate": 2.8354196231277213e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012392865400761366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0056080793030560015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03039381019771099, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.910952353477478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11580, + "weighted_flow_matching_loss": 0.02685701008886099, + "weighted_rkd_loss": 0.027687313221395015 + }, + { + "epoch": 3.372126854815246, + "flow_matching_loss": 0.028286951407790185, + "grad_norm": 0.3457241654396057, + "learning_rate": 2.835042913610147e-05, + "loss": 0.0566, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012611423525959254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005848515126854181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031071362271904945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9108031749725342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11590, + "weighted_flow_matching_loss": 0.028286951407790185, + "weighted_rkd_loss": 0.02829985525459051 + }, + { + "epoch": 3.375036368926389, + "flow_matching_loss": 0.02822699025273323, + "grad_norm": 0.2681426405906677, + "learning_rate": 2.8346657985417846e-05, + "loss": 0.0532, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011077364534139633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00521896299906075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027373692020773888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9106538772583008, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11600, + "weighted_flow_matching_loss": 0.02822699025273323, + "weighted_rkd_loss": 0.024927893653512 + }, + { + "epoch": 3.377945883037533, + "flow_matching_loss": 0.027468453533947466, + "grad_norm": 0.3346003293991089, + "learning_rate": 2.8342882780371932e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01147354645654559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0054964786395430565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02844357118010521, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9105044543743134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11610, + "weighted_flow_matching_loss": 0.027468453533947466, + "weighted_rkd_loss": 0.025897985510528087 + }, + { + "epoch": 3.380855397148676, + "flow_matching_loss": 0.027579985931515692, + "grad_norm": 0.2772121727466583, + "learning_rate": 2.833910352211053e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012039576563984156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005348095204681158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02942724823951721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9103549420833588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11620, + "weighted_flow_matching_loss": 0.027579985931515692, + "weighted_rkd_loss": 0.026789219304919242 + }, + { + "epoch": 3.38376491125982, + "flow_matching_loss": 0.02706612627953291, + "grad_norm": 0.229606032371521, + "learning_rate": 2.8335320211781667e-05, + "loss": 0.055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012494591809809209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057536831591278315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030742867104709148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9102053046226501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11630, + "weighted_flow_matching_loss": 0.02706612627953291, + "weighted_rkd_loss": 0.027982284873723985 + }, + { + "epoch": 3.386674425370963, + "flow_matching_loss": 0.029618673026561737, + "grad_norm": 0.24802923202514648, + "learning_rate": 2.8331532850534628e-05, + "loss": 0.0571, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012349244207143784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0055513945408165455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030249882861971854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.910055547952652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11640, + "weighted_flow_matching_loss": 0.029618673026561737, + "weighted_rkd_loss": 0.027529139630496503 + }, + { + "epoch": 3.3895839394821063, + "flow_matching_loss": 0.027282961271703245, + "grad_norm": 0.30708733201026917, + "learning_rate": 2.8327741439519906e-05, + "loss": 0.0549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01233081528916955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005704780062660575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03036641050130129, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9099056959152222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11650, + "weighted_flow_matching_loss": 0.027282961271703245, + "weighted_rkd_loss": 0.027630552649497986 + }, + { + "epoch": 3.39249345359325, + "flow_matching_loss": 0.029247322492301463, + "grad_norm": 0.3363018035888672, + "learning_rate": 2.8323945979889233e-05, + "loss": 0.0558, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01181091396138072, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005521829333156347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029143657349050046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9097557127475738, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11660, + "weighted_flow_matching_loss": 0.029247322492301463, + "weighted_rkd_loss": 0.026513595879077912 + }, + { + "epoch": 3.3954029677043933, + "flow_matching_loss": 0.027842298708856107, + "grad_norm": 0.2885417640209198, + "learning_rate": 2.8320146472795572e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012418476864695549, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005731739709153772, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030568693391978742, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9096056222915649, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11670, + "weighted_flow_matching_loss": 0.027842298708856107, + "weighted_rkd_loss": 0.027805553562939168 + }, + { + "epoch": 3.3983124818155366, + "flow_matching_loss": 0.028888147696852683, + "grad_norm": 0.3203212022781372, + "learning_rate": 2.831634291939311e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012628161814063787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005593598494306207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03084992226213217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9094554245471954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11680, + "weighted_flow_matching_loss": 0.028888147696852683, + "weighted_rkd_loss": 0.02805651556700468 + }, + { + "epoch": 3.4012219959266803, + "flow_matching_loss": 0.028462552465498447, + "grad_norm": 0.31435999274253845, + "learning_rate": 2.8312535320837266e-05, + "loss": 0.0552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01208263672888279, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005229912139475346, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02939518541097641, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9093051195144654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11690, + "weighted_flow_matching_loss": 0.028462552465498447, + "weighted_rkd_loss": 0.02672914881259203 + }, + { + "epoch": 3.4041315100378235, + "flow_matching_loss": 0.029696997813880445, + "grad_norm": 0.2642076015472412, + "learning_rate": 2.8308723678284687e-05, + "loss": 0.056, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011773050390183925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005413708975538611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028959809616208077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9091547071933747, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11700, + "weighted_flow_matching_loss": 0.029696997813880445, + "weighted_rkd_loss": 0.02632891610264778 + }, + { + "epoch": 3.4070410241489673, + "flow_matching_loss": 0.02748587504029274, + "grad_norm": 0.3314532935619354, + "learning_rate": 2.8304907992893254e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011617258563637734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00531214689835906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028546663746237756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9090041637420654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11710, + "weighted_flow_matching_loss": 0.02748587504029274, + "weighted_rkd_loss": 0.025948944874107836 + }, + { + "epoch": 3.4099505382601105, + "flow_matching_loss": 0.028279920108616352, + "grad_norm": 0.24422338604927063, + "learning_rate": 2.8301088265822066e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011698409914970398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00498556848615408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028382388316094874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9088535189628602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11720, + "weighted_flow_matching_loss": 0.028279920108616352, + "weighted_rkd_loss": 0.025795387290418147 + }, + { + "epoch": 3.4128600523712542, + "flow_matching_loss": 0.027913369052112103, + "grad_norm": 0.2870659828186035, + "learning_rate": 2.829726449823146e-05, + "loss": 0.0552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012115277256816626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005745130917057395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02997568529099226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9087027609348297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11730, + "weighted_flow_matching_loss": 0.027913369052112103, + "weighted_rkd_loss": 0.02723897099494934 + }, + { + "epoch": 3.4157695664823975, + "flow_matching_loss": 0.02959590535610914, + "grad_norm": 0.27173754572868347, + "learning_rate": 2.8293436691282993e-05, + "loss": 0.0567, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012252166774123907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005305410269647837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02980974391102791, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9085518836975097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11740, + "weighted_flow_matching_loss": 0.02959590535610914, + "weighted_rkd_loss": 0.027083767391741276 + }, + { + "epoch": 3.4186790805935408, + "flow_matching_loss": 0.02826238926500082, + "grad_norm": 0.3128926455974579, + "learning_rate": 2.828960484613945e-05, + "loss": 0.0571, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013042387925088406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005677525047212839, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03176230099052191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9084009051322937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11750, + "weighted_flow_matching_loss": 0.02826238926500082, + "weighted_rkd_loss": 0.028852966986596583 + }, + { + "epoch": 3.4215885947046845, + "flow_matching_loss": 0.027450861781835555, + "grad_norm": 0.3265913426876068, + "learning_rate": 2.828576896396485e-05, + "loss": 0.0561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01292358972132206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005667940992861986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03151512052863836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.908249807357788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11760, + "weighted_flow_matching_loss": 0.027450861781835555, + "weighted_rkd_loss": 0.02862368803471327 + }, + { + "epoch": 3.4244981088158277, + "flow_matching_loss": 0.02746716570109129, + "grad_norm": 0.30549681186676025, + "learning_rate": 2.828192904592443e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012221731338649988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005247998377308249, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029691461473703384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9080986022949219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11770, + "weighted_flow_matching_loss": 0.02746716570109129, + "weighted_rkd_loss": 0.026962766796350478 + }, + { + "epoch": 3.427407622926971, + "flow_matching_loss": 0.0272561552003026, + "grad_norm": 0.282575398683548, + "learning_rate": 2.8278085093184648e-05, + "loss": 0.0543, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012138285022228957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005560130812227726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02983670085668564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9079472899436951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11780, + "weighted_flow_matching_loss": 0.0272561552003026, + "weighted_rkd_loss": 0.027090140618383884 + }, + { + "epoch": 3.4303171370381147, + "flow_matching_loss": 0.027481824159622192, + "grad_norm": 0.2920219600200653, + "learning_rate": 2.827423710691321e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011422349698841571, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051207543350756165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027965453639626502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9077958464622498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11790, + "weighted_flow_matching_loss": 0.027481824159622192, + "weighted_rkd_loss": 0.02538683880120516 + }, + { + "epoch": 3.433226651149258, + "flow_matching_loss": 0.03130530994385481, + "grad_norm": 0.262966513633728, + "learning_rate": 2.827038508827902e-05, + "loss": 0.0583, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012065442558377982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005584208527579904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029715093597769738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9076443016529083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11800, + "weighted_flow_matching_loss": 0.03130530994385481, + "weighted_rkd_loss": 0.026970851235091686 + }, + { + "epoch": 3.4361361652604017, + "flow_matching_loss": 0.026135264337062834, + "grad_norm": 0.27414897084236145, + "learning_rate": 2.8266529038452223e-05, + "loss": 0.0524, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011839739698916673, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053012459073215725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028980725444853306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9074926614761353, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11810, + "weighted_flow_matching_loss": 0.026135264337062834, + "weighted_rkd_loss": 0.02629983238875866 + }, + { + "epoch": 3.439045679371545, + "flow_matching_loss": 0.026577433198690416, + "grad_norm": 0.2890843451023102, + "learning_rate": 2.826266895860419e-05, + "loss": 0.055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012716829404234886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005845804838463664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03127946350723505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9073408782482147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11820, + "weighted_flow_matching_loss": 0.026577433198690416, + "weighted_rkd_loss": 0.02838115766644478 + }, + { + "epoch": 3.4419551934826886, + "flow_matching_loss": 0.02674483973532915, + "grad_norm": 0.28821077942848206, + "learning_rate": 2.8258804849907508e-05, + "loss": 0.0533, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011987230274826288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005278922943398356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02925338353961706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9071890115737915, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11830, + "weighted_flow_matching_loss": 0.02674483973532915, + "weighted_rkd_loss": 0.02653832621872425 + }, + { + "epoch": 3.444864707593832, + "flow_matching_loss": 0.028248434513807298, + "grad_norm": 0.31266850233078003, + "learning_rate": 2.825493671353599e-05, + "loss": 0.055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012014051992446184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005475604347884655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029503708332777025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9070370256900787, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11840, + "weighted_flow_matching_loss": 0.028248434513807298, + "weighted_rkd_loss": 0.026760891452431677 + }, + { + "epoch": 3.447774221704975, + "flow_matching_loss": 0.02804385218769312, + "grad_norm": 0.32976168394088745, + "learning_rate": 2.8251064550664673e-05, + "loss": 0.0543, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01167022455483675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005613042833283543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02895349208265543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9068849205970764, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11850, + "weighted_flow_matching_loss": 0.02804385218769312, + "weighted_rkd_loss": 0.026257489807903767 + }, + { + "epoch": 3.450683735816119, + "flow_matching_loss": 0.029081770963966847, + "grad_norm": 0.3259187638759613, + "learning_rate": 2.8247188362469823e-05, + "loss": 0.0558, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011962869856506585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005592493899166584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029518233425915242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9067327082157135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11860, + "weighted_flow_matching_loss": 0.029081770963966847, + "weighted_rkd_loss": 0.02676520086824894 + }, + { + "epoch": 3.453593249927262, + "flow_matching_loss": 0.02601269632577896, + "grad_norm": 0.3435107171535492, + "learning_rate": 2.824330815012892e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011969906371086836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005423854780383408, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02936366740614176, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9065803766250611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11870, + "weighted_flow_matching_loss": 0.02601269632577896, + "weighted_rkd_loss": 0.02662042621523142 + }, + { + "epoch": 3.4565027640384054, + "flow_matching_loss": 0.029635803028941154, + "grad_norm": 0.28652992844581604, + "learning_rate": 2.8239423914820668e-05, + "loss": 0.0581, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012796663213521242, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005815786682069302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031409112922847274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9064279496669769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11880, + "weighted_flow_matching_loss": 0.029635803028941154, + "weighted_rkd_loss": 0.028470177948474885 + }, + { + "epoch": 3.459412278149549, + "flow_matching_loss": 0.027282948419451714, + "grad_norm": 0.33926841616630554, + "learning_rate": 2.8235535657724997e-05, + "loss": 0.0555, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012685587909072638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005734169669449329, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031105345487594603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9062753915786743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11890, + "weighted_flow_matching_loss": 0.027282948419451714, + "weighted_rkd_loss": 0.028190076164901257 + }, + { + "epoch": 3.4623217922606924, + "flow_matching_loss": 0.028568912856280805, + "grad_norm": 0.2878091633319855, + "learning_rate": 2.823164338002306e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012110959552228451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005556973582133651, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029778892733156682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9061227381229401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11900, + "weighted_flow_matching_loss": 0.028568912856280805, + "weighted_rkd_loss": 0.02698336262255907 + }, + { + "epoch": 3.465231306371836, + "flow_matching_loss": 0.025991646200418474, + "grad_norm": 0.3081654906272888, + "learning_rate": 2.8227747082897223e-05, + "loss": 0.0537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012379238288849593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005830437364056707, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030588913708925247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9059699714183808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11910, + "weighted_flow_matching_loss": 0.025991646200418474, + "weighted_rkd_loss": 0.02771264910697937 + }, + { + "epoch": 3.4681408204829793, + "flow_matching_loss": 0.028182985447347163, + "grad_norm": 0.3138817846775055, + "learning_rate": 2.8223846767531084e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012314470205456018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005637809820473194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030266750231385232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9058170795440674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11920, + "weighted_flow_matching_loss": 0.028182985447347163, + "weighted_rkd_loss": 0.027416085265576838 + }, + { + "epoch": 3.4710503345941226, + "flow_matching_loss": 0.025911031849682332, + "grad_norm": 0.2699715793132782, + "learning_rate": 2.821994243510945e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01146426573395729, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005360886943526566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02828941810876131, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9056640982627868, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11930, + "weighted_flow_matching_loss": 0.025911031849682332, + "weighted_rkd_loss": 0.025620619393885137 + }, + { + "epoch": 3.4739598487052663, + "flow_matching_loss": 0.028552266024053096, + "grad_norm": 0.26483380794525146, + "learning_rate": 2.821603408681835e-05, + "loss": 0.0567, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012685835827142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005668101925402879, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03103977404534817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9055109918117523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11940, + "weighted_flow_matching_loss": 0.028552266024053096, + "weighted_rkd_loss": 0.028106928430497648 + }, + { + "epoch": 3.4768693628164096, + "flow_matching_loss": 0.03181912768632174, + "grad_norm": 0.254560261964798, + "learning_rate": 2.8212121723845054e-05, + "loss": 0.0563, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011152411438524723, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004790264251641929, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027095087431371213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9053577780723572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11950, + "weighted_flow_matching_loss": 0.03181912768632174, + "weighted_rkd_loss": 0.024530839174985886 + }, + { + "epoch": 3.4797788769275533, + "flow_matching_loss": 0.02590877525508404, + "grad_norm": 0.26602450013160706, + "learning_rate": 2.820820534737802e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01162428930401802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051426135934889315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028391192294657232, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9052044451236725, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11960, + "weighted_flow_matching_loss": 0.02590877525508404, + "weighted_rkd_loss": 0.025699879601597787 + }, + { + "epoch": 3.4826883910386965, + "flow_matching_loss": 0.02842563223093748, + "grad_norm": 0.27157360315322876, + "learning_rate": 2.8204284958606946e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011637457273900509, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005475395545363426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028750310093164443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9050510168075562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11970, + "weighted_flow_matching_loss": 0.02842563223093748, + "weighted_rkd_loss": 0.026020583137869834 + }, + { + "epoch": 3.48559790514984, + "flow_matching_loss": 0.026459801755845547, + "grad_norm": 0.2587122619152069, + "learning_rate": 2.8200360558722738e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012227979488670826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005432150233536959, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029888109304010868, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9048974633216857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11980, + "weighted_flow_matching_loss": 0.026459801755845547, + "weighted_rkd_loss": 0.027045702934265135 + }, + { + "epoch": 3.4885074192609835, + "flow_matching_loss": 0.025293862260878087, + "grad_norm": 0.27674636244773865, + "learning_rate": 2.819643214891753e-05, + "loss": 0.0522, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011934631876647473, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005859113833867014, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029728377237915993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9047438085079194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 11990, + "weighted_flow_matching_loss": 0.025293862260878087, + "weighted_rkd_loss": 0.026896635815501214 + }, + { + "epoch": 3.4914169333721268, + "flow_matching_loss": 0.02681249175220728, + "grad_norm": 0.33524879813194275, + "learning_rate": 2.8192499730384668e-05, + "loss": 0.0568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013442054577171802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006257736729457975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03314184565097093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9045900404453278, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12000, + "weighted_flow_matching_loss": 0.02681249175220728, + "weighted_rkd_loss": 0.029979789257049562 + }, + { + "epoch": 3.4943264474832705, + "flow_matching_loss": 0.026602739281952382, + "grad_norm": 0.28304463624954224, + "learning_rate": 2.818856330431871e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011835047975182533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005441598128527403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029111694172024728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9044361531734466, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12010, + "weighted_flow_matching_loss": 0.026602739281952382, + "weighted_rkd_loss": 0.02632969617843628 + }, + { + "epoch": 3.4972359615944137, + "flow_matching_loss": 0.02771972753107548, + "grad_norm": 0.26612287759780884, + "learning_rate": 2.8184622871915446e-05, + "loss": 0.0568, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013042826391756534, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006121084513142705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03220673743635416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9042821645736694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12020, + "weighted_flow_matching_loss": 0.02771972753107548, + "weighted_rkd_loss": 0.02912395466119051 + }, + { + "epoch": 3.500145475705557, + "flow_matching_loss": 0.027078800648450852, + "grad_norm": 0.28448301553726196, + "learning_rate": 2.8180678434371874e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011378044076263905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005148078221827746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027904166467487813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9041280567646026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12030, + "weighted_flow_matching_loss": 0.027078800648450852, + "weighted_rkd_loss": 0.025228955782949924 + }, + { + "epoch": 3.5030549898167007, + "flow_matching_loss": 0.026789124682545663, + "grad_norm": 0.2952808141708374, + "learning_rate": 2.817672999288621e-05, + "loss": 0.0553, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012907637283205986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005753241153433919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03156851623207331, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9039738476276398, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12040, + "weighted_flow_matching_loss": 0.026789124682545663, + "weighted_rkd_loss": 0.028537140414118766 + }, + { + "epoch": 3.505964503927844, + "flow_matching_loss": 0.02673572041094303, + "grad_norm": 0.2772928774356842, + "learning_rate": 2.8172777548657886e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011537648923695087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005174584873020649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028249883092939854, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9038195312023163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12050, + "weighted_flow_matching_loss": 0.02673572041094303, + "weighted_rkd_loss": 0.025532791577279567 + }, + { + "epoch": 3.5088740180389877, + "flow_matching_loss": 0.027370340935885907, + "grad_norm": 0.24495691061019897, + "learning_rate": 2.8168821102887545e-05, + "loss": 0.0549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012399409990757703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005712933256290853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03051175270229578, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9036650836467743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12060, + "weighted_flow_matching_loss": 0.027370340935885907, + "weighted_rkd_loss": 0.027572578750550747 + }, + { + "epoch": 3.511783532150131, + "flow_matching_loss": 0.027283293567597865, + "grad_norm": 0.3051653504371643, + "learning_rate": 2.816486065677706e-05, + "loss": 0.0548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012405326776206493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005594047484919429, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030404700711369515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9035105407238007, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12070, + "weighted_flow_matching_loss": 0.027283293567597865, + "weighted_rkd_loss": 0.02747086137533188 + }, + { + "epoch": 3.514693046261274, + "flow_matching_loss": 0.029180257953703402, + "grad_norm": 0.2694762349128723, + "learning_rate": 2.8160896211529506e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011820072401314973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005551497871056199, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0291916424408555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9033558905124665, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12080, + "weighted_flow_matching_loss": 0.029180257953703402, + "weighted_rkd_loss": 0.026370465010404586 + }, + { + "epoch": 3.517602560372418, + "flow_matching_loss": 0.024138720519840718, + "grad_norm": 0.39774614572525024, + "learning_rate": 2.8156927768349175e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011949437856674194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005822377698495984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029721253365278245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9032011151313781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12090, + "weighted_flow_matching_loss": 0.024138720519840718, + "weighted_rkd_loss": 0.026844168640673162 + }, + { + "epoch": 3.520512074483561, + "flow_matching_loss": 0.02825115490704775, + "grad_norm": 0.33998823165893555, + "learning_rate": 2.815295532844158e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012404907960444689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0054597820620983836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03026959802955389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9030462443828583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12100, + "weighted_flow_matching_loss": 0.02825115490704775, + "weighted_rkd_loss": 0.02733487691730261 + }, + { + "epoch": 3.5234215885947044, + "flow_matching_loss": 0.02799985222518444, + "grad_norm": 0.2725108563899994, + "learning_rate": 2.8148978893013434e-05, + "loss": 0.0569, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013008294347673654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005955368699505925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03197195772081614, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9028912544250488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12110, + "weighted_flow_matching_loss": 0.02799985222518444, + "weighted_rkd_loss": 0.028867222741246222 + }, + { + "epoch": 3.526331102705848, + "flow_matching_loss": 0.02972047533839941, + "grad_norm": 0.35334789752960205, + "learning_rate": 2.814499846327268e-05, + "loss": 0.0572, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01243528090417385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0055966224987059835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03046718444675207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9027361512184143, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12120, + "weighted_flow_matching_loss": 0.02972047533839941, + "weighted_rkd_loss": 0.02750383336097002 + }, + { + "epoch": 3.5292406168169914, + "flow_matching_loss": 0.027745452523231507, + "grad_norm": 0.28595277667045593, + "learning_rate": 2.8141014040428468e-05, + "loss": 0.0554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012496715132147073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005614933464676142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030608363822102545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9025809466838837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12130, + "weighted_flow_matching_loss": 0.027745452523231507, + "weighted_rkd_loss": 0.02762657105922699 + }, + { + "epoch": 3.532150130928135, + "flow_matching_loss": 0.02809136565774679, + "grad_norm": 0.26682373881340027, + "learning_rate": 2.8137025625691163e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011964751314371825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005298221996054053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029227724112570286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9024256289005279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12140, + "weighted_flow_matching_loss": 0.02809136565774679, + "weighted_rkd_loss": 0.02637581340968609 + }, + { + "epoch": 3.5350596450392784, + "flow_matching_loss": 0.02761305756866932, + "grad_norm": 0.3511009216308594, + "learning_rate": 2.813303322027233e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011372678354382515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005201702681370079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027947059832513332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9022701978683472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12150, + "weighted_flow_matching_loss": 0.02761305756866932, + "weighted_rkd_loss": 0.02521574031561613 + }, + { + "epoch": 3.537969159150422, + "flow_matching_loss": 0.02607668414711952, + "grad_norm": 0.27793195843696594, + "learning_rate": 2.8129036825384763e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012442083936184645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005393389309756458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03027755729854107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9021146595478058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12160, + "weighted_flow_matching_loss": 0.02607668414711952, + "weighted_rkd_loss": 0.027313925698399542 + }, + { + "epoch": 3.5408786732615654, + "flow_matching_loss": 0.029521657153964043, + "grad_norm": 0.256722629070282, + "learning_rate": 2.8125036442242466e-05, + "loss": 0.0555, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011586340237408876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005615006247535348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028787686862051486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9019590079784393, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12170, + "weighted_flow_matching_loss": 0.029521657153964043, + "weighted_rkd_loss": 0.025965241156518458 + }, + { + "epoch": 3.5437881873727086, + "flow_matching_loss": 0.025957159325480463, + "grad_norm": 0.3279554545879364, + "learning_rate": 2.8121032072060635e-05, + "loss": 0.0539, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012599680572748184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0058334704488515856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031032831780612467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9018032431602478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12180, + "weighted_flow_matching_loss": 0.025957159325480463, + "weighted_rkd_loss": 0.027985533513128758 + }, + { + "epoch": 3.5466977014838523, + "flow_matching_loss": 0.02933406364172697, + "grad_norm": 0.31222259998321533, + "learning_rate": 2.811702371605571e-05, + "loss": 0.0556, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012052114680409431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004996798327192664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02910102792084217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9016473710536956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12190, + "weighted_flow_matching_loss": 0.02933406364172697, + "weighted_rkd_loss": 0.026238895393908022 + }, + { + "epoch": 3.5496072155949956, + "flow_matching_loss": 0.02594509720802307, + "grad_norm": 0.34008556604385376, + "learning_rate": 2.8113011375445308e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011798183899372816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005098619242198765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02869498748332262, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9014913976192475, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12200, + "weighted_flow_matching_loss": 0.02594509720802307, + "weighted_rkd_loss": 0.025868166238069534 + }, + { + "epoch": 3.552516729706139, + "flow_matching_loss": 0.027131596207618715, + "grad_norm": 0.2885300815105438, + "learning_rate": 2.8108995051448284e-05, + "loss": 0.056, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.013147974107414483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005763760162517428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03205970786511898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9013352990150452, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12210, + "weighted_flow_matching_loss": 0.027131596207618715, + "weighted_rkd_loss": 0.028896507620811463 + }, + { + "epoch": 3.5554262438172826, + "flow_matching_loss": 0.028029217943549158, + "grad_norm": 0.28589439392089844, + "learning_rate": 2.8104974745284685e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011157127190381288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005029606306925416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027343860641121865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9011791050434113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12220, + "weighted_flow_matching_loss": 0.028029217943549158, + "weighted_rkd_loss": 0.024641689285635947 + }, + { + "epoch": 3.558335757928426, + "flow_matching_loss": 0.02658377233892679, + "grad_norm": 0.2780078053474426, + "learning_rate": 2.8100950458175775e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011285151075571775, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004789346340112388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02735964860767126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9010227918624878, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12230, + "weighted_flow_matching_loss": 0.02658377233892679, + "weighted_rkd_loss": 0.024651660583913326 + }, + { + "epoch": 3.5612452720395695, + "flow_matching_loss": 0.02651757877320051, + "grad_norm": 0.28113624453544617, + "learning_rate": 2.8096922191344027e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01207858631387353, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005635341955348849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029792514443397523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9008663773536683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12240, + "weighted_flow_matching_loss": 0.02651757877320051, + "weighted_rkd_loss": 0.02683902159333229 + }, + { + "epoch": 3.564154786150713, + "flow_matching_loss": 0.026499926298856735, + "grad_norm": 0.27619847655296326, + "learning_rate": 2.809288994601312e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010492614563554525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004778141062706709, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025763370282948016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9007098436355591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12250, + "weighted_flow_matching_loss": 0.026499926298856735, + "weighted_rkd_loss": 0.02320532351732254 + }, + { + "epoch": 3.5670643002618565, + "flow_matching_loss": 0.027464785426855088, + "grad_norm": 0.2662321925163269, + "learning_rate": 2.8088853723407946e-05, + "loss": 0.0537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01185218570753932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005405934527516365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029110305570065977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9005531907081604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12260, + "weighted_flow_matching_loss": 0.027464785426855088, + "weighted_rkd_loss": 0.026215382479131223 + }, + { + "epoch": 3.5699738143729998, + "flow_matching_loss": 0.026035672053694725, + "grad_norm": 0.2887842059135437, + "learning_rate": 2.8084813524754602e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01114858891814947, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004866376612335443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027163554355502127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9003964364528656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12270, + "weighted_flow_matching_loss": 0.026035672053694725, + "weighted_rkd_loss": 0.02445788457989693 + }, + { + "epoch": 3.572883328484143, + "flow_matching_loss": 0.02658128272742033, + "grad_norm": 0.2666761875152588, + "learning_rate": 2.8080769351280395e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011644503148272634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005285676172934473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02857468221336603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9002395927906036, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12280, + "weighted_flow_matching_loss": 0.02658128272742033, + "weighted_rkd_loss": 0.025723999738693236 + }, + { + "epoch": 3.5757928425952867, + "flow_matching_loss": 0.025294272787868978, + "grad_norm": 0.3093181550502777, + "learning_rate": 2.8076721204213835e-05, + "loss": 0.0522, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012201012670993805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005465559754520654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02986758518964052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.9000826179981232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12290, + "weighted_flow_matching_loss": 0.025294272787868978, + "weighted_rkd_loss": 0.026883256807923316 + }, + { + "epoch": 3.57870235670643, + "flow_matching_loss": 0.027415824495255946, + "grad_norm": 0.26349812746047974, + "learning_rate": 2.8072669084784642e-05, + "loss": 0.0533, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01171921081840992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005349188670516014, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0287876108661294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8999255359172821, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12300, + "weighted_flow_matching_loss": 0.027415824495255946, + "weighted_rkd_loss": 0.02590679917484522 + }, + { + "epoch": 3.5816118708175733, + "flow_matching_loss": 0.02506019249558449, + "grad_norm": 0.29239121079444885, + "learning_rate": 2.8068612994223747e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011616716813296079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005005198973231017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0282386327162385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8997683584690094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12310, + "weighted_flow_matching_loss": 0.02506019249558449, + "weighted_rkd_loss": 0.02540820948779583 + }, + { + "epoch": 3.584521384928717, + "flow_matching_loss": 0.025681797601282596, + "grad_norm": 0.351337730884552, + "learning_rate": 2.8064552933763284e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01246136324480176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005482748383656144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03040547501295805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8996110558509827, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12320, + "weighted_flow_matching_loss": 0.025681797601282596, + "weighted_rkd_loss": 0.027353030070662497 + }, + { + "epoch": 3.5874308990398602, + "flow_matching_loss": 0.025059444084763526, + "grad_norm": 0.30144932866096497, + "learning_rate": 2.8060488904636586e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01129734069108963, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005369194084778428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02796387579292059, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8994536459445953, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12330, + "weighted_flow_matching_loss": 0.025059444084763526, + "weighted_rkd_loss": 0.025152184441685675 + }, + { + "epoch": 3.590340413151004, + "flow_matching_loss": 0.027347013913095, + "grad_norm": 0.292246013879776, + "learning_rate": 2.80564209080782e-05, + "loss": 0.0533, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011797736957669258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005297420732676983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02889289427548647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8992961347103119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12340, + "weighted_flow_matching_loss": 0.027347013913095, + "weighted_rkd_loss": 0.02598336637020111 + }, + { + "epoch": 3.593249927262147, + "flow_matching_loss": 0.027416902221739294, + "grad_norm": 0.27228760719299316, + "learning_rate": 2.8052348945323877e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01180382864549756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0054875562433153394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029095213674008847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8991385042667389, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12350, + "weighted_flow_matching_loss": 0.027416902221739294, + "weighted_rkd_loss": 0.026160511188209057 + }, + { + "epoch": 3.596159441373291, + "flow_matching_loss": 0.027832816168665887, + "grad_norm": 0.2815788686275482, + "learning_rate": 2.8048273017610574e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011352465115487575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051242390414699916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027829169668257235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8989807665348053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12360, + "weighted_flow_matching_loss": 0.027832816168665887, + "weighted_rkd_loss": 0.02501786109060049 + }, + { + "epoch": 3.599068955484434, + "flow_matching_loss": 0.026932628825306892, + "grad_norm": 0.25964149832725525, + "learning_rate": 2.804419312617645e-05, + "loss": 0.0549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012710334360599517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005659840675070882, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031080509535968304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8988229215145112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12370, + "weighted_flow_matching_loss": 0.026932628825306892, + "weighted_rkd_loss": 0.027935906872153284 + }, + { + "epoch": 3.6019784695955774, + "flow_matching_loss": 0.028829673863947393, + "grad_norm": 0.2898681163787842, + "learning_rate": 2.8040109272260862e-05, + "loss": 0.0561, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012412661500275136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0055652808398008345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03039060402661562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8986649632453918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12380, + "weighted_flow_matching_loss": 0.028829673863947393, + "weighted_rkd_loss": 0.027310889028012753 + }, + { + "epoch": 3.604887983706721, + "flow_matching_loss": 0.02650059647858143, + "grad_norm": 0.2863750755786896, + "learning_rate": 2.8036021457104384e-05, + "loss": 0.0523, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011705296486616135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052726796828210356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028683272562921046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.898506897687912, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12390, + "weighted_flow_matching_loss": 0.02650059647858143, + "weighted_rkd_loss": 0.02577202059328556 + }, + { + "epoch": 3.6077974978178644, + "flow_matching_loss": 0.029345091059803964, + "grad_norm": 0.292425274848938, + "learning_rate": 2.803192968194878e-05, + "loss": 0.0552, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01164948744699359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005510644777677954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02880961913615465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8983487188816071, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12400, + "weighted_flow_matching_loss": 0.029345091059803964, + "weighted_rkd_loss": 0.025881193205714227 + }, + { + "epoch": 3.6107070119290077, + "flow_matching_loss": 0.0278054840862751, + "grad_norm": 0.2445162683725357, + "learning_rate": 2.8027833948037032e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011301807314157485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053437584545463325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02794737312942743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8981904447078705, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12410, + "weighted_flow_matching_loss": 0.0278054840862751, + "weighted_rkd_loss": 0.025102140568196773 + }, + { + "epoch": 3.6136165260401514, + "flow_matching_loss": 0.026991719752550124, + "grad_norm": 0.2718099057674408, + "learning_rate": 2.802373425661331e-05, + "loss": 0.0536, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01219551870599389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005240832781419158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029631870426237585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8980320453643799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12420, + "weighted_flow_matching_loss": 0.026991719752550124, + "weighted_rkd_loss": 0.026610343530774117 + }, + { + "epoch": 3.6165260401512946, + "flow_matching_loss": 0.028754860535264014, + "grad_norm": 0.26854586601257324, + "learning_rate": 2.8019630608922988e-05, + "loss": 0.0554, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012038232665508986, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005643681879155338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02972014732658863, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8978735506534576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12430, + "weighted_flow_matching_loss": 0.028754860535264014, + "weighted_rkd_loss": 0.026684905961155893 + }, + { + "epoch": 3.619435554262438, + "flow_matching_loss": 0.027571089938282965, + "grad_norm": 0.27525097131729126, + "learning_rate": 2.801552300621265e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0112639089114964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0054659077664837245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02799372561275959, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8977149367332459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12440, + "weighted_flow_matching_loss": 0.027571089938282965, + "weighted_rkd_loss": 0.02513038869947195 + }, + { + "epoch": 3.6223450683735816, + "flow_matching_loss": 0.028372657112777235, + "grad_norm": 0.3411654233932495, + "learning_rate": 2.8011411449730084e-05, + "loss": 0.0541, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011712577566504479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052207197528332475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028645874932408333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8975562214851379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12450, + "weighted_flow_matching_loss": 0.028372657112777235, + "weighted_rkd_loss": 0.025711226835846902 + }, + { + "epoch": 3.6252545824847253, + "flow_matching_loss": 0.026418265886604786, + "grad_norm": 0.3503919243812561, + "learning_rate": 2.800729594072426e-05, + "loss": 0.0524, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011711028777062893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005492773046717048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028914830833673476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8973973870277405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12460, + "weighted_flow_matching_loss": 0.026418265886604786, + "weighted_rkd_loss": 0.02594806011766195 + }, + { + "epoch": 3.6281640965958686, + "flow_matching_loss": 0.028436385840177537, + "grad_norm": 0.2893011271953583, + "learning_rate": 2.8003176480445373e-05, + "loss": 0.0557, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012369603104889392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005604966869577765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03034417238086462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8972384512424469, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12470, + "weighted_flow_matching_loss": 0.028436385840177537, + "weighted_rkd_loss": 0.027225897274911405 + }, + { + "epoch": 3.631073610707012, + "flow_matching_loss": 0.026877039670944215, + "grad_norm": 0.2720698118209839, + "learning_rate": 2.7999053070144793e-05, + "loss": 0.0527, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011612758412957192, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005591328768059611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02881684564054012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8970794141292572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12480, + "weighted_flow_matching_loss": 0.026877039670944215, + "weighted_rkd_loss": 0.02585099097341299 + }, + { + "epoch": 3.6339831248181556, + "flow_matching_loss": 0.02644737549126148, + "grad_norm": 0.25207117199897766, + "learning_rate": 2.7994925711075117e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011908992193639278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005577021325007081, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02939500566571951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.896920257806778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12490, + "weighted_flow_matching_loss": 0.02644737549126148, + "weighted_rkd_loss": 0.0263650331646204 + }, + { + "epoch": 3.636892638929299, + "flow_matching_loss": 0.025766540691256522, + "grad_norm": 0.29830068349838257, + "learning_rate": 2.7990794404490118e-05, + "loss": 0.052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012089756689965725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005046041216701269, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02922555450350046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8967609822750091, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12500, + "weighted_flow_matching_loss": 0.025766540691256522, + "weighted_rkd_loss": 0.026208346337080003 + }, + { + "epoch": 3.639802153040442, + "flow_matching_loss": 0.02864628694951534, + "grad_norm": 0.34062105417251587, + "learning_rate": 2.798665915164478e-05, + "loss": 0.0562, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01250607306137681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005712228128686547, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030724374018609522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8966016113758087, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12510, + "weighted_flow_matching_loss": 0.02864628694951534, + "weighted_rkd_loss": 0.02754747588187456 + }, + { + "epoch": 3.642711667151586, + "flow_matching_loss": 0.027072853967547418, + "grad_norm": 0.3474093973636627, + "learning_rate": 2.7982519953795284e-05, + "loss": 0.055, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01263094050809741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005869433656334877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.031131314858794213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8964421331882477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12520, + "weighted_flow_matching_loss": 0.027072853967547418, + "weighted_rkd_loss": 0.027907355315983297 + }, + { + "epoch": 3.645621181262729, + "flow_matching_loss": 0.02628643661737442, + "grad_norm": 0.2502788305282593, + "learning_rate": 2.7978376812199013e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011031572893261909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005056371865794063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02711951769888401, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.896282535791397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12530, + "weighted_flow_matching_loss": 0.02628643661737442, + "weighted_rkd_loss": 0.024306782707571983 + }, + { + "epoch": 3.6485306953738723, + "flow_matching_loss": 0.02833595909178257, + "grad_norm": 0.22650402784347534, + "learning_rate": 2.797422972811454e-05, + "loss": 0.0548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012133886385709048, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005297480151057243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02956525273621082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8961228489875793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12540, + "weighted_flow_matching_loss": 0.02833595909178257, + "weighted_rkd_loss": 0.02649410180747509 + }, + { + "epoch": 3.651440209485016, + "flow_matching_loss": 0.02524854354560375, + "grad_norm": 0.33042705059051514, + "learning_rate": 2.7970078702801635e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011967917252331973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00534253369551152, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02927836813032627, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8959630370140076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12550, + "weighted_flow_matching_loss": 0.02524854354560375, + "weighted_rkd_loss": 0.026232211478054525 + }, + { + "epoch": 3.6543497235961593, + "flow_matching_loss": 0.027265935204923153, + "grad_norm": 0.2686721682548523, + "learning_rate": 2.7965923737521283e-05, + "loss": 0.0549, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012557264138013124, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005767538165673614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03088206648826599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8958031177520752, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12560, + "weighted_flow_matching_loss": 0.027265935204923153, + "weighted_rkd_loss": 0.027664310485124587 + }, + { + "epoch": 3.657259237707303, + "flow_matching_loss": 0.027033516578376292, + "grad_norm": 0.2712067663669586, + "learning_rate": 2.796176483353564e-05, + "loss": 0.0532, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01175735667347908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005708168121054768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0292228814214468, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8956430971622467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12570, + "weighted_flow_matching_loss": 0.027033516578376292, + "weighted_rkd_loss": 0.026173322275280953 + }, + { + "epoch": 3.6601687518184463, + "flow_matching_loss": 0.026865703612565996, + "grad_norm": 0.3100869357585907, + "learning_rate": 2.795760199210808e-05, + "loss": 0.0531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011975393164902925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0052993704564869406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029250156693160535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8954829633235931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12580, + "weighted_flow_matching_loss": 0.026865703612565996, + "weighted_rkd_loss": 0.026192979514598848 + }, + { + "epoch": 3.66307826592959, + "flow_matching_loss": 0.0267293993383646, + "grad_norm": 0.25387656688690186, + "learning_rate": 2.7953435214503157e-05, + "loss": 0.0535, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012169357296079398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005562996817752719, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0299017108976841, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.89532271027565, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12590, + "weighted_flow_matching_loss": 0.0267293993383646, + "weighted_rkd_loss": 0.02677177619189024 + }, + { + "epoch": 3.6659877800407332, + "flow_matching_loss": 0.027392338030040264, + "grad_norm": 0.3153534531593323, + "learning_rate": 2.7949264501986634e-05, + "loss": 0.0534, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011905057355761528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005286634480580687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029096749052405356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8951623678207398, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12600, + "weighted_flow_matching_loss": 0.027392338030040264, + "weighted_rkd_loss": 0.026046359352767466 + }, + { + "epoch": 3.6688972941518765, + "flow_matching_loss": 0.02643189877271652, + "grad_norm": 0.2867465615272522, + "learning_rate": 2.794508985582546e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011210813093930483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0055176008492708204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027939227223396302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8950019121170044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12610, + "weighted_flow_matching_loss": 0.02643189877271652, + "weighted_rkd_loss": 0.025005638785660266 + }, + { + "epoch": 3.67180680826302, + "flow_matching_loss": 0.02572328522801399, + "grad_norm": 0.2663887143135071, + "learning_rate": 2.7940911277287777e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011482676025480032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005165186664089561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028130538761615753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.894841343164444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12620, + "weighted_flow_matching_loss": 0.02572328522801399, + "weighted_rkd_loss": 0.025172423757612705 + }, + { + "epoch": 3.6747163223741635, + "flow_matching_loss": 0.026350127160549165, + "grad_norm": 0.3195253312587738, + "learning_rate": 2.793672876764294e-05, + "loss": 0.0524, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01190609224140644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005296170944347978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029108355566859246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.894680666923523, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12630, + "weighted_flow_matching_loss": 0.026350127160549165, + "weighted_rkd_loss": 0.026042680814862253 + }, + { + "epoch": 3.6776258364853067, + "flow_matching_loss": 0.02796841412782669, + "grad_norm": 0.28530144691467285, + "learning_rate": 2.7932542328161478e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010888065956532956, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004832833423279226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0266089653596282, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8945198953151703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12640, + "weighted_flow_matching_loss": 0.02796841412782669, + "weighted_rkd_loss": 0.023802299052476883 + }, + { + "epoch": 3.6805353505964504, + "flow_matching_loss": 0.026085536740720273, + "grad_norm": 0.28245359659194946, + "learning_rate": 2.792835196011512e-05, + "loss": 0.0528, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012196190375834703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005469011468812823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02986139226704836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8943589925765991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12650, + "weighted_flow_matching_loss": 0.026085536740720273, + "weighted_rkd_loss": 0.026706817001104353 + }, + { + "epoch": 3.6834448647075937, + "flow_matching_loss": 0.025784032046794893, + "grad_norm": 0.24573473632335663, + "learning_rate": 2.792415766477679e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012451891787350178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005536827584728598, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0304406113922596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8941979944705963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12660, + "weighted_flow_matching_loss": 0.025784032046794893, + "weighted_rkd_loss": 0.02722000386565924 + }, + { + "epoch": 3.6863543788187374, + "flow_matching_loss": 0.028027025610208513, + "grad_norm": 0.28782275319099426, + "learning_rate": 2.79199594434206e-05, + "loss": 0.0504, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010187968611717224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004674926050938666, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025050863064825533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8940368950366974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12670, + "weighted_flow_matching_loss": 0.028027025610208513, + "weighted_rkd_loss": 0.022396328672766686 + }, + { + "epoch": 3.6892638929298807, + "flow_matching_loss": 0.027778849564492704, + "grad_norm": 0.30332300066947937, + "learning_rate": 2.7915757297321867e-05, + "loss": 0.0532, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011643077619373799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005131897469982505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02841805275529623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.893875676393509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12680, + "weighted_flow_matching_loss": 0.027778849564492704, + "weighted_rkd_loss": 0.025402160175144673 + }, + { + "epoch": 3.6921734070410244, + "flow_matching_loss": 0.02618682198226452, + "grad_norm": 0.2622012495994568, + "learning_rate": 2.7911551227757085e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011562467645853758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005465654958970845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028590590134263037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8937143504619598, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12690, + "weighted_flow_matching_loss": 0.02618682198226452, + "weighted_rkd_loss": 0.02555176168680191 + }, + { + "epoch": 3.6950829211521676, + "flow_matching_loss": 0.027789635211229326, + "grad_norm": 0.36006036400794983, + "learning_rate": 2.790734123600395e-05, + "loss": 0.0548, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01235045911744237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005564433475956321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03026535175740719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8935529232025147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12700, + "weighted_flow_matching_loss": 0.027789635211229326, + "weighted_rkd_loss": 0.027043644897639753 + }, + { + "epoch": 3.697992435263311, + "flow_matching_loss": 0.029256391711533068, + "grad_norm": 0.2528729736804962, + "learning_rate": 2.7903127323341347e-05, + "loss": 0.054, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01130783036351204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005107303475961089, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027722964249551296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8933913826942443, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12710, + "weighted_flow_matching_loss": 0.029256391711533068, + "weighted_rkd_loss": 0.024767505750060082 + }, + { + "epoch": 3.7009019493744546, + "flow_matching_loss": 0.026313228905200957, + "grad_norm": 0.2564309239387512, + "learning_rate": 2.7898909491049353e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01095514725893736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047759017907083035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026686196215450764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8932297348976135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12720, + "weighted_flow_matching_loss": 0.026313228905200957, + "weighted_rkd_loss": 0.023836841620504857 + }, + { + "epoch": 3.703811463485598, + "flow_matching_loss": 0.0273601021617651, + "grad_norm": 0.2895980179309845, + "learning_rate": 2.789468774040923e-05, + "loss": 0.0545, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012263027392327786, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005813894188031554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030339948274195193, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8930679857730865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12730, + "weighted_flow_matching_loss": 0.0273601021617651, + "weighted_rkd_loss": 0.027095660008490084 + }, + { + "epoch": 3.706720977596741, + "flow_matching_loss": 0.028166887909173967, + "grad_norm": 0.2513412833213806, + "learning_rate": 2.7890462072703433e-05, + "loss": 0.0564, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012793628871440888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.006031367927789688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03161862548440695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8929061174392701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12740, + "weighted_flow_matching_loss": 0.028166887909173967, + "weighted_rkd_loss": 0.02823255266994238 + }, + { + "epoch": 3.709630491707885, + "flow_matching_loss": 0.028321937099099158, + "grad_norm": 0.26597100496292114, + "learning_rate": 2.7886232489215617e-05, + "loss": 0.0531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01136667001992464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005064710672013461, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027798050455749033, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8927441537380219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12750, + "weighted_flow_matching_loss": 0.028321937099099158, + "weighted_rkd_loss": 0.024816630221903325 + }, + { + "epoch": 3.712540005819028, + "flow_matching_loss": 0.026867646723985672, + "grad_norm": 0.35197243094444275, + "learning_rate": 2.7881998991230606e-05, + "loss": 0.0544, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012521640677005053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005788855254650116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030832136794924737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8925820589065552, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12760, + "weighted_flow_matching_loss": 0.026867646723985672, + "weighted_rkd_loss": 0.02752019837498665 + }, + { + "epoch": 3.715449519930172, + "flow_matching_loss": 0.0266194187104702, + "grad_norm": 0.2303682565689087, + "learning_rate": 2.7877761580034436e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01104914015159011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005235198792070151, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027333479002118112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8924198806285858, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12770, + "weighted_flow_matching_loss": 0.0266194187104702, + "weighted_rkd_loss": 0.024393046088516713 + }, + { + "epoch": 3.718359034041315, + "flow_matching_loss": 0.02454210855066776, + "grad_norm": 0.23693984746932983, + "learning_rate": 2.787352025691431e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011741577461361886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005211515305563808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028694669902324676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8922575950622559, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12780, + "weighted_flow_matching_loss": 0.02454210855066776, + "weighted_rkd_loss": 0.025603056140244006 + }, + { + "epoch": 3.7212685481524588, + "flow_matching_loss": 0.02612473350018263, + "grad_norm": 0.3174750804901123, + "learning_rate": 2.7869275023158642e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01153044318780303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005323377810418606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028384264558553696, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8920951902866363, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12790, + "weighted_flow_matching_loss": 0.02612473350018263, + "weighted_rkd_loss": 0.025321424566209318 + }, + { + "epoch": 3.724178062263602, + "flow_matching_loss": 0.026133371889591216, + "grad_norm": 0.31716516613960266, + "learning_rate": 2.7865025880057014e-05, + "loss": 0.052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01178652998059988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005422074487432838, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028995134495198728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8919326782226562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12800, + "weighted_flow_matching_loss": 0.026133371889591216, + "weighted_rkd_loss": 0.025861670449376106 + }, + { + "epoch": 3.7270875763747453, + "flow_matching_loss": 0.028272339329123496, + "grad_norm": 0.2406080961227417, + "learning_rate": 2.78607728289002e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011125051137059926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00505329214502126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027303394116461276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8917700707912445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12810, + "weighted_flow_matching_loss": 0.028272339329123496, + "weighted_rkd_loss": 0.024348295480012893 + }, + { + "epoch": 3.729997090485889, + "flow_matching_loss": 0.026903258077800273, + "grad_norm": 0.34759247303009033, + "learning_rate": 2.7856515870980176e-05, + "loss": 0.0511, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011007546726614238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005078351218253374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027093444764614106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8916073441505432, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12820, + "weighted_flow_matching_loss": 0.026903258077800273, + "weighted_rkd_loss": 0.024156699329614638 + }, + { + "epoch": 3.7329066045970323, + "flow_matching_loss": 0.025487993471324445, + "grad_norm": 0.3217254877090454, + "learning_rate": 2.785225500759008e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010383191332221032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004555617272853851, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02532200012356043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8914445161819458, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12830, + "weighted_flow_matching_loss": 0.025487993471324445, + "weighted_rkd_loss": 0.022573154419660568 + }, + { + "epoch": 3.7358161187081755, + "flow_matching_loss": 0.026557007990777492, + "grad_norm": 0.27959319949150085, + "learning_rate": 2.7847990240024263e-05, + "loss": 0.0509, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011120570823550224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005016554379835725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02725769579410553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8912815749645233, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12840, + "weighted_flow_matching_loss": 0.026557007990777492, + "weighted_rkd_loss": 0.024294292740523814 + }, + { + "epoch": 3.7387256328193192, + "flow_matching_loss": 0.026666930690407754, + "grad_norm": 0.30147984623908997, + "learning_rate": 2.784372156957824e-05, + "loss": 0.0531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01196575267240405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0057197475340217355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02965125273913145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8911185324192047, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12850, + "weighted_flow_matching_loss": 0.026666930690407754, + "weighted_rkd_loss": 0.02642271537333727 + }, + { + "epoch": 3.7416351469304625, + "flow_matching_loss": 0.027637873589992524, + "grad_norm": 0.30764859914779663, + "learning_rate": 2.7839448997548728e-05, + "loss": 0.0522, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011043970845639706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005451939883641899, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027539881505072118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8909553766250611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12860, + "weighted_flow_matching_loss": 0.027637873589992524, + "weighted_rkd_loss": 0.024536703899502754 + }, + { + "epoch": 3.744544661041606, + "flow_matching_loss": 0.02769598364830017, + "grad_norm": 0.30152761936187744, + "learning_rate": 2.783517252523361e-05, + "loss": 0.052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011104132514446974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051182886119931935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027326553501188754, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8907921314239502, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12870, + "weighted_flow_matching_loss": 0.02769598364830017, + "weighted_rkd_loss": 0.02434226404875517 + }, + { + "epoch": 3.7474541751527495, + "flow_matching_loss": 0.025384044088423253, + "grad_norm": 0.37975114583969116, + "learning_rate": 2.7830892153931975e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011873556673526764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005290054995566606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02903716843575239, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8906287431716919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12880, + "weighted_flow_matching_loss": 0.025384044088423253, + "weighted_rkd_loss": 0.025861234962940217 + }, + { + "epoch": 3.750363689263893, + "flow_matching_loss": 0.026357887126505376, + "grad_norm": 0.29256582260131836, + "learning_rate": 2.7826607884944083e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010675756540149451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004794514738023281, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026146027818322183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8904652774333954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12890, + "weighted_flow_matching_loss": 0.026357887126505376, + "weighted_rkd_loss": 0.023282164707779884 + }, + { + "epoch": 3.7532732033750364, + "flow_matching_loss": 0.026252189464867116, + "grad_norm": 0.22459913790225983, + "learning_rate": 2.782231971957138e-05, + "loss": 0.0525, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012010885402560234, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00544297385495156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029464744590222837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8903016984462738, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12900, + "weighted_flow_matching_loss": 0.026252189464867116, + "weighted_rkd_loss": 0.02623253222554922 + }, + { + "epoch": 3.7561827174861797, + "flow_matching_loss": 0.029183872044086456, + "grad_norm": 0.34118202328681946, + "learning_rate": 2.78180276591165e-05, + "loss": 0.0539, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011252670921385289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005289388308301568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02779473029077053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8901380062103271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12910, + "weighted_flow_matching_loss": 0.029183872044086456, + "weighted_rkd_loss": 0.02474106978625059 + }, + { + "epoch": 3.7590922315973234, + "flow_matching_loss": 0.02675092127174139, + "grad_norm": 0.31345441937446594, + "learning_rate": 2.781373170488326e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011733091901987791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005526159983128309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028992343880236148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8899742066860199, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12920, + "weighted_flow_matching_loss": 0.02675092127174139, + "weighted_rkd_loss": 0.025802367925643922 + }, + { + "epoch": 3.7620017457084667, + "flow_matching_loss": 0.025260388664901257, + "grad_norm": 0.27355408668518066, + "learning_rate": 2.7809431858176655e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011434638686478139, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005444141826592386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02831341903656721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8898102939128876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12930, + "weighted_flow_matching_loss": 0.025260388664901257, + "weighted_rkd_loss": 0.02519352827221155 + }, + { + "epoch": 3.76491125981961, + "flow_matching_loss": 0.02699974160641432, + "grad_norm": 0.3041273355484009, + "learning_rate": 2.7805128120302865e-05, + "loss": 0.0537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012241229973733425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005554422782734036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.03003688305616379, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.889646303653717, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12940, + "weighted_flow_matching_loss": 0.02699974160641432, + "weighted_rkd_loss": 0.026722276210784913 + }, + { + "epoch": 3.7678207739307537, + "flow_matching_loss": 0.026192852668464182, + "grad_norm": 0.2571048140525818, + "learning_rate": 2.7800820492569254e-05, + "loss": 0.0508, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011319994181394576, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005053876386955381, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027693864516913892, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8894821882247925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12950, + "weighted_flow_matching_loss": 0.026192852668464182, + "weighted_rkd_loss": 0.024633194878697397 + }, + { + "epoch": 3.770730288041897, + "flow_matching_loss": 0.028155535459518433, + "grad_norm": 0.31451869010925293, + "learning_rate": 2.779650897628436e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011371613014489412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005058455700054765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02780168205499649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8893179655075073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12960, + "weighted_flow_matching_loss": 0.028155535459518433, + "weighted_rkd_loss": 0.0247245280072093 + }, + { + "epoch": 3.77363980215304, + "flow_matching_loss": 0.025127497501671314, + "grad_norm": 0.27652695775032043, + "learning_rate": 2.7792193572757915e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010734603460878133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004736871225759387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02620607800781727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8891536355018616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12970, + "weighted_flow_matching_loss": 0.025127497501671314, + "weighted_rkd_loss": 0.023301263712346552 + }, + { + "epoch": 3.776549316264184, + "flow_matching_loss": 0.02806299477815628, + "grad_norm": 0.272841215133667, + "learning_rate": 2.7787874283300812e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010798399429768324, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00511685167439282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026713650673627853, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8889892101287842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12980, + "weighted_flow_matching_loss": 0.02806299477815628, + "weighted_rkd_loss": 0.02374823559075594 + }, + { + "epoch": 3.7794588303753276, + "flow_matching_loss": 0.025094636902213096, + "grad_norm": 0.25580427050590515, + "learning_rate": 2.7783551109225155e-05, + "loss": 0.0486, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010786572378128767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004816307313740253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026389452069997786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8888246655464173, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 12990, + "weighted_flow_matching_loss": 0.025094636902213096, + "weighted_rkd_loss": 0.023455584794282912 + }, + { + "epoch": 3.782368344486471, + "flow_matching_loss": 0.027650189585983753, + "grad_norm": 0.31307724118232727, + "learning_rate": 2.7779224051844195e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011476153321564198, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005542214657180011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028494520857930184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8886600017547608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13000, + "weighted_flow_matching_loss": 0.027650189585983753, + "weighted_rkd_loss": 0.025321905873715876 + }, + { + "epoch": 3.785277858597614, + "flow_matching_loss": 0.02583218701183796, + "grad_norm": 0.2928609251976013, + "learning_rate": 2.777489311247239e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011543862894177437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005450590839609503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02853831648826599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8884952604770661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13010, + "weighted_flow_matching_loss": 0.02583218701183796, + "weighted_rkd_loss": 0.025356335565447807 + }, + { + "epoch": 3.788187372708758, + "flow_matching_loss": 0.028342588990926742, + "grad_norm": 0.26585856080055237, + "learning_rate": 2.7770558292425355e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011179441586136818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005236310278996825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027595193311572074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8883303940296173, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13020, + "weighted_flow_matching_loss": 0.028342588990926742, + "weighted_rkd_loss": 0.024513649940490722 + }, + { + "epoch": 3.791096886819901, + "flow_matching_loss": 0.027699468284845354, + "grad_norm": 0.25343307852745056, + "learning_rate": 2.77662195930199e-05, + "loss": 0.0518, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011031467095017433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0050396338338032365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02710256800055504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.888165420293808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13030, + "weighted_flow_matching_loss": 0.027699468284845354, + "weighted_rkd_loss": 0.02407166678458452 + }, + { + "epoch": 3.7940064009310444, + "flow_matching_loss": 0.024504953995347023, + "grad_norm": 0.26873233914375305, + "learning_rate": 2.7761877015574005e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01109085874632001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005206267512403428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02738798502832651, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8880003392696381, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13040, + "weighted_flow_matching_loss": 0.024504953995347023, + "weighted_rkd_loss": 0.02432053554803133 + }, + { + "epoch": 3.796915915042188, + "flow_matching_loss": 0.026512499153614043, + "grad_norm": 0.2872370779514313, + "learning_rate": 2.775753056140683e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011391067784279585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004853834304958582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027635969594120978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.887835168838501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13050, + "weighted_flow_matching_loss": 0.026512499153614043, + "weighted_rkd_loss": 0.024536203779280185 + }, + { + "epoch": 3.7998254291533313, + "flow_matching_loss": 0.025839274749159813, + "grad_norm": 0.30622076988220215, + "learning_rate": 2.775318023183871e-05, + "loss": 0.0504, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0112481027841568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005203901138156652, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027700106799602508, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8876698732376098, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13060, + "weighted_flow_matching_loss": 0.025839274749159813, + "weighted_rkd_loss": 0.024588498286902905 + }, + { + "epoch": 3.8027349432644746, + "flow_matching_loss": 0.027239222638309003, + "grad_norm": 0.3254048228263855, + "learning_rate": 2.774882602819117e-05, + "loss": 0.0533, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011850484926253558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005648763990029693, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029349734261631965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8875044822692871, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13070, + "weighted_flow_matching_loss": 0.027239222638309003, + "weighted_rkd_loss": 0.02604797650128603 + }, + { + "epoch": 3.8056444573756183, + "flow_matching_loss": 0.026949534378945827, + "grad_norm": 0.2815423905849457, + "learning_rate": 2.7744467951786892e-05, + "loss": 0.0531, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011956427711993456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005535547900944948, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029448403231799602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8873389899730683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13080, + "weighted_flow_matching_loss": 0.026949534378945827, + "weighted_rkd_loss": 0.026130637153983116 + }, + { + "epoch": 3.8085539714867616, + "flow_matching_loss": 0.02723674848675728, + "grad_norm": 0.3389657735824585, + "learning_rate": 2.7740106003949747e-05, + "loss": 0.0521, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0114368281327188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005116612371057272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027990268357098103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8871733844280243, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13090, + "weighted_flow_matching_loss": 0.02723674848675728, + "weighted_rkd_loss": 0.024832194857299327 + }, + { + "epoch": 3.8114634855979053, + "flow_matching_loss": 0.02753823958337307, + "grad_norm": 0.2747952342033386, + "learning_rate": 2.7735740186004785e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010543664125725627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004822030989453196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02590935956686735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8870076715946198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13100, + "weighted_flow_matching_loss": 0.02753823958337307, + "weighted_rkd_loss": 0.02298174351453781 + }, + { + "epoch": 3.8143729997090485, + "flow_matching_loss": 0.026598897017538547, + "grad_norm": 0.2559396028518677, + "learning_rate": 2.7731370499278222e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01131754433736205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00513487549033016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027769964188337326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8868418455123901, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13110, + "weighted_flow_matching_loss": 0.026598897017538547, + "weighted_rkd_loss": 0.024627568759024145 + }, + { + "epoch": 3.8172825138201922, + "flow_matching_loss": 0.02602277249097824, + "grad_norm": 0.284682959318161, + "learning_rate": 2.772699694509745e-05, + "loss": 0.0499, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010859125386923551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005203937878832221, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02692218907177448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8866759181022644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13120, + "weighted_flow_matching_loss": 0.02602277249097824, + "weighted_rkd_loss": 0.02387124076485634 + }, + { + "epoch": 3.8201920279313355, + "flow_matching_loss": 0.02899115402251482, + "grad_norm": 0.2459946721792221, + "learning_rate": 2.7722619524791046e-05, + "loss": 0.0555, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012073588650673628, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005783565435558557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029930742643773555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.886509895324707, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13130, + "weighted_flow_matching_loss": 0.02899115402251482, + "weighted_rkd_loss": 0.026533975079655648 + }, + { + "epoch": 3.8231015420424788, + "flow_matching_loss": 0.026887511648237706, + "grad_norm": 0.2986243665218353, + "learning_rate": 2.7718238239688747e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011230198945850135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005031233397312462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027491631358861922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8863437533378601, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13140, + "weighted_flow_matching_loss": 0.026887511648237706, + "weighted_rkd_loss": 0.024366985075175762 + }, + { + "epoch": 3.8260110561536225, + "flow_matching_loss": 0.024695346131920815, + "grad_norm": 0.25489720702171326, + "learning_rate": 2.7713853091121484e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010517847631126643, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004737789509817958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025773485004901887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8861775100231171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13150, + "weighted_flow_matching_loss": 0.024695346131920815, + "weighted_rkd_loss": 0.02283989954739809 + }, + { + "epoch": 3.8289205702647657, + "flow_matching_loss": 0.02713260632008314, + "grad_norm": 0.3367641270160675, + "learning_rate": 2.7709464080421338e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012007293477654456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005200594710186124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02921518199145794, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8860111713409424, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13160, + "weighted_flow_matching_loss": 0.02713260632008314, + "weighted_rkd_loss": 0.025885043665766716 + }, + { + "epoch": 3.831830084375909, + "flow_matching_loss": 0.028591879084706305, + "grad_norm": 0.264400839805603, + "learning_rate": 2.770507120892158e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010604563541710377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048204994760453704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026029626652598382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8858447074890137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13170, + "weighted_flow_matching_loss": 0.028591879084706305, + "weighted_rkd_loss": 0.023058293014764787 + }, + { + "epoch": 3.8347395984870527, + "flow_matching_loss": 0.02629763577133417, + "grad_norm": 0.2869505286216736, + "learning_rate": 2.7700674477956643e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011520834174007177, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051119168754667045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028153585083782674, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8856781482696533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13180, + "weighted_flow_matching_loss": 0.02629763577133417, + "weighted_rkd_loss": 0.024934931471943856 + }, + { + "epoch": 3.837649112598196, + "flow_matching_loss": 0.026066239923238754, + "grad_norm": 0.29450103640556335, + "learning_rate": 2.7696273888862145e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011381490621715784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005338452290743589, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028101433254778384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8855114817619324, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13190, + "weighted_flow_matching_loss": 0.026066239923238754, + "weighted_rkd_loss": 0.02488398402929306 + }, + { + "epoch": 3.8405586267093397, + "flow_matching_loss": 0.02562877759337425, + "grad_norm": 0.30632349848747253, + "learning_rate": 2.7691869442974864e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010691441874951125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004628208908252418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026011092588305473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8853447198867798, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13200, + "weighted_flow_matching_loss": 0.02562877759337425, + "weighted_rkd_loss": 0.02302872743457556 + }, + { + "epoch": 3.843468140820483, + "flow_matching_loss": 0.027065898664295672, + "grad_norm": 0.30354923009872437, + "learning_rate": 2.7687461141632754e-05, + "loss": 0.0519, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011461319494992494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005165392532944679, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028088031709194182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8851778447628021, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13210, + "weighted_flow_matching_loss": 0.027065898664295672, + "weighted_rkd_loss": 0.024863009341061115 + }, + { + "epoch": 3.8463776549316266, + "flow_matching_loss": 0.029636430181562902, + "grad_norm": 0.26124095916748047, + "learning_rate": 2.7683048986174945e-05, + "loss": 0.053, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01067721825093031, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004996818071231246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026351254805922507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8850108504295349, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13220, + "weighted_flow_matching_loss": 0.029636430181562902, + "weighted_rkd_loss": 0.023321189731359483 + }, + { + "epoch": 3.84928716904277, + "flow_matching_loss": 0.0265014810487628, + "grad_norm": 0.2956845462322235, + "learning_rate": 2.7678632977941727e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011509701982140541, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005457698041573167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028477101773023605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8848437666893005, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13230, + "weighted_flow_matching_loss": 0.0265014810487628, + "weighted_rkd_loss": 0.025197694636881353 + }, + { + "epoch": 3.852196683153913, + "flow_matching_loss": 0.025496860221028327, + "grad_norm": 0.2878592908382416, + "learning_rate": 2.767421311827457e-05, + "loss": 0.0515, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012002412229776382, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053315350320190195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029336359351873398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8846765697002411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13240, + "weighted_flow_matching_loss": 0.025496860221028327, + "weighted_rkd_loss": 0.025953212380409242 + }, + { + "epoch": 3.855106197265057, + "flow_matching_loss": 0.02733662407845259, + "grad_norm": 0.23083443939685822, + "learning_rate": 2.7669789408516117e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011188747733831406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005189539864659309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02756703533232212, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.88450927734375, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13250, + "weighted_flow_matching_loss": 0.02733662407845259, + "weighted_rkd_loss": 0.024383392371237277 + }, + { + "epoch": 3.8580157113762, + "flow_matching_loss": 0.02774400282651186, + "grad_norm": 0.3537035584449768, + "learning_rate": 2.766536185001016e-05, + "loss": 0.0537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011741571314632892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0058869221713393925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029370064847171307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8843418657779694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13260, + "weighted_flow_matching_loss": 0.02774400282651186, + "weighted_rkd_loss": 0.02597304228693247 + }, + { + "epoch": 3.8609252254873434, + "flow_matching_loss": 0.027409464307129382, + "grad_norm": 0.3216320276260376, + "learning_rate": 2.7660930444101692e-05, + "loss": 0.052, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011286353040486575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005205520382151008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027778226509690286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.884174358844757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13270, + "weighted_flow_matching_loss": 0.027409464307129382, + "weighted_rkd_loss": 0.024560685642063616 + }, + { + "epoch": 3.863834739598487, + "flow_matching_loss": 0.023977941274642943, + "grad_norm": 0.24741171300411224, + "learning_rate": 2.7656495192136842e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010874421522021293, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004848738433793187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02659758161753416, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8840067505836486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13280, + "weighted_flow_matching_loss": 0.023977941274642943, + "weighted_rkd_loss": 0.023512439243495463 + }, + { + "epoch": 3.8667442537096304, + "flow_matching_loss": 0.027785122394561768, + "grad_norm": 0.3262527585029602, + "learning_rate": 2.7652056095462926e-05, + "loss": 0.0502, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010295009054243565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004744384507648647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0253344027325511, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8838390231132507, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13290, + "weighted_flow_matching_loss": 0.027785122394561768, + "weighted_rkd_loss": 0.022391477786004543 + }, + { + "epoch": 3.869653767820774, + "flow_matching_loss": 0.027058246172964574, + "grad_norm": 0.2396751046180725, + "learning_rate": 2.764761315542843e-05, + "loss": 0.0507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010834058700129389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00509370700456202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0267618240788579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8836711943149567, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13300, + "weighted_flow_matching_loss": 0.027058246172964574, + "weighted_rkd_loss": 0.023648701049387454 + }, + { + "epoch": 3.8725632819319173, + "flow_matching_loss": 0.024874612502753735, + "grad_norm": 0.3051430583000183, + "learning_rate": 2.7643166373383e-05, + "loss": 0.0479, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010572504531592131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004862090293318033, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02600709944963455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8835032701492309, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13310, + "weighted_flow_matching_loss": 0.024874612502753735, + "weighted_rkd_loss": 0.0229773685336113 + }, + { + "epoch": 3.875472796043061, + "flow_matching_loss": 0.026364354789257048, + "grad_norm": 0.3123011589050293, + "learning_rate": 2.763871575067744e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011044285725802183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004976904578506946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027065475843846797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8833352327346802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13320, + "weighted_flow_matching_loss": 0.026364354789257048, + "weighted_rkd_loss": 0.02390781007707119 + }, + { + "epoch": 3.8783823101542043, + "flow_matching_loss": 0.027595572732388975, + "grad_norm": 0.3243313133716583, + "learning_rate": 2.763426128866375e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011143726017326116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004997543036006391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027284994907677173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8831670939922333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13330, + "weighted_flow_matching_loss": 0.027595572732388975, + "weighted_rkd_loss": 0.0240972300991416 + }, + { + "epoch": 3.8812918242653476, + "flow_matching_loss": 0.026950480230152608, + "grad_norm": 0.3119725286960602, + "learning_rate": 2.7629802988695064e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011318116262555123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004856833117082715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027493065781891347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8829988479614258, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13340, + "weighted_flow_matching_loss": 0.026950480230152608, + "weighted_rkd_loss": 0.024276230484247208 + }, + { + "epoch": 3.8842013383764913, + "flow_matching_loss": 0.026143408566713332, + "grad_norm": 0.27021923661231995, + "learning_rate": 2.7625340852125703e-05, + "loss": 0.0511, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011528588179498912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005265194713138044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02832237109541893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8828304946422577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13350, + "weighted_flow_matching_loss": 0.026143408566713332, + "weighted_rkd_loss": 0.025003904476761818 + }, + { + "epoch": 3.8871108524876345, + "flow_matching_loss": 0.027535645104944707, + "grad_norm": 0.2842022478580475, + "learning_rate": 2.7620874880311148e-05, + "loss": 0.0537, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012005998473614454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005682131578214467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029694128409028055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.882662034034729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13360, + "weighted_flow_matching_loss": 0.027535645104944707, + "weighted_rkd_loss": 0.02620977833867073 + }, + { + "epoch": 3.890020366598778, + "flow_matching_loss": 0.028099789470434188, + "grad_norm": 0.3006730079650879, + "learning_rate": 2.7616405074608033e-05, + "loss": 0.0551, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012531703058630227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005485527333803475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.030548933148384094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8824934780597686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13370, + "weighted_flow_matching_loss": 0.028099789470434188, + "weighted_rkd_loss": 0.026959181204438208 + }, + { + "epoch": 3.8929298807099215, + "flow_matching_loss": 0.02744514439254999, + "grad_norm": 0.33325904607772827, + "learning_rate": 2.7611931436374173e-05, + "loss": 0.0526, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011472654901444912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005592239135876298, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028537549078464508, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8823248088359833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13380, + "weighted_flow_matching_loss": 0.02744514439254999, + "weighted_rkd_loss": 0.025179416686296464 + }, + { + "epoch": 3.895839394821065, + "flow_matching_loss": 0.025937585346400737, + "grad_norm": 0.2911183536052704, + "learning_rate": 2.7607453966968545e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010812304075807333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004774776962585747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02639938537031412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8821560442447662, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13390, + "weighted_flow_matching_loss": 0.025937585346400737, + "weighted_rkd_loss": 0.023288249410688876 + }, + { + "epoch": 3.898748908932208, + "flow_matching_loss": 0.026606130972504617, + "grad_norm": 0.24650049209594727, + "learning_rate": 2.760297266775128e-05, + "loss": 0.0511, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01133974976837635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005122477794066072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027801977656781673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8819871604442596, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13400, + "weighted_flow_matching_loss": 0.026606130972504617, + "weighted_rkd_loss": 0.02452101241797209 + }, + { + "epoch": 3.9016584230433518, + "flow_matching_loss": 0.026829608716070652, + "grad_norm": 0.27516040205955505, + "learning_rate": 2.759848754008368e-05, + "loss": 0.05, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010685746558010578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004890084359794855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026261577568948268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8818181872367858, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13410, + "weighted_flow_matching_loss": 0.026829608716070652, + "weighted_rkd_loss": 0.023157962039113043 + }, + { + "epoch": 3.9045679371544955, + "flow_matching_loss": 0.02648954074829817, + "grad_norm": 0.2834342122077942, + "learning_rate": 2.759399858532821e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011310422420501709, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005182699859142303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027803544886410236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8816490948200226, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13420, + "weighted_flow_matching_loss": 0.02648954074829817, + "weighted_rkd_loss": 0.024513011798262596 + }, + { + "epoch": 3.9074774512656387, + "flow_matching_loss": 0.02557466421276331, + "grad_norm": 0.25082552433013916, + "learning_rate": 2.75895058048485e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011379907187074424, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005104226199910045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027864040434360506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8814799129962921, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13430, + "weighted_flow_matching_loss": 0.02557466421276331, + "weighted_rkd_loss": 0.0245617201551795 + }, + { + "epoch": 3.910386965376782, + "flow_matching_loss": 0.025784014165401457, + "grad_norm": 0.24345818161964417, + "learning_rate": 2.7585009200009326e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.012026670761406422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005338552338071167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029391893930733205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8813106179237366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13440, + "weighted_flow_matching_loss": 0.025784014165401457, + "weighted_rkd_loss": 0.025903512351214887 + }, + { + "epoch": 3.9132964794879257, + "flow_matching_loss": 0.02520954478532076, + "grad_norm": 0.25271257758140564, + "learning_rate": 2.7580508772176648e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01129067074507475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005119637353345752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02770097889006138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.881141209602356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13450, + "weighted_flow_matching_loss": 0.02520954478532076, + "weighted_rkd_loss": 0.024408460967242716 + }, + { + "epoch": 3.916205993599069, + "flow_matching_loss": 0.025905280001461507, + "grad_norm": 0.23015373945236206, + "learning_rate": 2.7576004522717577e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011396454926580191, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005115919606760144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027908829599618913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8809717118740081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13460, + "weighted_flow_matching_loss": 0.025905280001461507, + "weighted_rkd_loss": 0.024586981907486916 + }, + { + "epoch": 3.919115507710212, + "flow_matching_loss": 0.026285044848918915, + "grad_norm": 0.29232650995254517, + "learning_rate": 2.7571496453000377e-05, + "loss": 0.0512, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011456923745572566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005406520143151283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028320367448031902, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8808021068572998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13470, + "weighted_flow_matching_loss": 0.026285044848918915, + "weighted_rkd_loss": 0.024944673106074334 + }, + { + "epoch": 3.922025021821356, + "flow_matching_loss": 0.025993804819881915, + "grad_norm": 0.2580817639827728, + "learning_rate": 2.7566984564394487e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010532837733626366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00471512523945421, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025780800729990005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8806323945522309, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13480, + "weighted_flow_matching_loss": 0.025993804819881915, + "weighted_rkd_loss": 0.022703469917178153 + }, + { + "epoch": 3.924934535932499, + "flow_matching_loss": 0.02667270489037037, + "grad_norm": 0.3246138393878937, + "learning_rate": 2.7562468858270498e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010559430252760649, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004810742940753699, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025929603725671768, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8804625570774078, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13490, + "weighted_flow_matching_loss": 0.02667270489037037, + "weighted_rkd_loss": 0.02283000685274601 + }, + { + "epoch": 3.9278440500436425, + "flow_matching_loss": 0.02748116385191679, + "grad_norm": 0.2473859339952469, + "learning_rate": 2.7557949336000165e-05, + "loss": 0.0529, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01174035482108593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005354415997862816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02883512508124113, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8802926540374756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13500, + "weighted_flow_matching_loss": 0.02748116385191679, + "weighted_rkd_loss": 0.025383511930704115 + }, + { + "epoch": 3.930753564154786, + "flow_matching_loss": 0.025733083300292493, + "grad_norm": 0.2619945704936981, + "learning_rate": 2.7553425998956395e-05, + "loss": 0.049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010869968123733998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004688425688073039, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026428362168371678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8801226258277893, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13510, + "weighted_flow_matching_loss": 0.025733083300292493, + "weighted_rkd_loss": 0.023260247334837912 + }, + { + "epoch": 3.9336630782659294, + "flow_matching_loss": 0.025045795552432536, + "grad_norm": 0.2810313105583191, + "learning_rate": 2.7548898848513264e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010194444097578526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004512765863910318, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024901654198765756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8799524903297424, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13520, + "weighted_flow_matching_loss": 0.025045795552432536, + "weighted_rkd_loss": 0.021912340261042117 + }, + { + "epoch": 3.936572592377073, + "flow_matching_loss": 0.026045184396207333, + "grad_norm": 0.2888442277908325, + "learning_rate": 2.754436788604599e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010375222470611335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004557674890384078, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025308119878172875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8797822654247284, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13530, + "weighted_flow_matching_loss": 0.026045184396207333, + "weighted_rkd_loss": 0.02226560693234205 + }, + { + "epoch": 3.9394821064882164, + "flow_matching_loss": 0.024245224706828595, + "grad_norm": 0.3187105357646942, + "learning_rate": 2.7539833112930973e-05, + "loss": 0.05, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011970165837556123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0053492027800530195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.029289534501731395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8796119332313538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13540, + "weighted_flow_matching_loss": 0.024245224706828595, + "weighted_rkd_loss": 0.025763316079974174 + }, + { + "epoch": 3.94239162059936, + "flow_matching_loss": 0.023990041762590408, + "grad_norm": 0.2853356897830963, + "learning_rate": 2.753529453054575e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010832265485078096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004764628643169999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026429159753024577, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8794414937496186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13550, + "weighted_flow_matching_loss": 0.023990041762590408, + "weighted_rkd_loss": 0.0232429476454854 + }, + { + "epoch": 3.9453011347105034, + "flow_matching_loss": 0.025456681102514266, + "grad_norm": 0.27157384157180786, + "learning_rate": 2.7530752140269028e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0106365573592484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004797770036384463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026070884615182876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8792709350585938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13560, + "weighted_flow_matching_loss": 0.025456681102514266, + "weighted_rkd_loss": 0.022923398204147814 + }, + { + "epoch": 3.9482106488216466, + "flow_matching_loss": 0.026987804099917412, + "grad_norm": 0.3157619833946228, + "learning_rate": 2.7526205943480657e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010741491708904505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048298671608790755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02631285060197115, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8791002929210663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13570, + "weighted_flow_matching_loss": 0.026987804099917412, + "weighted_rkd_loss": 0.023131596110761165 + }, + { + "epoch": 3.9511201629327903, + "flow_matching_loss": 0.02497386857867241, + "grad_norm": 0.25263896584510803, + "learning_rate": 2.7521655941561653e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01085695754736662, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004987777629867196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026701692678034305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8789295375347137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13580, + "weighted_flow_matching_loss": 0.02497386857867241, + "weighted_rkd_loss": 0.023468819819390772 + }, + { + "epoch": 3.9540296770439336, + "flow_matching_loss": 0.02526441402733326, + "grad_norm": 0.24320349097251892, + "learning_rate": 2.751710213589419e-05, + "loss": 0.049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010964266397058964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005099950823932886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027028483524918556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.878758680820465, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13590, + "weighted_flow_matching_loss": 0.02526441402733326, + "weighted_rkd_loss": 0.02375149503350258 + }, + { + "epoch": 3.956939191155077, + "flow_matching_loss": 0.026482907868921757, + "grad_norm": 0.28962045907974243, + "learning_rate": 2.7512544527861597e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010802294313907623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048992080381140115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026503796502947808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8785877227783203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13600, + "weighted_flow_matching_loss": 0.026482907868921757, + "weighted_rkd_loss": 0.023285842314362527 + }, + { + "epoch": 3.9598487052662206, + "flow_matching_loss": 0.025038906000554563, + "grad_norm": 0.2544526755809784, + "learning_rate": 2.7507983118848353e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010217871610075235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004571904265321791, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025007647462189198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8784166574478149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13610, + "weighted_flow_matching_loss": 0.025038906000554563, + "weighted_rkd_loss": 0.021967073529958726 + }, + { + "epoch": 3.962758219377364, + "flow_matching_loss": 0.02663461621850729, + "grad_norm": 0.29080986976623535, + "learning_rate": 2.7503417910240085e-05, + "loss": 0.049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010430309642106294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004579412844032049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025440032221376897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8782454848289489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13620, + "weighted_flow_matching_loss": 0.02663461621850729, + "weighted_rkd_loss": 0.022342541627585887 + }, + { + "epoch": 3.9656677334885075, + "flow_matching_loss": 0.025336898118257522, + "grad_norm": 0.252439945936203, + "learning_rate": 2.7498848903423593e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010565129481256008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004702611872926354, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025832870788872243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8780742168426514, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13630, + "weighted_flow_matching_loss": 0.025336898118257522, + "weighted_rkd_loss": 0.022683191858232022 + }, + { + "epoch": 3.968577247599651, + "flow_matching_loss": 0.02528584636747837, + "grad_norm": 0.2938862144947052, + "learning_rate": 2.7494276099786817e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01031401939690113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004610835830681026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025238874740898608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8779028415679931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13640, + "weighted_flow_matching_loss": 0.02528584636747837, + "weighted_rkd_loss": 0.022157238982617854 + }, + { + "epoch": 3.9714867617107945, + "flow_matching_loss": 0.024278416112065316, + "grad_norm": 0.23732997477054596, + "learning_rate": 2.7489699500718854e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010657562594860791, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048279882408678535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026143113523721694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8777313530445099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13650, + "weighted_flow_matching_loss": 0.024278416112065316, + "weighted_rkd_loss": 0.022946584783494473 + }, + { + "epoch": 3.9743962758219378, + "flow_matching_loss": 0.025296938233077527, + "grad_norm": 0.32873013615608215, + "learning_rate": 2.7485119107609953e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010626099351793528, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00473380929324776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025986008159816266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8775597810745239, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13660, + "weighted_flow_matching_loss": 0.025296938233077527, + "weighted_rkd_loss": 0.022804319113492965 + }, + { + "epoch": 3.977305789933081, + "flow_matching_loss": 0.025995568372309208, + "grad_norm": 0.27453693747520447, + "learning_rate": 2.748053492185152e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010910182911902665, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005233066156506538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027053431794047355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8773880898952484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13670, + "weighted_flow_matching_loss": 0.025995568372309208, + "weighted_rkd_loss": 0.023736338876187802 + }, + { + "epoch": 3.9802153040442247, + "flow_matching_loss": 0.024907124042510987, + "grad_norm": 0.2974410355091095, + "learning_rate": 2.7475946944836103e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010705790482461452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004881242895498872, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02629282381385565, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8772162973880768, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13680, + "weighted_flow_matching_loss": 0.024907124042510987, + "weighted_rkd_loss": 0.02306445064023137 + }, + { + "epoch": 3.983124818155368, + "flow_matching_loss": 0.028122493624687196, + "grad_norm": 0.24483849108219147, + "learning_rate": 2.7471355177957414e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010951258055865765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004883778211660683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026786294393241407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.877044403553009, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13690, + "weighted_flow_matching_loss": 0.028122493624687196, + "weighted_rkd_loss": 0.02349273841828108 + }, + { + "epoch": 3.9860343322665113, + "flow_matching_loss": 0.025647283345460892, + "grad_norm": 0.3021547198295593, + "learning_rate": 2.746675962261031e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011124011687934398, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004956273036077619, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027204296179115774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8768724024295806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13700, + "weighted_flow_matching_loss": 0.025647283345460892, + "weighted_rkd_loss": 0.023854794912040234 + }, + { + "epoch": 3.988943846377655, + "flow_matching_loss": 0.026729082316160203, + "grad_norm": 0.2618725001811981, + "learning_rate": 2.74621602801908e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010555586963891982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004939533304423094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026050707325339317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8767002999782563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13710, + "weighted_flow_matching_loss": 0.026729082316160203, + "weighted_rkd_loss": 0.022838589362800123 + }, + { + "epoch": 3.9918533604887982, + "flow_matching_loss": 0.02442317605018616, + "grad_norm": 0.29877761006355286, + "learning_rate": 2.745755715209604e-05, + "loss": 0.0493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01156714465469122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005203589727170765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028337879478931426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8765281140804291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13720, + "weighted_flow_matching_loss": 0.02442317605018616, + "weighted_rkd_loss": 0.024838989973068236 + }, + { + "epoch": 3.994762874599942, + "flow_matching_loss": 0.025001793913543225, + "grad_norm": 0.236439511179924, + "learning_rate": 2.7452950239724337e-05, + "loss": 0.0479, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010575834475457669, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004985278053209185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026136947236955164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.876355791091919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13730, + "weighted_flow_matching_loss": 0.025001793913543225, + "weighted_rkd_loss": 0.022905278019607067 + }, + { + "epoch": 3.997672388711085, + "flow_matching_loss": 0.025899525545537472, + "grad_norm": 0.22634510695934296, + "learning_rate": 2.7448339544475156e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00964586641639471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004500518599525094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023792251385748385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8761833906173706, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13740, + "weighted_flow_matching_loss": 0.025899525545537472, + "weighted_rkd_loss": 0.020846372470259666 + }, + { + "epoch": 4.000581902822229, + "flow_matching_loss": 0.025850954838097094, + "grad_norm": 0.2577328085899353, + "learning_rate": 2.7443725067749098e-05, + "loss": 0.049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01095092697069049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004522606893442571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026424461044371128, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8760108649730682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13750, + "weighted_flow_matching_loss": 0.025850954838097094, + "weighted_rkd_loss": 0.02314807828515768 + }, + { + "epoch": 4.003491416933372, + "flow_matching_loss": 0.024740167334675787, + "grad_norm": 0.24064131081104279, + "learning_rate": 2.7439106810947923e-05, + "loss": 0.0493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011552809458225965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004957794956862926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02806341405957937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8758382499217987, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13760, + "weighted_flow_matching_loss": 0.024740167334675787, + "weighted_rkd_loss": 0.024578945152461527 + }, + { + "epoch": 4.0064009310445154, + "flow_matching_loss": 0.025714928843081, + "grad_norm": 0.3358056843280792, + "learning_rate": 2.7434484775474532e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0109451356343925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004811665951274335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02670193687081337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8756655275821685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13770, + "weighted_flow_matching_loss": 0.025714928843081, + "weighted_rkd_loss": 0.023381834477186204 + }, + { + "epoch": 4.009310445155659, + "flow_matching_loss": 0.02597453333437443, + "grad_norm": 0.25621163845062256, + "learning_rate": 2.7429858962732984e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009993646945804357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004610219784080982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024597513489425182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8754926919937134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13780, + "weighted_flow_matching_loss": 0.02597453333437443, + "weighted_rkd_loss": 0.0215348644182086 + }, + { + "epoch": 4.012219959266803, + "flow_matching_loss": 0.02670933213084936, + "grad_norm": 0.27030056715011597, + "learning_rate": 2.7425229374128475e-05, + "loss": 0.0513, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011319909058511258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00543067273683846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02807049062103033, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8753197848796844, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13790, + "weighted_flow_matching_loss": 0.02670933213084936, + "weighted_rkd_loss": 0.024570723436772822 + }, + { + "epoch": 4.015129473377946, + "flow_matching_loss": 0.025632675923407077, + "grad_norm": 0.27908310294151306, + "learning_rate": 2.742059601106735e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010733682755380869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004892133269459009, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02635949868708849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8751467525959015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13800, + "weighted_flow_matching_loss": 0.025632675923407077, + "weighted_rkd_loss": 0.023068449832499026 + }, + { + "epoch": 4.018038987489089, + "flow_matching_loss": 0.026392412930727006, + "grad_norm": 0.24978119134902954, + "learning_rate": 2.74159588749571e-05, + "loss": 0.0499, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011064684204757213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004792852234095335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02692221999168396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8749736249446869, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13810, + "weighted_flow_matching_loss": 0.026392412930727006, + "weighted_rkd_loss": 0.02355639822781086 + }, + { + "epoch": 4.020948501600233, + "flow_matching_loss": 0.025651768408715726, + "grad_norm": 0.27135899662971497, + "learning_rate": 2.7411317967206373e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010743553610518574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005049735051579773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026536842808127403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8748003959655761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13820, + "weighted_flow_matching_loss": 0.025651768408715726, + "weighted_rkd_loss": 0.02321435008198023 + }, + { + "epoch": 4.023858015711376, + "flow_matching_loss": 0.027787457779049875, + "grad_norm": 0.267410546541214, + "learning_rate": 2.7406673289224945e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010423673875629902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005085248500108719, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025932595878839493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8746270596981048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13830, + "weighted_flow_matching_loss": 0.027787457779049875, + "weighted_rkd_loss": 0.02268131449818611 + }, + { + "epoch": 4.02676752982252, + "flow_matching_loss": 0.026312771998345853, + "grad_norm": 0.297344446182251, + "learning_rate": 2.740202484242376e-05, + "loss": 0.0492, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010755683202296496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004633847787044942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02614521384239197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8744536221027375, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13840, + "weighted_flow_matching_loss": 0.026312771998345853, + "weighted_rkd_loss": 0.022862821444869043 + }, + { + "epoch": 4.029677043933663, + "flow_matching_loss": 0.025811100006103517, + "grad_norm": 0.2671239674091339, + "learning_rate": 2.7397372628214873e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010650689899921417, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005080541386269033, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02638192120939493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8742800772190094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13850, + "weighted_flow_matching_loss": 0.025811100006103517, + "weighted_rkd_loss": 0.023065163381397726 + }, + { + "epoch": 4.032586558044806, + "flow_matching_loss": 0.02676030918955803, + "grad_norm": 0.27547699213027954, + "learning_rate": 2.739271664801152e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011488580610603094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005413212743587792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0283903744071722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8741064310073853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13860, + "weighted_flow_matching_loss": 0.02676030918955803, + "weighted_rkd_loss": 0.024816170148551465 + }, + { + "epoch": 4.03549607215595, + "flow_matching_loss": 0.02560875527560711, + "grad_norm": 0.27027684450149536, + "learning_rate": 2.738805690322806e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010549677535891533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004606089624576271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02570544444024563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8739326775074006, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13870, + "weighted_flow_matching_loss": 0.02560875527560711, + "weighted_rkd_loss": 0.02246476076543331 + }, + { + "epoch": 4.038405586267094, + "flow_matching_loss": 0.027936185151338576, + "grad_norm": 0.2634887099266052, + "learning_rate": 2.7383393395280004e-05, + "loss": 0.0514, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010966269671916962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004960744734853506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02689328379929066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8737588286399841, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13880, + "weighted_flow_matching_loss": 0.027936185151338576, + "weighted_rkd_loss": 0.023498271591961384 + }, + { + "epoch": 4.041315100378237, + "flow_matching_loss": 0.024311605654656888, + "grad_norm": 0.3250862956047058, + "learning_rate": 2.7378726125584e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011086386162787676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005301223555579782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027473995462059975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8735848665237427, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13890, + "weighted_flow_matching_loss": 0.024311605654656888, + "weighted_rkd_loss": 0.024000858888030052 + }, + { + "epoch": 4.04422461448938, + "flow_matching_loss": 0.022902352921664714, + "grad_norm": 0.27509406208992004, + "learning_rate": 2.7374055095557832e-05, + "loss": 0.0458, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010795464552938939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004596986784599721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026187916100025178, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8734108209609985, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13900, + "weighted_flow_matching_loss": 0.022902352921664714, + "weighted_rkd_loss": 0.022872783616185187 + }, + { + "epoch": 4.047134128600524, + "flow_matching_loss": 0.025949794985353946, + "grad_norm": 0.3032803237438202, + "learning_rate": 2.7369380306620446e-05, + "loss": 0.05, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011156733613461256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005285320850089192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027598788030445576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8732366681098938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13910, + "weighted_flow_matching_loss": 0.025949794985353946, + "weighted_rkd_loss": 0.024100198410451412 + }, + { + "epoch": 4.0500436427116675, + "flow_matching_loss": 0.02839778158813715, + "grad_norm": 0.24457134306430817, + "learning_rate": 2.7364701760191923e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010270848590880632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004770008195191622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025311705470085145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8730624139308929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13920, + "weighted_flow_matching_loss": 0.02839778158813715, + "weighted_rkd_loss": 0.02209879644215107 + }, + { + "epoch": 4.05295315682281, + "flow_matching_loss": 0.028441774658858776, + "grad_norm": 0.29482516646385193, + "learning_rate": 2.7360019457693468e-05, + "loss": 0.0517, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010806981846690178, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005019381223246455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026633344404399396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8728880405426025, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13930, + "weighted_flow_matching_loss": 0.028441774658858776, + "weighted_rkd_loss": 0.023247898928821085 + }, + { + "epoch": 4.055862670933954, + "flow_matching_loss": 0.02475856654345989, + "grad_norm": 0.2513660788536072, + "learning_rate": 2.7355333400547457e-05, + "loss": 0.0485, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011124082189053298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004988854983821511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027237019129097463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.872713589668274, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13940, + "weighted_flow_matching_loss": 0.02475856654345989, + "weighted_rkd_loss": 0.02377011701464653 + }, + { + "epoch": 4.058772185045098, + "flow_matching_loss": 0.027132314629852772, + "grad_norm": 0.27089688181877136, + "learning_rate": 2.7350643590177378e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010454719141125678, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00487023750320077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0257796760648489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8725390136241913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13950, + "weighted_flow_matching_loss": 0.027132314629852772, + "weighted_rkd_loss": 0.02249375209212303 + }, + { + "epoch": 4.0616816991562406, + "flow_matching_loss": 0.024771552346646786, + "grad_norm": 0.2069694697856903, + "learning_rate": 2.7345950028007878e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010609442833811045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048347367905080315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02605362217873335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8723643481731415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13960, + "weighted_flow_matching_loss": 0.024771552346646786, + "weighted_rkd_loss": 0.02272831704467535 + }, + { + "epoch": 4.064591213267384, + "flow_matching_loss": 0.027689249813556673, + "grad_norm": 0.2524145543575287, + "learning_rate": 2.7341252715464736e-05, + "loss": 0.0494, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010269620269536973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004343725484795868, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024882966093719004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8721895754337311, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13970, + "weighted_flow_matching_loss": 0.027689249813556673, + "weighted_rkd_loss": 0.021702705509960653 + }, + { + "epoch": 4.067500727378528, + "flow_matching_loss": 0.02533657643944025, + "grad_norm": 0.2405501753091812, + "learning_rate": 2.7336551653974868e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009751404356211423, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004450774611905217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023953583277761935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8720147013664246, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13980, + "weighted_flow_matching_loss": 0.02533657643944025, + "weighted_rkd_loss": 0.02088791746646166 + }, + { + "epoch": 4.070410241489671, + "flow_matching_loss": 0.025483424961566924, + "grad_norm": 0.2823944389820099, + "learning_rate": 2.733184684496634e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011563833709806204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005365464254282415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028493131697177886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8718397319316864, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 13990, + "weighted_flow_matching_loss": 0.025483424961566924, + "weighted_rkd_loss": 0.024841455928981303 + }, + { + "epoch": 4.0733197556008145, + "flow_matching_loss": 0.02752930000424385, + "grad_norm": 0.2692056894302368, + "learning_rate": 2.7327138289868337e-05, + "loss": 0.049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010009228903800248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004632885497994721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02465134356170893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8716646611690522, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14000, + "weighted_flow_matching_loss": 0.02752930000424385, + "weighted_rkd_loss": 0.021487684547901155 + }, + { + "epoch": 4.076229269711958, + "flow_matching_loss": 0.025286842696368696, + "grad_norm": 0.2641235291957855, + "learning_rate": 2.7322425990111208e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010420664492994548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004984112689271569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025825441628694535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8714894652366638, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14010, + "weighted_flow_matching_loss": 0.025286842696368696, + "weighted_rkd_loss": 0.02250662799924612 + }, + { + "epoch": 4.079138783823102, + "flow_matching_loss": 0.026843438856303692, + "grad_norm": 0.2575127184391022, + "learning_rate": 2.7317709947126416e-05, + "loss": 0.049, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010429683513939381, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004569375538267195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02542874254286289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8713141977787018, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14020, + "weighted_flow_matching_loss": 0.026843438856303692, + "weighted_rkd_loss": 0.022156476974487305 + }, + { + "epoch": 4.082048297934245, + "flow_matching_loss": 0.025663585402071476, + "grad_norm": 0.2802469730377197, + "learning_rate": 2.731299016234657e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010361603088676929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004500507074408233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025223712995648385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8711388170719147, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14030, + "weighted_flow_matching_loss": 0.025663585402071476, + "weighted_rkd_loss": 0.02197334486991167 + }, + { + "epoch": 4.084957812045388, + "flow_matching_loss": 0.027666745893657207, + "grad_norm": 0.26903238892555237, + "learning_rate": 2.730826663720542e-05, + "loss": 0.0516, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01114916978403926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051618922967463735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02746023181825876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8709633350372314, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14040, + "weighted_flow_matching_loss": 0.027666745893657207, + "weighted_rkd_loss": 0.02391683906316757 + }, + { + "epoch": 4.087867326156532, + "flow_matching_loss": 0.0258947366848588, + "grad_norm": 0.27158263325691223, + "learning_rate": 2.730353937313785e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009471770329400896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004140453203581274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023083993792533876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8707877457141876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14050, + "weighted_flow_matching_loss": 0.0258947366848588, + "weighted_rkd_loss": 0.02010123785585165 + }, + { + "epoch": 4.090776840267675, + "flow_matching_loss": 0.025340497493743896, + "grad_norm": 0.26930174231529236, + "learning_rate": 2.729880837157987e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010111880861222744, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004417397966608405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024641159921884537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8706120610237121, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14060, + "weighted_flow_matching_loss": 0.025340497493743896, + "weighted_rkd_loss": 0.021452925726771356 + }, + { + "epoch": 4.093686354378819, + "flow_matching_loss": 0.02311752736568451, + "grad_norm": 0.23091097176074982, + "learning_rate": 2.729407363396864e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009223790792748331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004225900350138545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022673482075333595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.870436280965805, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14070, + "weighted_flow_matching_loss": 0.02311752736568451, + "weighted_rkd_loss": 0.01973587367683649 + }, + { + "epoch": 4.096595868489962, + "flow_matching_loss": 0.024654848873615264, + "grad_norm": 0.2851642966270447, + "learning_rate": 2.7289335161742444e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010937034524977208, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004906858783215284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026780927926301955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.870260375738144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14080, + "weighted_flow_matching_loss": 0.024654848873615264, + "weighted_rkd_loss": 0.02330620400607586 + }, + { + "epoch": 4.099505382601105, + "flow_matching_loss": 0.025647461228072644, + "grad_norm": 0.2862446904182434, + "learning_rate": 2.7284592956340707e-05, + "loss": 0.0485, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010679043736308813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004883503541350365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026241591200232506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8700843870639801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14090, + "weighted_flow_matching_loss": 0.025647461228072644, + "weighted_rkd_loss": 0.02283227350562811 + }, + { + "epoch": 4.102414896712249, + "flow_matching_loss": 0.02814022172242403, + "grad_norm": 0.27202752232551575, + "learning_rate": 2.727984701920399e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010151757206767798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004592357203364372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02489587161689997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8699082970619202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14100, + "weighted_flow_matching_loss": 0.02814022172242403, + "weighted_rkd_loss": 0.0216571481898427 + }, + { + "epoch": 4.105324410823393, + "flow_matching_loss": 0.025097398646175863, + "grad_norm": 0.2748108506202698, + "learning_rate": 2.7275097351773977e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01096166716888547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004796389141120017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02671972382813692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8697321057319641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14110, + "weighted_flow_matching_loss": 0.025097398646175863, + "weighted_rkd_loss": 0.023239029571413994 + }, + { + "epoch": 4.108233924934536, + "flow_matching_loss": 0.024354479648172854, + "grad_norm": 0.2888869643211365, + "learning_rate": 2.727034395549349e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010418562777340412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00474543203599751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02558255735784769, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8695558071136474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14120, + "weighted_flow_matching_loss": 0.024354479648172854, + "weighted_rkd_loss": 0.022245525009930135 + }, + { + "epoch": 4.111143439045679, + "flow_matching_loss": 0.02641559559851885, + "grad_norm": 0.250958651304245, + "learning_rate": 2.726558683180649e-05, + "loss": 0.0507, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011476933863013983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004924405482597649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027878273278474808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8693794012069702, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14130, + "weighted_flow_matching_loss": 0.02641559559851885, + "weighted_rkd_loss": 0.024236787296831607 + }, + { + "epoch": 4.114052953156823, + "flow_matching_loss": 0.025261562131345273, + "grad_norm": 0.21850183606147766, + "learning_rate": 2.7260825982158067e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010202966723591089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004350853920914233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024756787344813347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8692029178142547, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14140, + "weighted_flow_matching_loss": 0.025261562131345273, + "weighted_rkd_loss": 0.021518774889409543 + }, + { + "epoch": 4.1169624672679666, + "flow_matching_loss": 0.025677351281046867, + "grad_norm": 0.2765999436378479, + "learning_rate": 2.725606140799444e-05, + "loss": 0.0487, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010839520208537578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004792350716888905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026471390947699548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8690263092517853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14150, + "weighted_flow_matching_loss": 0.025677351281046867, + "weighted_rkd_loss": 0.02300438769161701 + }, + { + "epoch": 4.119871981379109, + "flow_matching_loss": 0.025745120830833912, + "grad_norm": 0.2971050441265106, + "learning_rate": 2.7251293110762957e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010867115575820208, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004971412499435246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02670564390718937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8688496053218842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14160, + "weighted_flow_matching_loss": 0.025745120830833912, + "weighted_rkd_loss": 0.023203126341104507 + }, + { + "epoch": 4.122781495490253, + "flow_matching_loss": 0.02528984975069761, + "grad_norm": 0.26150748133659363, + "learning_rate": 2.7246521091912106e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011034912429749966, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004956094711087644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027025920152664185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8686728060245514, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14170, + "weighted_flow_matching_loss": 0.02528984975069761, + "weighted_rkd_loss": 0.02347668670117855 + }, + { + "epoch": 4.125691009601397, + "flow_matching_loss": 0.023738156259059905, + "grad_norm": 0.25165247917175293, + "learning_rate": 2.7241745352891497e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010997011419385671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00509354192763567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027087564952671526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8684958994388581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14180, + "weighted_flow_matching_loss": 0.023738156259059905, + "weighted_rkd_loss": 0.023525523580610753 + }, + { + "epoch": 4.12860052371254, + "flow_matching_loss": 0.024718990549445152, + "grad_norm": 0.28778788447380066, + "learning_rate": 2.7236965895151874e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010719309374690057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049155926797539, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02635421194136143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8683188855648041, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14190, + "weighted_flow_matching_loss": 0.024718990549445152, + "weighted_rkd_loss": 0.022883850894868375 + }, + { + "epoch": 4.131510037823683, + "flow_matching_loss": 0.024996065348386765, + "grad_norm": 0.2378811538219452, + "learning_rate": 2.7232182720145105e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009993658307939768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004536792170256376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02452410850673914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8681417942047119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14200, + "weighted_flow_matching_loss": 0.024996065348386765, + "weighted_rkd_loss": 0.02129046358168125 + }, + { + "epoch": 4.134419551934827, + "flow_matching_loss": 0.026071410067379473, + "grad_norm": 0.23575878143310547, + "learning_rate": 2.7227395829324208e-05, + "loss": 0.0497, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011100029526278376, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005022139684297144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027222198992967607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8679645955562592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14210, + "weighted_flow_matching_loss": 0.026071410067379473, + "weighted_rkd_loss": 0.023628062661737205 + }, + { + "epoch": 4.137329066045971, + "flow_matching_loss": 0.02442625965923071, + "grad_norm": 0.24136683344841003, + "learning_rate": 2.7222605224143298e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010330547299236058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004410124383866787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025071219354867936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8677872896194458, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14220, + "weighted_flow_matching_loss": 0.02442625965923071, + "weighted_rkd_loss": 0.02175645958632231 + }, + { + "epoch": 4.1402385801571135, + "flow_matching_loss": 0.02372407503426075, + "grad_norm": 0.22800993919372559, + "learning_rate": 2.721781090605764e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010622241720557212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004752759402617812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025997242704033852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8676098704338073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14230, + "weighted_flow_matching_loss": 0.02372407503426075, + "weighted_rkd_loss": 0.022555428184568883 + }, + { + "epoch": 4.143148094268257, + "flow_matching_loss": 0.0247527327388525, + "grad_norm": 0.29153379797935486, + "learning_rate": 2.721301287652362e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010631481651216745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004564423300325871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02582738660275936, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8674323678016662, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14240, + "weighted_flow_matching_loss": 0.0247527327388525, + "weighted_rkd_loss": 0.022403576783835887 + }, + { + "epoch": 4.146057608379401, + "flow_matching_loss": 0.026750050857663153, + "grad_norm": 0.2754031717777252, + "learning_rate": 2.7208211136998758e-05, + "loss": 0.051, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011228617653250695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005479823984205723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02793705929070711, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8672547578811646, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14250, + "weighted_flow_matching_loss": 0.026750050857663153, + "weighted_rkd_loss": 0.024228520318865777 + }, + { + "epoch": 4.148967122490544, + "flow_matching_loss": 0.025653071328997613, + "grad_norm": 0.25990694761276245, + "learning_rate": 2.7203405688941693e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009630721108987928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045840554405003784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02384549770504236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8670770525932312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14260, + "weighted_flow_matching_loss": 0.025653071328997613, + "weighted_rkd_loss": 0.02067579384893179 + }, + { + "epoch": 4.1518766366016875, + "flow_matching_loss": 0.024339147470891477, + "grad_norm": 0.25041231513023376, + "learning_rate": 2.719859653381219e-05, + "loss": 0.0455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010020273691043257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004398468998260796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02443901654332876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8668992340564727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14270, + "weighted_flow_matching_loss": 0.024339147470891477, + "weighted_rkd_loss": 0.021186201274394988 + }, + { + "epoch": 4.154786150712831, + "flow_matching_loss": 0.02533050049096346, + "grad_norm": 0.28906238079071045, + "learning_rate": 2.719378367307114e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01056278133764863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046800751704722645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025805638171732426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8667213201522828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14280, + "weighted_flow_matching_loss": 0.02533050049096346, + "weighted_rkd_loss": 0.02236628495156765 + }, + { + "epoch": 4.157695664823974, + "flow_matching_loss": 0.025051201693713665, + "grad_norm": 0.30374670028686523, + "learning_rate": 2.7188967108180574e-05, + "loss": 0.0493, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011450810078531503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005100549920462072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028002170100808143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.86654332280159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14290, + "weighted_flow_matching_loss": 0.025051201693713665, + "weighted_rkd_loss": 0.02426513023674488 + }, + { + "epoch": 4.160605178935118, + "flow_matching_loss": 0.025504163652658462, + "grad_norm": 0.27108827233314514, + "learning_rate": 2.7184146840603634e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010004958463832736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004563414514996112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024573331326246263, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8663652062416076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14300, + "weighted_flow_matching_loss": 0.025504163652658462, + "weighted_rkd_loss": 0.021289448253810406 + }, + { + "epoch": 4.163514693046261, + "flow_matching_loss": 0.025215638615190984, + "grad_norm": 0.2563879191875458, + "learning_rate": 2.7179322871804584e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010173705080524087, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004297844390384853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024645254760980607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8661869943141938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14310, + "weighted_flow_matching_loss": 0.025215638615190984, + "weighted_rkd_loss": 0.02134734708815813 + }, + { + "epoch": 4.166424207157405, + "flow_matching_loss": 0.02526633031666279, + "grad_norm": 0.26973241567611694, + "learning_rate": 2.717449520324882e-05, + "loss": 0.0483, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010868704039603471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004896626644767821, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0266340347006917, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8660086810588836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14320, + "weighted_flow_matching_loss": 0.02526633031666279, + "weighted_rkd_loss": 0.023065255768597127 + }, + { + "epoch": 4.169333721268548, + "flow_matching_loss": 0.024185852147638798, + "grad_norm": 0.29746323823928833, + "learning_rate": 2.7169663836402864e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009927714057266713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004279383807443082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02413481194525957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8658302664756775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14330, + "weighted_flow_matching_loss": 0.024185852147638798, + "weighted_rkd_loss": 0.020896603353321552 + }, + { + "epoch": 4.172243235379692, + "flow_matching_loss": 0.023485785163939, + "grad_norm": 0.2992391586303711, + "learning_rate": 2.7164828772734352e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009606562880799174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004156315932050347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02336944192647934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8656517446041108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14340, + "weighted_flow_matching_loss": 0.023485785163939, + "weighted_rkd_loss": 0.020229725074023007 + }, + { + "epoch": 4.175152749490835, + "flow_matching_loss": 0.027729791589081287, + "grad_norm": 0.23607535660266876, + "learning_rate": 2.7159990013712057e-05, + "loss": 0.0491, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010105379531159997, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004529034812003374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0247397942468524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8654731392860413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14350, + "weighted_flow_matching_loss": 0.027729791589081287, + "weighted_rkd_loss": 0.02141168788075447 + }, + { + "epoch": 4.178062263601978, + "flow_matching_loss": 0.026470547541975975, + "grad_norm": 0.27728790044784546, + "learning_rate": 2.7155147560805863e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010074641462415457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004195147729478777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02434443086385727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8652944266796112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14360, + "weighted_flow_matching_loss": 0.026470547541975975, + "weighted_rkd_loss": 0.0210650734603405 + }, + { + "epoch": 4.180971777713122, + "flow_matching_loss": 0.025504324212670328, + "grad_norm": 0.23425708711147308, + "learning_rate": 2.7150301415486777e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009587301686406136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004377889120951295, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023552492447197437, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.865115612745285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14370, + "weighted_flow_matching_loss": 0.025504324212670328, + "weighted_rkd_loss": 0.02037561181932688 + }, + { + "epoch": 4.183881291824266, + "flow_matching_loss": 0.026509815268218516, + "grad_norm": 0.23403200507164001, + "learning_rate": 2.7145451579226932e-05, + "loss": 0.0485, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01035877075046301, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004649687721394002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025367229245603085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8649366915225982, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14380, + "weighted_flow_matching_loss": 0.026509815268218516, + "weighted_rkd_loss": 0.021941115334630012 + }, + { + "epoch": 4.186790805935408, + "flow_matching_loss": 0.025694734789431094, + "grad_norm": 0.2495538592338562, + "learning_rate": 2.7140598053499577e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010298640280961991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004425845900550484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025023126415908335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8647576749324799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14390, + "weighted_flow_matching_loss": 0.025694734789431094, + "weighted_rkd_loss": 0.021638973616063595 + }, + { + "epoch": 4.189700320046552, + "flow_matching_loss": 0.027022233791649343, + "grad_norm": 0.28836339712142944, + "learning_rate": 2.7135740839779087e-05, + "loss": 0.0506, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011044059321284294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005140976142138242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027229094691574574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8645785629749299, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14400, + "weighted_flow_matching_loss": 0.027022233791649343, + "weighted_rkd_loss": 0.02354159504175186 + }, + { + "epoch": 4.192609834157696, + "flow_matching_loss": 0.025321079418063164, + "grad_norm": 0.24145013093948364, + "learning_rate": 2.7130879939540957e-05, + "loss": 0.0455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009576733689755201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00422030845656991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02337377592921257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8643993556499481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14410, + "weighted_flow_matching_loss": 0.025321079418063164, + "weighted_rkd_loss": 0.02020421903580427 + }, + { + "epoch": 4.195519348268839, + "flow_matching_loss": 0.02727307453751564, + "grad_norm": 0.27195361256599426, + "learning_rate": 2.7126015354261798e-05, + "loss": 0.0495, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010567602701485157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004610995203256607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02574620060622692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8642200410366059, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14420, + "weighted_flow_matching_loss": 0.02727307453751564, + "weighted_rkd_loss": 0.022250417061150073 + }, + { + "epoch": 4.198428862379982, + "flow_matching_loss": 0.025427788868546485, + "grad_norm": 0.26348504424095154, + "learning_rate": 2.712114708541934e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010403654351830483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004666510596871376, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025473818928003312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8640406250953674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14430, + "weighted_flow_matching_loss": 0.025427788868546485, + "weighted_rkd_loss": 0.022010352090001108 + }, + { + "epoch": 4.201338376491126, + "flow_matching_loss": 0.02623079065233469, + "grad_norm": 0.22794626653194427, + "learning_rate": 2.7116275134492445e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011056157294660807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004993997304700315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02710631173104048, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8638611137866974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14440, + "weighted_flow_matching_loss": 0.02623079065233469, + "weighted_rkd_loss": 0.023416164331138134 + }, + { + "epoch": 4.20424789060227, + "flow_matching_loss": 0.02791642677038908, + "grad_norm": 0.2064303606748581, + "learning_rate": 2.7111399502961073e-05, + "loss": 0.0501, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010469173779711127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047001820523291825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02563852909952402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8636814892292023, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14450, + "weighted_flow_matching_loss": 0.02791642677038908, + "weighted_rkd_loss": 0.022143658623099326 + }, + { + "epoch": 4.207157404713413, + "flow_matching_loss": 0.024057147838175297, + "grad_norm": 0.2557750940322876, + "learning_rate": 2.710652019230631e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009723426587879658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004171187500469386, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023618040792644023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8635017812252045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14460, + "weighted_flow_matching_loss": 0.024057147838175297, + "weighted_rkd_loss": 0.020394166745245457 + }, + { + "epoch": 4.210066918824556, + "flow_matching_loss": 0.02475877422839403, + "grad_norm": 0.22772645950317383, + "learning_rate": 2.710163720401037e-05, + "loss": 0.0475, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010817740857601166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004763807775452733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026399289816617967, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8633219718933105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14470, + "weighted_flow_matching_loss": 0.02475877422839403, + "weighted_rkd_loss": 0.02279120348393917 + }, + { + "epoch": 4.2129764329357, + "flow_matching_loss": 0.02414097711443901, + "grad_norm": 0.2159639298915863, + "learning_rate": 2.709675053955658e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009571435395628214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042026711395010356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023345541767776014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8631420612335206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14480, + "weighted_flow_matching_loss": 0.02414097711443901, + "weighted_rkd_loss": 0.020150566659867763 + }, + { + "epoch": 4.215885947046843, + "flow_matching_loss": 0.024881643988192082, + "grad_norm": 0.24110464751720428, + "learning_rate": 2.7091860200429362e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009478453360497952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042293128091841934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02318621948361397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8629620432853699, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14490, + "weighted_flow_matching_loss": 0.024881643988192082, + "weighted_rkd_loss": 0.02000885996967554 + }, + { + "epoch": 4.2187954611579865, + "flow_matching_loss": 0.02584895268082619, + "grad_norm": 0.22652918100357056, + "learning_rate": 2.7086966188114285e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00987211843021214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00456796707585454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024312203750014305, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8627819299697876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14500, + "weighted_flow_matching_loss": 0.02584895268082619, + "weighted_rkd_loss": 0.020976217463612556 + }, + { + "epoch": 4.22170497526913, + "flow_matching_loss": 0.02455844935029745, + "grad_norm": 0.2633052468299866, + "learning_rate": 2.708206850409802e-05, + "loss": 0.0458, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010028525069355965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004538865061476827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024595915339887142, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8626017153263092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14510, + "weighted_flow_matching_loss": 0.02455844935029745, + "weighted_rkd_loss": 0.02121655233204365 + }, + { + "epoch": 4.224614489380274, + "flow_matching_loss": 0.02514205239713192, + "grad_norm": 0.29935145378112793, + "learning_rate": 2.707716714986835e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010863775201141835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005207985499873757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026935535855591296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8624214053153991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14520, + "weighted_flow_matching_loss": 0.02514205239713192, + "weighted_rkd_loss": 0.023229799419641494 + }, + { + "epoch": 4.227524003491417, + "flow_matching_loss": 0.027129537984728815, + "grad_norm": 0.3719078302383423, + "learning_rate": 2.707226212691418e-05, + "loss": 0.0503, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010939212329685689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005019083432853222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02689750846475363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8622409999370575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14530, + "weighted_flow_matching_loss": 0.027129537984728815, + "weighted_rkd_loss": 0.023192102462053298 + }, + { + "epoch": 4.2304335176025605, + "flow_matching_loss": 0.027891813032329082, + "grad_norm": 0.28072524070739746, + "learning_rate": 2.7067353436725525e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009293512720614672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004375715623609722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02296274118125439, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8620604932308197, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14540, + "weighted_flow_matching_loss": 0.027891813032329082, + "weighted_rkd_loss": 0.019795213267207144 + }, + { + "epoch": 4.233343031713704, + "flow_matching_loss": 0.024988529831171037, + "grad_norm": 0.2377452254295349, + "learning_rate": 2.706244108079352e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009655251493677497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00434178663417697, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023652289621531965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8618798851966858, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14550, + "weighted_flow_matching_loss": 0.024988529831171037, + "weighted_rkd_loss": 0.02038535652682185 + }, + { + "epoch": 4.236252545824847, + "flow_matching_loss": 0.026244460977613925, + "grad_norm": 0.31719955801963806, + "learning_rate": 2.7057525060610396e-05, + "loss": 0.0488, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010603450890630483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004994754050858319, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02620165590196848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8616991758346557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14560, + "weighted_flow_matching_loss": 0.026244460977613925, + "weighted_rkd_loss": 0.02257788982242346 + }, + { + "epoch": 4.239162059935991, + "flow_matching_loss": 0.02340759839862585, + "grad_norm": 0.25945746898651123, + "learning_rate": 2.7052605377669527e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01110210344195366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005117653822526336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027321860566735266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8615183651447296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14570, + "weighted_flow_matching_loss": 0.02340759839862585, + "weighted_rkd_loss": 0.023538407310843466 + }, + { + "epoch": 4.242071574047134, + "flow_matching_loss": 0.02546585388481617, + "grad_norm": 0.2736358344554901, + "learning_rate": 2.704768203346537e-05, + "loss": 0.0494, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01124484995380044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005253201141022146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02774290107190609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8613374590873718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14580, + "weighted_flow_matching_loss": 0.02546585388481617, + "weighted_rkd_loss": 0.02389603890478611 + }, + { + "epoch": 4.244981088158277, + "flow_matching_loss": 0.02674269340932369, + "grad_norm": 0.2228291928768158, + "learning_rate": 2.7042755029493513e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01004186188802123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004576517548412084, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02466024160385132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8611564576625824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14590, + "weighted_flow_matching_loss": 0.02674269340932369, + "weighted_rkd_loss": 0.021236290223896503 + }, + { + "epoch": 4.247890602269421, + "flow_matching_loss": 0.02596956752240658, + "grad_norm": 0.24316470324993134, + "learning_rate": 2.703782436725065e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0089778040535748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004037579754367471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021993187814950944, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8609753549098969, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14600, + "weighted_flow_matching_loss": 0.02596956752240658, + "weighted_rkd_loss": 0.01893557347357273 + }, + { + "epoch": 4.250800116380565, + "flow_matching_loss": 0.024890468269586564, + "grad_norm": 0.2298266887664795, + "learning_rate": 2.7032890048234585e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009896766440942884, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004427932621911168, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02422146573662758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8607941508293152, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14610, + "weighted_flow_matching_loss": 0.024890468269586564, + "weighted_rkd_loss": 0.020849883556365967 + }, + { + "epoch": 4.2537096304917075, + "flow_matching_loss": 0.025152990780770778, + "grad_norm": 0.2425304502248764, + "learning_rate": 2.702795207394423e-05, + "loss": 0.048, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010838877875357866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004857934429310262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026535690389573575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8606128454208374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14620, + "weighted_flow_matching_loss": 0.025152990780770778, + "weighted_rkd_loss": 0.022837034612894058 + }, + { + "epoch": 4.256619144602851, + "flow_matching_loss": 0.024669033475220203, + "grad_norm": 0.22501808404922485, + "learning_rate": 2.702301044587962e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009852443030104041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00437787405680865, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024082760140299798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8604314446449279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14630, + "weighted_flow_matching_loss": 0.024669033475220203, + "weighted_rkd_loss": 0.020721558853983878 + }, + { + "epoch": 4.259528658713995, + "flow_matching_loss": 0.02594271134585142, + "grad_norm": 0.2835503816604614, + "learning_rate": 2.7018065165541885e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010331794805824756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045994235901162025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025263013131916524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8602499425411224, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14640, + "weighted_flow_matching_loss": 0.02594271134585142, + "weighted_rkd_loss": 0.021732419170439242 + }, + { + "epoch": 4.262438172825139, + "flow_matching_loss": 0.025520492158830167, + "grad_norm": 0.23239398002624512, + "learning_rate": 2.701311623443327e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009564220439642668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004385626502335071, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023514067567884923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8600683510303497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14650, + "weighted_flow_matching_loss": 0.025520492158830167, + "weighted_rkd_loss": 0.020223673526197673 + }, + { + "epoch": 4.265347686936281, + "flow_matching_loss": 0.02513970360159874, + "grad_norm": 0.2688872814178467, + "learning_rate": 2.700816365405713e-05, + "loss": 0.0481, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010888810083270074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004970922926440835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02674854323267937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8598866522312164, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14660, + "weighted_flow_matching_loss": 0.02513970360159874, + "weighted_rkd_loss": 0.023000697419047356 + }, + { + "epoch": 4.268257201047425, + "flow_matching_loss": 0.025027424097061157, + "grad_norm": 0.2803601920604706, + "learning_rate": 2.7003207425917926e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009792139939963817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004339134762994945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023923414386808874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8597048580646515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14670, + "weighted_flow_matching_loss": 0.025027424097061157, + "weighted_rkd_loss": 0.020567152090370654 + }, + { + "epoch": 4.271166715158569, + "flow_matching_loss": 0.026846374198794366, + "grad_norm": 0.25311416387557983, + "learning_rate": 2.6998247551521238e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009665533574298025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004569262638688088, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023900329880416394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.859522956609726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14680, + "weighted_flow_matching_loss": 0.026846374198794366, + "weighted_rkd_loss": 0.02054288350045681 + }, + { + "epoch": 4.274076229269712, + "flow_matching_loss": 0.02643683645874262, + "grad_norm": 0.25242680311203003, + "learning_rate": 2.6993284032373732e-05, + "loss": 0.0469, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0097498026676476, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042824322590604424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023782037757337095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8593409657478333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14690, + "weighted_flow_matching_loss": 0.02643683645874262, + "weighted_rkd_loss": 0.020436860900372267 + }, + { + "epoch": 4.276985743380855, + "flow_matching_loss": 0.026503799483180046, + "grad_norm": 0.23635374009609222, + "learning_rate": 2.6988316869983208e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010508163459599018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005002195667475462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026018522307276727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8591588735580444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14700, + "weighted_flow_matching_loss": 0.026503799483180046, + "weighted_rkd_loss": 0.02235415019094944 + }, + { + "epoch": 4.279895257491999, + "flow_matching_loss": 0.024997349828481674, + "grad_norm": 0.29543164372444153, + "learning_rate": 2.698334606585855e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010303173772990703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004553106520324946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025159453973174095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8589766800403595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14710, + "weighted_flow_matching_loss": 0.024997349828481674, + "weighted_rkd_loss": 0.021611350402235985 + }, + { + "epoch": 4.282804771603142, + "flow_matching_loss": 0.024646823480725288, + "grad_norm": 0.24977122247219086, + "learning_rate": 2.6978371621509757e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01079941214993596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004683476616628468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026282300986349583, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8587943911552429, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14720, + "weighted_flow_matching_loss": 0.024646823480725288, + "weighted_rkd_loss": 0.02257117033004761 + }, + { + "epoch": 4.285714285714286, + "flow_matching_loss": 0.02698616832494736, + "grad_norm": 0.2890014052391052, + "learning_rate": 2.697339353844794e-05, + "loss": 0.0505, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011181220365688204, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004991473653353751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02735391426831484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8586120009422302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14730, + "weighted_flow_matching_loss": 0.02698616832494736, + "weighted_rkd_loss": 0.023486456368118525 + }, + { + "epoch": 4.288623799825429, + "flow_matching_loss": 0.02474036831408739, + "grad_norm": 0.2782863974571228, + "learning_rate": 2.6968411818185303e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009664642252027988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004496889188885689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02382617350667715, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8584295213222504, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14740, + "weighted_flow_matching_loss": 0.02474036831408739, + "weighted_rkd_loss": 0.020453031361103057 + }, + { + "epoch": 4.291533313936573, + "flow_matching_loss": 0.02544203754514456, + "grad_norm": 0.2526555359363556, + "learning_rate": 2.6963426462235167e-05, + "loss": 0.0455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009537468338385225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004288552515208721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02336348928511143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8582469403743744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14750, + "weighted_flow_matching_loss": 0.02544203754514456, + "weighted_rkd_loss": 0.020051633473485708 + }, + { + "epoch": 4.294442828047716, + "flow_matching_loss": 0.027412707172334193, + "grad_norm": 0.27111586928367615, + "learning_rate": 2.6958437472111947e-05, + "loss": 0.0496, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010553666763007641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004730040556751192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025837374106049537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8580642580986023, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14760, + "weighted_flow_matching_loss": 0.027412707172334193, + "weighted_rkd_loss": 0.022170142270624638 + }, + { + "epoch": 4.2973523421588595, + "flow_matching_loss": 0.024017275683581828, + "grad_norm": 0.279776930809021, + "learning_rate": 2.695344484933116e-05, + "loss": 0.0461, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010524378484115004, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004668723628856242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025717480480670928, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8578814744949341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14770, + "weighted_flow_matching_loss": 0.024017275683581828, + "weighted_rkd_loss": 0.022062491811811923 + }, + { + "epoch": 4.300261856270003, + "flow_matching_loss": 0.026203926280140878, + "grad_norm": 0.2724754214286804, + "learning_rate": 2.6948448595409443e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010231771413236857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004659414570778608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02512295711785555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8576985955238342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14780, + "weighted_flow_matching_loss": 0.026203926280140878, + "weighted_rkd_loss": 0.021547933388501406 + }, + { + "epoch": 4.303171370381146, + "flow_matching_loss": 0.02551556918770075, + "grad_norm": 0.26123762130737305, + "learning_rate": 2.694344871186453e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009666518867015838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004245602479204536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023578639887273312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8575156211853028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14790, + "weighted_flow_matching_loss": 0.02551556918770075, + "weighted_rkd_loss": 0.020218990743160248 + }, + { + "epoch": 4.30608088449229, + "flow_matching_loss": 0.02460280880331993, + "grad_norm": 0.20571322739124298, + "learning_rate": 2.6938445200215237e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00981038697063923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004287175135686994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023907949216663837, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8573325514793396, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14800, + "weighted_flow_matching_loss": 0.02460280880331993, + "weighted_rkd_loss": 0.02049711365252733 + }, + { + "epoch": 4.3089903986034335, + "flow_matching_loss": 0.025856344401836394, + "grad_norm": 0.2974574565887451, + "learning_rate": 2.6933438061981513e-05, + "loss": 0.0498, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.011176635697484016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005532580614089966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.027885851636528968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8571493744850158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14810, + "weighted_flow_matching_loss": 0.025856344401836394, + "weighted_rkd_loss": 0.023902191407978536 + }, + { + "epoch": 4.311899912714576, + "flow_matching_loss": 0.025888963229954244, + "grad_norm": 0.31139862537384033, + "learning_rate": 2.6928427298684395e-05, + "loss": 0.0478, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010453988332301378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004671311448328197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025579287856817245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8569661021232605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14820, + "weighted_flow_matching_loss": 0.025888963229954244, + "weighted_rkd_loss": 0.021920528262853622 + }, + { + "epoch": 4.31480942682572, + "flow_matching_loss": 0.024166360311210156, + "grad_norm": 0.21117982268333435, + "learning_rate": 2.6923412911846008e-05, + "loss": 0.0453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010151729825884104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004382687620818615, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02468614708632231, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8567827403545379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14830, + "weighted_flow_matching_loss": 0.024166360311210156, + "weighted_rkd_loss": 0.021150752902030945 + }, + { + "epoch": 4.317718940936864, + "flow_matching_loss": 0.02448077332228422, + "grad_norm": 0.24857567250728607, + "learning_rate": 2.6918394902989604e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01078952318057418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049784148810431365, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026557461358606815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8565992653369904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14840, + "weighted_flow_matching_loss": 0.02448077332228422, + "weighted_rkd_loss": 0.02274914048612118 + }, + { + "epoch": 4.320628455048007, + "flow_matching_loss": 0.026677615754306317, + "grad_norm": 0.26399341225624084, + "learning_rate": 2.6913373273639508e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009889122564345598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004615490953437984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024393736198544503, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8564157068729401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14850, + "weighted_flow_matching_loss": 0.026677615754306317, + "weighted_rkd_loss": 0.020891203172504903 + }, + { + "epoch": 4.32353796915915, + "flow_matching_loss": 0.02459931243211031, + "grad_norm": 0.29560914635658264, + "learning_rate": 2.690834802532117e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00935925398953259, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042661547427997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0229846628382802, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8562320530414581, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14860, + "weighted_flow_matching_loss": 0.02459931243211031, + "weighted_rkd_loss": 0.01968015283346176 + }, + { + "epoch": 4.326447483270294, + "flow_matching_loss": 0.026206566393375395, + "grad_norm": 0.2372213751077652, + "learning_rate": 2.6903319159561125e-05, + "loss": 0.0465, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009663235954940319, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043701023096218705, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023696574196219445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8560482919216156, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14870, + "weighted_flow_matching_loss": 0.026206566393375395, + "weighted_rkd_loss": 0.02028544768691063 + }, + { + "epoch": 4.329356997381438, + "flow_matching_loss": 0.025021732784807682, + "grad_norm": 0.2574147582054138, + "learning_rate": 2.6898286677887005e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010601298976689577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048136070370674135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026016204804182052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8558644354343414, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14880, + "weighted_flow_matching_loss": 0.025021732784807682, + "weighted_rkd_loss": 0.022266454808413983 + }, + { + "epoch": 4.3322665114925805, + "flow_matching_loss": 0.025054767914116382, + "grad_norm": 0.27123090624809265, + "learning_rate": 2.689325058182755e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010595317091792822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004873714013956487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026064348034560682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8556804835796357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14890, + "weighted_flow_matching_loss": 0.025054767914116382, + "weighted_rkd_loss": 0.022302708961069583 + }, + { + "epoch": 4.335176025603724, + "flow_matching_loss": 0.025597444362938405, + "grad_norm": 0.2542797327041626, + "learning_rate": 2.6888210872912594e-05, + "loss": 0.0461, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009778433479368687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004392236773855984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02394910380244255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8554964244365693, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14900, + "weighted_flow_matching_loss": 0.025597444362938405, + "weighted_rkd_loss": 0.020488414727151395 + }, + { + "epoch": 4.338085539714868, + "flow_matching_loss": 0.025493309274315833, + "grad_norm": 0.2544277608394623, + "learning_rate": 2.688316755267306e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010341723263263703, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004498053551651537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02518150005489588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8553122580051422, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14910, + "weighted_flow_matching_loss": 0.025493309274315833, + "weighted_rkd_loss": 0.021538094244897367 + }, + { + "epoch": 4.340995053826011, + "flow_matching_loss": 0.02538454569876194, + "grad_norm": 0.233974888920784, + "learning_rate": 2.6878120622640988e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00898410426452756, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040299098938703535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021998118609189987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.855128037929535, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14920, + "weighted_flow_matching_loss": 0.02538454569876194, + "weighted_rkd_loss": 0.018811235763132574 + }, + { + "epoch": 4.343904567937154, + "flow_matching_loss": 0.024801870435476304, + "grad_norm": 0.22500917315483093, + "learning_rate": 2.6873070084349486e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009400968020781875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004211828741244972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023013764619827272, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8549436926841736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14930, + "weighted_flow_matching_loss": 0.024801870435476304, + "weighted_rkd_loss": 0.019675440900027753 + }, + { + "epoch": 4.346814082048298, + "flow_matching_loss": 0.025640005245804787, + "grad_norm": 0.25627055764198303, + "learning_rate": 2.6868015939332788e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010324915405362844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0048533720895648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02550320327281952, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8547592461109161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14940, + "weighted_flow_matching_loss": 0.025640005245804787, + "weighted_rkd_loss": 0.021799160353839397 + }, + { + "epoch": 4.349723596159441, + "flow_matching_loss": 0.025099471025168896, + "grad_norm": 0.31156471371650696, + "learning_rate": 2.68629581891262e-05, + "loss": 0.0467, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010408227797597647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004460749612189829, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02527720481157303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8545747101306915, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14950, + "weighted_flow_matching_loss": 0.025099471025168896, + "weighted_rkd_loss": 0.02160110156983137 + }, + { + "epoch": 4.352633110270585, + "flow_matching_loss": 0.02382022012025118, + "grad_norm": 0.23170167207717896, + "learning_rate": 2.685789683526614e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009820449072867632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004318964970298112, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023959863185882568, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8543900668621063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14960, + "weighted_flow_matching_loss": 0.02382022012025118, + "weighted_rkd_loss": 0.0204709867015481 + }, + { + "epoch": 4.355542624381728, + "flow_matching_loss": 0.024414898827672005, + "grad_norm": 0.2058415561914444, + "learning_rate": 2.685283187929011e-05, + "loss": 0.0465, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010434505715966224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0049308400833979246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025799851678311825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8542053580284119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14970, + "weighted_flow_matching_loss": 0.024414898827672005, + "weighted_rkd_loss": 0.022038483805954457 + }, + { + "epoch": 4.358452138492872, + "flow_matching_loss": 0.02693708408623934, + "grad_norm": 0.25274693965911865, + "learning_rate": 2.6847763322736713e-05, + "loss": 0.0466, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00930292229168117, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004471618053503335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023077462799847126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8540205121040344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14980, + "weighted_flow_matching_loss": 0.02693708408623934, + "weighted_rkd_loss": 0.019708677567541598 + }, + { + "epoch": 4.361361652604015, + "flow_matching_loss": 0.024173790775239467, + "grad_norm": 0.26941314339637756, + "learning_rate": 2.684269116714564e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010256431112065912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004681185446679592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025194047950208187, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8538355827331543, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 14990, + "weighted_flow_matching_loss": 0.024173790775239467, + "weighted_rkd_loss": 0.021511607989668845 + }, + { + "epoch": 4.364271166715159, + "flow_matching_loss": 0.02385709621012211, + "grad_norm": 0.2335941344499588, + "learning_rate": 2.6837615414057677e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009323449153453112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004079484031535685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022726382315158843, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8536505699157715, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15000, + "weighted_flow_matching_loss": 0.02385709621012211, + "weighted_rkd_loss": 0.01940036043524742 + }, + { + "epoch": 4.367180680826302, + "flow_matching_loss": 0.0243483429774642, + "grad_norm": 0.2780650854110718, + "learning_rate": 2.6832536065014708e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0099476114846766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004332356480881571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024227579496800898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8534654378890991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15010, + "weighted_flow_matching_loss": 0.0243483429774642, + "weighted_rkd_loss": 0.020677411928772926 + }, + { + "epoch": 4.370090194937445, + "flow_matching_loss": 0.02696578297764063, + "grad_norm": 0.2688899636268616, + "learning_rate": 2.68274531215597e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00957334809936583, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004130183812230825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02327687982469797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.853280222415924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15020, + "weighted_flow_matching_loss": 0.02696578297764063, + "weighted_rkd_loss": 0.019861760642379524 + }, + { + "epoch": 4.372999709048589, + "flow_matching_loss": 0.026084090769290923, + "grad_norm": 0.2261257767677307, + "learning_rate": 2.6822366585236716e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010123248491436244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004643817688338459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024890314415097238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8530949056148529, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15030, + "weighted_flow_matching_loss": 0.026084090769290923, + "weighted_rkd_loss": 0.02123379912227392 + }, + { + "epoch": 4.3759092231597325, + "flow_matching_loss": 0.0259066266939044, + "grad_norm": 0.22872626781463623, + "learning_rate": 2.6817276457590924e-05, + "loss": 0.0482, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010713900718837977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004689396568574011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026117197796702386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.852909505367279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15040, + "weighted_flow_matching_loss": 0.0259066266939044, + "weighted_rkd_loss": 0.022275580465793608 + }, + { + "epoch": 4.378818737270876, + "flow_matching_loss": 0.025034806132316588, + "grad_norm": 0.18406201899051666, + "learning_rate": 2.6812182740168555e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009704860066995025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044354560086503625, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02384517639875412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8527239918708801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15050, + "weighted_flow_matching_loss": 0.025034806132316588, + "weighted_rkd_loss": 0.020333396922796964 + }, + { + "epoch": 4.381728251382019, + "flow_matching_loss": 0.023547392152249814, + "grad_norm": 0.24820353090763092, + "learning_rate": 2.6807085434516953e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01075086034834385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004733058111742139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026234778575599194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8525383770465851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15060, + "weighted_flow_matching_loss": 0.023547392152249814, + "weighted_rkd_loss": 0.02236617747694254 + }, + { + "epoch": 4.384637765493163, + "flow_matching_loss": 0.024386686086654664, + "grad_norm": 0.2713903784751892, + "learning_rate": 2.6801984542184544e-05, + "loss": 0.0489, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01170515948906541, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005389065481722355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.028799384459853174, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8523526847362518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15070, + "weighted_flow_matching_loss": 0.024386686086654664, + "weighted_rkd_loss": 0.02454738859087229 + }, + { + "epoch": 4.3875472796043065, + "flow_matching_loss": 0.0244203669950366, + "grad_norm": 0.25769728422164917, + "learning_rate": 2.6796880064720845e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00960621596314013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004410347575321794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02362277954816818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.852166885137558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15080, + "weighted_flow_matching_loss": 0.0244203669950366, + "weighted_rkd_loss": 0.020130466017872096 + }, + { + "epoch": 4.390456793715449, + "flow_matching_loss": 0.025821737200021743, + "grad_norm": 0.27022549510002136, + "learning_rate": 2.6791772003676462e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009766985941678286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00413627706002444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023670248873531818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.851980984210968, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15090, + "weighted_flow_matching_loss": 0.025821737200021743, + "weighted_rkd_loss": 0.020166642777621747 + }, + { + "epoch": 4.393366307826593, + "flow_matching_loss": 0.024201085045933723, + "grad_norm": 0.2731785774230957, + "learning_rate": 2.6786660360603087e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010975435003638267, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045656114583835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0265164814889431, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8517949998378753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15100, + "weighted_flow_matching_loss": 0.024201085045933723, + "weighted_rkd_loss": 0.02258658464998007 + }, + { + "epoch": 4.396275821937737, + "flow_matching_loss": 0.02483952511101961, + "grad_norm": 0.23080173134803772, + "learning_rate": 2.6781545137053503e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009673198126256466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004330109036527574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023676505498588084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8516089141368866, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15110, + "weighted_flow_matching_loss": 0.02483952511101961, + "weighted_rkd_loss": 0.02016314547508955 + }, + { + "epoch": 4.3991853360488795, + "flow_matching_loss": 0.02430947348475456, + "grad_norm": 0.23726914823055267, + "learning_rate": 2.677642633458158e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009099653549492359, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004002879490144551, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022202186286449432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8514227271080017, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15120, + "weighted_flow_matching_loss": 0.02430947348475456, + "weighted_rkd_loss": 0.018903546035289764 + }, + { + "epoch": 4.402094850160023, + "flow_matching_loss": 0.024970916099846363, + "grad_norm": 0.201747864484787, + "learning_rate": 2.6771303954742274e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008990617841482163, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004189008311368525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022170243971049785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8512364506721497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15130, + "weighted_flow_matching_loss": 0.024970916099846363, + "weighted_rkd_loss": 0.01887216866016388 + }, + { + "epoch": 4.405004364271167, + "flow_matching_loss": 0.027041755244135858, + "grad_norm": 0.21550238132476807, + "learning_rate": 2.6766177999091633e-05, + "loss": 0.0471, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009560782788321375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004414721764624119, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023536287527531386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8510500729084015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15140, + "weighted_flow_matching_loss": 0.027041755244135858, + "weighted_rkd_loss": 0.020030581671744586 + }, + { + "epoch": 4.40791387838231, + "flow_matching_loss": 0.023690756782889365, + "grad_norm": 0.2337247133255005, + "learning_rate": 2.676104846918678e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009203439950942994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004155026166699827, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022561906091868877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8508635997772217, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15150, + "weighted_flow_matching_loss": 0.023690756782889365, + "weighted_rkd_loss": 0.019197175465524196 + }, + { + "epoch": 4.4108233924934535, + "flow_matching_loss": 0.02458545994013548, + "grad_norm": 0.23553313314914703, + "learning_rate": 2.675591536658594e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009547342639416456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004389579547569156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023484264872968195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8506770312786103, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15160, + "weighted_flow_matching_loss": 0.02458545994013548, + "weighted_rkd_loss": 0.01997756538912654 + }, + { + "epoch": 4.413732906604597, + "flow_matching_loss": 0.02537373695522547, + "grad_norm": 0.22483685612678528, + "learning_rate": 2.6750778692848405e-05, + "loss": 0.0461, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010016647400334477, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043631632346659895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024396458081901072, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8504903614521027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15170, + "weighted_flow_matching_loss": 0.02537373695522547, + "weighted_rkd_loss": 0.02074907412752509 + }, + { + "epoch": 4.416642420715741, + "flow_matching_loss": 0.026055147312581538, + "grad_norm": 0.20653411746025085, + "learning_rate": 2.6745638449534562e-05, + "loss": 0.0458, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00948283369652927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042087466455996035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023174413852393626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8503036022186279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15180, + "weighted_flow_matching_loss": 0.026055147312581538, + "weighted_rkd_loss": 0.019705392327159642 + }, + { + "epoch": 4.419551934826884, + "flow_matching_loss": 0.02308435309678316, + "grad_norm": 0.2964051067829132, + "learning_rate": 2.6740494638205888e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009568980801850558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004302512877620757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023440474644303322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8501167476177216, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15190, + "weighted_flow_matching_loss": 0.02308435309678316, + "weighted_rkd_loss": 0.019927050173282623 + }, + { + "epoch": 4.422461448938027, + "flow_matching_loss": 0.023992104083299638, + "grad_norm": 0.290096640586853, + "learning_rate": 2.6735347260424928e-05, + "loss": 0.0453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010278247576206923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004568683239631355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025125178322196007, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.849929791688919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15200, + "weighted_flow_matching_loss": 0.023992104083299638, + "weighted_rkd_loss": 0.021354597993195057 + }, + { + "epoch": 4.425370963049171, + "flow_matching_loss": 0.024765956588089465, + "grad_norm": 0.28152358531951904, + "learning_rate": 2.6730196317755326e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01013447493314743, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004370104684494436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02463905457407236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8497427463531494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15210, + "weighted_flow_matching_loss": 0.024765956588089465, + "weighted_rkd_loss": 0.020936904940754175 + }, + { + "epoch": 4.428280477160314, + "flow_matching_loss": 0.021643693372607232, + "grad_norm": 0.27810904383659363, + "learning_rate": 2.6725041811761805e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01003363197669387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00427801648620516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024345280416309834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8495556056499481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15220, + "weighted_flow_matching_loss": 0.021643693372607232, + "weighted_rkd_loss": 0.020682658068835735 + }, + { + "epoch": 4.431189991271458, + "flow_matching_loss": 0.024485724233090877, + "grad_norm": 0.2479282170534134, + "learning_rate": 2.671988374401016e-05, + "loss": 0.0468, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010847081895917654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004617371154017747, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026311535201966763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8493683576583863, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15230, + "weighted_flow_matching_loss": 0.024485724233090877, + "weighted_rkd_loss": 0.022348236851394178 + }, + { + "epoch": 4.434099505382601, + "flow_matching_loss": 0.02529911044985056, + "grad_norm": 0.2777361571788788, + "learning_rate": 2.671472211606728e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010025826469063758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004206458199769258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024258110858500003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8491810202598572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15240, + "weighted_flow_matching_loss": 0.02529911044985056, + "weighted_rkd_loss": 0.0205995699390769 + }, + { + "epoch": 4.437009019493744, + "flow_matching_loss": 0.02457043472677469, + "grad_norm": 0.2587687075138092, + "learning_rate": 2.6709556929501128e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009816183149814606, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004257483268156647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023889849521219732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8489935874938965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15250, + "weighted_flow_matching_loss": 0.02457043472677469, + "weighted_rkd_loss": 0.02028233651071787 + }, + { + "epoch": 4.439918533604888, + "flow_matching_loss": 0.02361346799880266, + "grad_norm": 0.27120593190193176, + "learning_rate": 2.670438818588076e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009681534161791206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004350367770530284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023713436163961887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8488060593605041, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15260, + "weighted_flow_matching_loss": 0.02361346799880266, + "weighted_rkd_loss": 0.02012808220461011 + }, + { + "epoch": 4.442828047716032, + "flow_matching_loss": 0.025960027426481246, + "grad_norm": 0.2235565036535263, + "learning_rate": 2.6699215886776287e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010213722288608552, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00475706069264561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025184504874050617, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8486184418201447, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15270, + "weighted_flow_matching_loss": 0.025960027426481246, + "weighted_rkd_loss": 0.021372009813785554 + }, + { + "epoch": 4.445737561827175, + "flow_matching_loss": 0.023508013039827347, + "grad_norm": 0.2756088972091675, + "learning_rate": 2.6694040033758933e-05, + "loss": 0.0447, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01016017897054553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046820695977658035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025002427212893964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.848430722951889, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15280, + "weighted_flow_matching_loss": 0.023508013039827347, + "weighted_rkd_loss": 0.021212864294648172 + }, + { + "epoch": 4.448647075938318, + "flow_matching_loss": 0.023216148652136327, + "grad_norm": 0.25014638900756836, + "learning_rate": 2.6688860628400982e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010528128687292337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004655555891804397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025711813010275364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8482428967952729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15290, + "weighted_flow_matching_loss": 0.023216148652136327, + "weighted_rkd_loss": 0.02180993165820837 + }, + { + "epoch": 4.451556590049462, + "flow_matching_loss": 0.023235779814422132, + "grad_norm": 0.2924257516860962, + "learning_rate": 2.6683677672275797e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010001970175653695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004552643257193267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024556583538651467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8480549931526185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15300, + "weighted_flow_matching_loss": 0.023235779814422132, + "weighted_rkd_loss": 0.020825224369764327 + }, + { + "epoch": 4.4544661041606055, + "flow_matching_loss": 0.02540802601724863, + "grad_norm": 0.2709493935108185, + "learning_rate": 2.6678491166957825e-05, + "loss": 0.0464, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010064288973808289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046111676841974255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02473974581807852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8478669822216034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15310, + "weighted_flow_matching_loss": 0.02540802601724863, + "weighted_rkd_loss": 0.02097598500549793 + }, + { + "epoch": 4.457375618271748, + "flow_matching_loss": 0.02403584439307451, + "grad_norm": 0.2354045957326889, + "learning_rate": 2.6673301114022592e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010393942054361106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004613741487264633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025401625595986844, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8476788878440857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15320, + "weighted_flow_matching_loss": 0.02403584439307451, + "weighted_rkd_loss": 0.021532442793250083 + }, + { + "epoch": 4.460285132382892, + "flow_matching_loss": 0.02574784550815821, + "grad_norm": 0.26043152809143066, + "learning_rate": 2.666810751504669e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010487984213978053, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004823677428066731, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025799645483493804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8474906861782074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15330, + "weighted_flow_matching_loss": 0.02574784550815821, + "weighted_rkd_loss": 0.021864886954426767 + }, + { + "epoch": 4.463194646494036, + "flow_matching_loss": 0.025694814324378968, + "grad_norm": 0.25240135192871094, + "learning_rate": 2.6662910371607807e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009340183809399605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042839697794988755, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02296433728188276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8473023772239685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15340, + "weighted_flow_matching_loss": 0.025694814324378968, + "weighted_rkd_loss": 0.01945775095373392 + }, + { + "epoch": 4.466104160605179, + "flow_matching_loss": 0.024625479616224764, + "grad_norm": 0.2376009225845337, + "learning_rate": 2.6657709685284696e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010250930115580559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004579676920548082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0250815374776721, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8471140086650848, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15350, + "weighted_flow_matching_loss": 0.024625479616224764, + "weighted_rkd_loss": 0.021246871910989285 + }, + { + "epoch": 4.469013674716322, + "flow_matching_loss": 0.024740970693528652, + "grad_norm": 0.27553558349609375, + "learning_rate": 2.6652505457657183e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009548826655372978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004323214967735111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02342086862772703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8469255268573761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15360, + "weighted_flow_matching_loss": 0.024740970693528652, + "weighted_rkd_loss": 0.01983562931418419 + }, + { + "epoch": 4.471923188827466, + "flow_matching_loss": 0.024012264981865882, + "grad_norm": 0.2657083570957184, + "learning_rate": 2.664729769030618e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010314088501036166, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004366230010055006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024994407221674918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8467369556427002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15370, + "weighted_flow_matching_loss": 0.024012264981865882, + "weighted_rkd_loss": 0.02116364147514105 + }, + { + "epoch": 4.47483270293861, + "flow_matching_loss": 0.02548341955989599, + "grad_norm": 0.2766512334346771, + "learning_rate": 2.6642086384813667e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009769589733332395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004609669605270028, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024148849211633204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8465482711791992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15380, + "weighted_flow_matching_loss": 0.02548341955989599, + "weighted_rkd_loss": 0.020443179458379746 + }, + { + "epoch": 4.4777422170497525, + "flow_matching_loss": 0.022160044685006142, + "grad_norm": 0.251535564661026, + "learning_rate": 2.6636871542762703e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009740084502846002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045918476767838005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024072016403079034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8463595151901245, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15390, + "weighted_flow_matching_loss": 0.022160044685006142, + "weighted_rkd_loss": 0.020373547077178956 + }, + { + "epoch": 4.480651731160896, + "flow_matching_loss": 0.024755993112921714, + "grad_norm": 0.30717340111732483, + "learning_rate": 2.6631653165737418e-05, + "loss": 0.0463, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010341536905616521, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047314786585047845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025414552725851535, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8461706340312958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15400, + "weighted_flow_matching_loss": 0.024755993112921714, + "weighted_rkd_loss": 0.021505006402730942 + }, + { + "epoch": 4.48356124527204, + "flow_matching_loss": 0.02497327942401171, + "grad_norm": 0.267496258020401, + "learning_rate": 2.662643125532302e-05, + "loss": 0.0477, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010867109149694442, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0051743107382208105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026908529177308082, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8459816873073578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15410, + "weighted_flow_matching_loss": 0.02497327942401171, + "weighted_rkd_loss": 0.02276406604796648 + }, + { + "epoch": 4.486470759383183, + "flow_matching_loss": 0.026571796834468843, + "grad_norm": 0.23226036131381989, + "learning_rate": 2.6621205813105777e-05, + "loss": 0.0476, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010036963317543269, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004795610788278282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02486953753978014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8457926332950592, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15420, + "weighted_flow_matching_loss": 0.026571796834468843, + "weighted_rkd_loss": 0.0210345484316349 + }, + { + "epoch": 4.4893802734943264, + "flow_matching_loss": 0.023325778543949127, + "grad_norm": 0.252435564994812, + "learning_rate": 2.6615976840673056e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010132992453873158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004520499613136053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024786484614014625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8456034898757935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15430, + "weighted_flow_matching_loss": 0.023325778543949127, + "weighted_rkd_loss": 0.02095952443778515 + }, + { + "epoch": 4.49228978760547, + "flow_matching_loss": 0.02404319904744625, + "grad_norm": 0.38508644700050354, + "learning_rate": 2.6610744339613265e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009752496657893062, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004577459604479373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02408245261758566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8454142451286316, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15440, + "weighted_flow_matching_loss": 0.02404319904744625, + "weighted_rkd_loss": 0.020359564386308194 + }, + { + "epoch": 4.495199301716613, + "flow_matching_loss": 0.02568835634738207, + "grad_norm": 0.23233291506767273, + "learning_rate": 2.6605508311515916e-05, + "loss": 0.0484, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010996212903410196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00489633260294795, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.026888758689165116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.845224916934967, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15450, + "weighted_flow_matching_loss": 0.02568835634738207, + "weighted_rkd_loss": 0.02272706590592861 + }, + { + "epoch": 4.498108815827757, + "flow_matching_loss": 0.023485148325562477, + "grad_norm": 0.33467286825180054, + "learning_rate": 2.6600268757971566e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010271165054291487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004663410200737417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025205740332603456, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8450354814529419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15460, + "weighted_flow_matching_loss": 0.023485148325562477, + "weighted_rkd_loss": 0.021299562230706216 + }, + { + "epoch": 4.5010183299389, + "flow_matching_loss": 0.02467732448130846, + "grad_norm": 0.26512786746025085, + "learning_rate": 2.659502568057185e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010148674529045821, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004626134014688432, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024923483096063136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8448459506034851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15470, + "weighted_flow_matching_loss": 0.02467732448130846, + "weighted_rkd_loss": 0.021056528203189373 + }, + { + "epoch": 4.503927844050043, + "flow_matching_loss": 0.025651215575635433, + "grad_norm": 0.25713104009628296, + "learning_rate": 2.658977908090949e-05, + "loss": 0.047, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010214756382629276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047897602897137405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025219272635877133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8446563422679901, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15480, + "weighted_flow_matching_loss": 0.025651215575635433, + "weighted_rkd_loss": 0.021301756706088783 + }, + { + "epoch": 4.506837358161187, + "flow_matching_loss": 0.024610585719347, + "grad_norm": 0.2144567370414734, + "learning_rate": 2.6584528960578245e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010472826194018126, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046641633845865725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02560981586575508, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8444666206836701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15490, + "weighted_flow_matching_loss": 0.024610585719347, + "weighted_rkd_loss": 0.021626669354736804 + }, + { + "epoch": 4.509746872272331, + "flow_matching_loss": 0.02535865269601345, + "grad_norm": 0.26016765832901, + "learning_rate": 2.657927532117298e-05, + "loss": 0.0455, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009615204762667418, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004609257192350924, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023839666694402694, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8442768156528473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15500, + "weighted_flow_matching_loss": 0.02535865269601345, + "weighted_rkd_loss": 0.020127279125154017 + }, + { + "epoch": 4.512656386383474, + "flow_matching_loss": 0.024383540637791155, + "grad_norm": 0.22458888590335846, + "learning_rate": 2.6574018164289605e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01006589513272047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004811026738025248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024942816980183125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.844086903333664, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15510, + "weighted_flow_matching_loss": 0.024383540637791155, + "weighted_rkd_loss": 0.02105395756661892 + }, + { + "epoch": 4.515565900494617, + "flow_matching_loss": 0.02468877099454403, + "grad_norm": 0.3081466853618622, + "learning_rate": 2.6568757491525103e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009430853091180325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004170129401609301, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02303183563053608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8438969135284424, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15520, + "weighted_flow_matching_loss": 0.02468877099454403, + "weighted_rkd_loss": 0.01943650459870696 + }, + { + "epoch": 4.518475414605761, + "flow_matching_loss": 0.0250669265165925, + "grad_norm": 0.2524046003818512, + "learning_rate": 2.6563493304477534e-05, + "loss": 0.0462, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01015744125470519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0047650607768446205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025079943239688873, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8437068164348602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15530, + "weighted_flow_matching_loss": 0.0250669265165925, + "weighted_rkd_loss": 0.021160320192575455 + }, + { + "epoch": 4.521384928716905, + "flow_matching_loss": 0.024086585827171804, + "grad_norm": 0.26447293162345886, + "learning_rate": 2.655822560474601e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00932056768797338, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004205753491260111, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022846888937056064, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8435166358947754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15540, + "weighted_flow_matching_loss": 0.024086585827171804, + "weighted_rkd_loss": 0.019271801225841045 + }, + { + "epoch": 4.524294442828047, + "flow_matching_loss": 0.023899142630398273, + "grad_norm": 0.3126222491264343, + "learning_rate": 2.655295439393073e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009786828747019171, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004591439664363861, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024165097065269946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.84332634806633, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15550, + "weighted_flow_matching_loss": 0.023899142630398273, + "weighted_rkd_loss": 0.020379038527607918 + }, + { + "epoch": 4.527203956939191, + "flow_matching_loss": 0.025447890907526017, + "grad_norm": 0.2214842140674591, + "learning_rate": 2.654767967363294e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009500386379659175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004122923477552831, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023123696446418762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8431359827518463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15560, + "weighted_flow_matching_loss": 0.025447890907526017, + "weighted_rkd_loss": 0.019496432039886713 + }, + { + "epoch": 4.530113471050335, + "flow_matching_loss": 0.024639133363962173, + "grad_norm": 0.25200945138931274, + "learning_rate": 2.6542401445454967e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010203844122588634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004414201667532325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02482188995927572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8429455041885376, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15570, + "weighted_flow_matching_loss": 0.024639133363962173, + "weighted_rkd_loss": 0.020923558343201875 + }, + { + "epoch": 4.5330229851614785, + "flow_matching_loss": 0.024625243805348872, + "grad_norm": 0.2542215585708618, + "learning_rate": 2.6537119711000187e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009724253602325917, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042754263849928975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023723933659493924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8427549481391907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15580, + "weighted_flow_matching_loss": 0.024625243805348872, + "weighted_rkd_loss": 0.019993468280881645 + }, + { + "epoch": 4.535932499272621, + "flow_matching_loss": 0.02227950431406498, + "grad_norm": 0.24073176085948944, + "learning_rate": 2.6531834471873068e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009303928352892399, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004011719906702638, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022619576752185823, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8425642848014832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15590, + "weighted_flow_matching_loss": 0.02227950431406498, + "weighted_rkd_loss": 0.019058451056480408 + }, + { + "epoch": 4.538842013383765, + "flow_matching_loss": 0.024366100318729877, + "grad_norm": 0.25690463185310364, + "learning_rate": 2.6526545729679115e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00965876542031765, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00445370334200561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023771234229207037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.842373538017273, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15600, + "weighted_flow_matching_loss": 0.024366100318729877, + "weighted_rkd_loss": 0.020024308934807777 + }, + { + "epoch": 4.541751527494909, + "flow_matching_loss": 0.02450087685137987, + "grad_norm": 0.2535628080368042, + "learning_rate": 2.6521253486024908e-05, + "loss": 0.0443, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009671843331307172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004208888276480138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023552574776113035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8421826958656311, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15610, + "weighted_flow_matching_loss": 0.02450087685137987, + "weighted_rkd_loss": 0.01983544435352087 + }, + { + "epoch": 4.5446610416060516, + "flow_matching_loss": 0.02340175323188305, + "grad_norm": 0.27531319856643677, + "learning_rate": 2.651595774251809e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009772313060238957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004381315596401691, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023925941437482834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8419917583465576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15620, + "weighted_flow_matching_loss": 0.02340175323188305, + "weighted_rkd_loss": 0.020145390741527082 + }, + { + "epoch": 4.547570555717195, + "flow_matching_loss": 0.024756474047899248, + "grad_norm": 0.28954795002937317, + "learning_rate": 2.6510658500767378e-05, + "loss": 0.0474, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.01089677084237337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005110441241413355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02690398283302784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8418007314205169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15630, + "weighted_flow_matching_loss": 0.024756474047899248, + "weighted_rkd_loss": 0.022647942416369915 + }, + { + "epoch": 4.550480069828339, + "flow_matching_loss": 0.02670335490256548, + "grad_norm": 0.25840213894844055, + "learning_rate": 2.650535576238253e-05, + "loss": 0.0461, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009450085368007422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004178146831691265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02307831756770611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8416095972061157, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15640, + "weighted_flow_matching_loss": 0.02670335490256548, + "weighted_rkd_loss": 0.019422979839146138 + }, + { + "epoch": 4.553389583939482, + "flow_matching_loss": 0.02367959339171648, + "grad_norm": 0.29256919026374817, + "learning_rate": 2.650004952897438e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009843689063563943, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004365131747908891, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024052509851753712, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8414183855056763, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15650, + "weighted_flow_matching_loss": 0.02367959339171648, + "weighted_rkd_loss": 0.02023819265887141 + }, + { + "epoch": 4.5562990980506255, + "flow_matching_loss": 0.024504273012280465, + "grad_norm": 0.26709598302841187, + "learning_rate": 2.649473980215483e-05, + "loss": 0.0459, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010318106226623058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004760027024894952, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025396239757537842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8412270784378052, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15660, + "weighted_flow_matching_loss": 0.024504273012280465, + "weighted_rkd_loss": 0.021364040672779083 + }, + { + "epoch": 4.559208612161769, + "flow_matching_loss": 0.024032644741237164, + "grad_norm": 0.24545221030712128, + "learning_rate": 2.648942658353683e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0099628078751266, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004758415441028774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024684031121432783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.841035658121109, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15670, + "weighted_flow_matching_loss": 0.024032644741237164, + "weighted_rkd_loss": 0.02076022531837225 + }, + { + "epoch": 4.562118126272912, + "flow_matching_loss": 0.024488053284585476, + "grad_norm": 0.2657988965511322, + "learning_rate": 2.648410987473439e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0093266811221838, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041804966051131485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02283385880291462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8408441603183746, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15680, + "weighted_flow_matching_loss": 0.024488053284585476, + "weighted_rkd_loss": 0.019199617858976125 + }, + { + "epoch": 4.565027640384056, + "flow_matching_loss": 0.02419473771005869, + "grad_norm": 0.2790115475654602, + "learning_rate": 2.6478789677362596e-05, + "loss": 0.0457, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010309402551501989, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004909021174535155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025527825579047203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8406525671482086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15690, + "weighted_flow_matching_loss": 0.02419473771005869, + "weighted_rkd_loss": 0.021460057236254215 + }, + { + "epoch": 4.567937154495199, + "flow_matching_loss": 0.02377888225018978, + "grad_norm": 0.2944582402706146, + "learning_rate": 2.6473465993037572e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010379382176324725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004637600993737579, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0253963652998209, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8404608845710755, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15700, + "weighted_flow_matching_loss": 0.02377888225018978, + "weighted_rkd_loss": 0.02134466217830777 + }, + { + "epoch": 4.570846668606343, + "flow_matching_loss": 0.024138614535331726, + "grad_norm": 0.2120058387517929, + "learning_rate": 2.6468138823376522e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009956435579806567, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042484521865844725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02416132315993309, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8402691006660461, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15710, + "weighted_flow_matching_loss": 0.024138614535331726, + "weighted_rkd_loss": 0.02030202969908714 + }, + { + "epoch": 4.573756182717486, + "flow_matching_loss": 0.023638609796762466, + "grad_norm": 0.2331274151802063, + "learning_rate": 2.6462808169997695e-05, + "loss": 0.0454, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010609633941203355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00468941954895854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025908687524497508, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8400772213935852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15720, + "weighted_flow_matching_loss": 0.023638609796762466, + "weighted_rkd_loss": 0.02176536973565817 + }, + { + "epoch": 4.57666569682863, + "flow_matching_loss": 0.025533375702798365, + "grad_norm": 0.2891859710216522, + "learning_rate": 2.6457474034520402e-05, + "loss": 0.0479, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010783213842660188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.005096374917775393, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02666280250996351, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8398852646350861, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15730, + "weighted_flow_matching_loss": 0.025533375702798365, + "weighted_rkd_loss": 0.022393756825476886 + }, + { + "epoch": 4.579575210939773, + "flow_matching_loss": 0.02307699965313077, + "grad_norm": 0.23698191344738007, + "learning_rate": 2.6452136418565016e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00946886371821165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004198829247616232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02313655652105808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8396931946277618, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15740, + "weighted_flow_matching_loss": 0.02307699965313077, + "weighted_rkd_loss": 0.019427653029561044 + }, + { + "epoch": 4.582484725050916, + "flow_matching_loss": 0.024047712795436382, + "grad_norm": 0.20746184885501862, + "learning_rate": 2.6446795323752962e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009597430238500237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043390276376158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023533887602388857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8395010471343994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15750, + "weighted_flow_matching_loss": 0.024047712795436382, + "weighted_rkd_loss": 0.019756703078746794 + }, + { + "epoch": 4.58539423916206, + "flow_matching_loss": 0.021862179785966874, + "grad_norm": 0.24655532836914062, + "learning_rate": 2.6441450751706725e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00960278743878007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040836634812876586, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023289238475263117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8393088042736053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15760, + "weighted_flow_matching_loss": 0.021862179785966874, + "weighted_rkd_loss": 0.019546869024634362 + }, + { + "epoch": 4.588303753273204, + "flow_matching_loss": 0.0215591911226511, + "grad_norm": 0.2430257797241211, + "learning_rate": 2.6436102704049843e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009291900973767041, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004351454810239375, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02293525654822588, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8391164600849151, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15770, + "weighted_flow_matching_loss": 0.0215591911226511, + "weighted_rkd_loss": 0.01924536917358637 + }, + { + "epoch": 4.591213267384346, + "flow_matching_loss": 0.025249832682311533, + "grad_norm": 0.2432827204465866, + "learning_rate": 2.6430751182406913e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008975175628438591, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004056737851351499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022007089108228683, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8389240205287933, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15780, + "weighted_flow_matching_loss": 0.025249832682311533, + "weighted_rkd_loss": 0.018462321441620588 + }, + { + "epoch": 4.59412278149549, + "flow_matching_loss": 0.025224927067756652, + "grad_norm": 0.3087514340877533, + "learning_rate": 2.6425396188403585e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009785334207117557, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004119977331720292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02369064576923847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8387314975261688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15790, + "weighted_flow_matching_loss": 0.025224927067756652, + "weighted_rkd_loss": 0.019870087131857873 + }, + { + "epoch": 4.597032295606634, + "flow_matching_loss": 0.02283346764743328, + "grad_norm": 0.2257852852344513, + "learning_rate": 2.6420037723666565e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010068897157907486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004400029242970049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024537823721766473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8385388791561127, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15800, + "weighted_flow_matching_loss": 0.02283346764743328, + "weighted_rkd_loss": 0.020575834810733794 + }, + { + "epoch": 4.599941809717777, + "flow_matching_loss": 0.024492674693465232, + "grad_norm": 0.2333386391401291, + "learning_rate": 2.641467578982361e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00936083127744496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003987572551704943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02270923536270857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8383461713790894, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15810, + "weighted_flow_matching_loss": 0.024492674693465232, + "weighted_rkd_loss": 0.01903816033154726 + }, + { + "epoch": 4.60285132382892, + "flow_matching_loss": 0.02382706757634878, + "grad_norm": 0.2880517244338989, + "learning_rate": 2.6409310388503542e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009672951232641936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004495896329171955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02384179849177599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8381533741950988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15820, + "weighted_flow_matching_loss": 0.02382706757634878, + "weighted_rkd_loss": 0.019983086828142406 + }, + { + "epoch": 4.605760837940064, + "flow_matching_loss": 0.02287756875157356, + "grad_norm": 0.2587370276451111, + "learning_rate": 2.640394152133622e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009070100169628859, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00388026365544647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022020464204251765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8379604697227478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15830, + "weighted_flow_matching_loss": 0.02287756875157356, + "weighted_rkd_loss": 0.01845230218023062 + }, + { + "epoch": 4.608670352051208, + "flow_matching_loss": 0.025267404317855836, + "grad_norm": 0.2545614242553711, + "learning_rate": 2.6398569189952573e-05, + "loss": 0.046, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010095988772809506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004606860387139022, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024798837676644325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8377674877643585, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15840, + "weighted_flow_matching_loss": 0.025267404317855836, + "weighted_rkd_loss": 0.020775585528463127 + }, + { + "epoch": 4.611579866162351, + "flow_matching_loss": 0.022848261706531047, + "grad_norm": 0.2349938452243805, + "learning_rate": 2.6393193395984563e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009868469787761569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0044883785769343374, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024225318059325218, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8375744223594666, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15850, + "weighted_flow_matching_loss": 0.022848261706531047, + "weighted_rkd_loss": 0.020290489494800567 + }, + { + "epoch": 4.614489380273494, + "flow_matching_loss": 0.025736811570823193, + "grad_norm": 0.2647935748100281, + "learning_rate": 2.6387814141065216e-05, + "loss": 0.0461, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010010974854230881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004283329425379634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024305279180407525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.837381225824356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15860, + "weighted_flow_matching_loss": 0.025736811570823193, + "weighted_rkd_loss": 0.0203527744859457 + }, + { + "epoch": 4.617398894384638, + "flow_matching_loss": 0.022959168441593646, + "grad_norm": 0.24877876043319702, + "learning_rate": 2.6382431426828615e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009412860451266169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004152104770764708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022977825626730917, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8371879696846009, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15870, + "weighted_flow_matching_loss": 0.022959168441593646, + "weighted_rkd_loss": 0.01923665665090084 + }, + { + "epoch": 4.620308408495781, + "flow_matching_loss": 0.0255180936306715, + "grad_norm": 0.23999418318271637, + "learning_rate": 2.6377045254909885e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009524735016748309, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004334096540696919, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023383566364645957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.836994606256485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15880, + "weighted_flow_matching_loss": 0.0255180936306715, + "weighted_rkd_loss": 0.01957201100885868 + }, + { + "epoch": 4.6232179226069245, + "flow_matching_loss": 0.024970978312194348, + "grad_norm": 0.213243767619133, + "learning_rate": 2.6371655626945196e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009114685282111169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004125709808431566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022355080395936967, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8368011474609375, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15890, + "weighted_flow_matching_loss": 0.024970978312194348, + "weighted_rkd_loss": 0.018706658855080605 + }, + { + "epoch": 4.626127436718068, + "flow_matching_loss": 0.02465899847447872, + "grad_norm": 0.29156750440597534, + "learning_rate": 2.6366262544571778e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009249759139493108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004220415721647441, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022719934210181238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8366076111793518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15900, + "weighted_flow_matching_loss": 0.02465899847447872, + "weighted_rkd_loss": 0.019007638655602933 + }, + { + "epoch": 4.629036950829212, + "flow_matching_loss": 0.024616202153265477, + "grad_norm": 0.21325214207172394, + "learning_rate": 2.636086600942791e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009743576031178236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004268846823833883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02375599890947342, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.83641397356987, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15910, + "weighted_flow_matching_loss": 0.024616202153265477, + "weighted_rkd_loss": 0.01986994789913297 + }, + { + "epoch": 4.631946464940355, + "flow_matching_loss": 0.02435575518757105, + "grad_norm": 0.21951662003993988, + "learning_rate": 2.6355466023152913e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00923520028591156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003978638700209558, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02244903929531574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.836220246553421, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15920, + "weighted_flow_matching_loss": 0.02435575518757105, + "weighted_rkd_loss": 0.018772335071116687 + }, + { + "epoch": 4.6348559790514985, + "flow_matching_loss": 0.02305495385080576, + "grad_norm": 0.2713705897331238, + "learning_rate": 2.6350062587387164e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009559596609324216, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004263030528090894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023382223770022394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.836026418209076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15930, + "weighted_flow_matching_loss": 0.02305495385080576, + "weighted_rkd_loss": 0.019548134971410037 + }, + { + "epoch": 4.637765493162642, + "flow_matching_loss": 0.02534981369972229, + "grad_norm": 0.27431216835975647, + "learning_rate": 2.634465570377208e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009463622607290745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004383764369413257, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023311009258031846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8358325183391571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15940, + "weighted_flow_matching_loss": 0.02534981369972229, + "weighted_rkd_loss": 0.019483840093016624 + }, + { + "epoch": 4.640675007273785, + "flow_matching_loss": 0.02518989108502865, + "grad_norm": 0.23102059960365295, + "learning_rate": 2.633924537395013e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00902441432699561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00392603650689125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021974865160882474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8356385111808777, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15950, + "weighted_flow_matching_loss": 0.02518989108502865, + "weighted_rkd_loss": 0.018363052420318125 + }, + { + "epoch": 4.643584521384929, + "flow_matching_loss": 0.026155663840472698, + "grad_norm": 0.23599129915237427, + "learning_rate": 2.6333831599564835e-05, + "loss": 0.0472, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010203911922872066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004773923940956592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025181747600436212, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8354444146156311, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15960, + "weighted_flow_matching_loss": 0.026155663840472698, + "weighted_rkd_loss": 0.02103797122836113 + }, + { + "epoch": 4.646494035496072, + "flow_matching_loss": 0.023947573453187942, + "grad_norm": 0.3200235664844513, + "learning_rate": 2.632841438226075e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010022839158773422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004499575984664261, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024545254185795784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8352502286434174, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15970, + "weighted_flow_matching_loss": 0.023947573453187942, + "weighted_rkd_loss": 0.020501278527081012 + }, + { + "epoch": 4.649403549607215, + "flow_matching_loss": 0.02732427679002285, + "grad_norm": 0.19882401823997498, + "learning_rate": 2.6322993723683486e-05, + "loss": 0.0473, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009821833483874798, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004315470601432026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023959137685596942, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.835055947303772, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15980, + "weighted_flow_matching_loss": 0.02732427679002285, + "weighted_rkd_loss": 0.020007171668112278 + }, + { + "epoch": 4.652313063718359, + "flow_matching_loss": 0.024811545759439467, + "grad_norm": 0.1967383623123169, + "learning_rate": 2.6317569625479695e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009266829630360008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004325700248591602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022859359718859197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8348615646362305, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 15990, + "weighted_flow_matching_loss": 0.024811545759439467, + "weighted_rkd_loss": 0.01908452631905675 + }, + { + "epoch": 4.655222577829503, + "flow_matching_loss": 0.023878200724720956, + "grad_norm": 0.23154835402965546, + "learning_rate": 2.6312142089297076e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009953204914927483, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004206192540004849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024112602695822716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8346671044826508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16000, + "weighted_flow_matching_loss": 0.023878200724720956, + "weighted_rkd_loss": 0.02012593634426594 + }, + { + "epoch": 4.6581320919406455, + "flow_matching_loss": 0.02272343970835209, + "grad_norm": 0.23171882331371307, + "learning_rate": 2.6306711116784366e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009606628585606814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00428786464035511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023501121811568736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8344725489616394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16010, + "weighted_flow_matching_loss": 0.02272343970835209, + "weighted_rkd_loss": 0.019611148349940776 + }, + { + "epoch": 4.661041606051789, + "flow_matching_loss": 0.024465534277260302, + "grad_norm": 0.2773049473762512, + "learning_rate": 2.6301276709591358e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009617232531309129, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004345862381160259, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023580327257514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8342779040336609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16020, + "weighted_flow_matching_loss": 0.024465534277260302, + "weighted_rkd_loss": 0.019672497548162936 + }, + { + "epoch": 4.663951120162933, + "flow_matching_loss": 0.025160209834575654, + "grad_norm": 0.33362773060798645, + "learning_rate": 2.6295838869368878e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009985368326306343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0045055739348754285, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024476310424506664, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8340831637382508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16030, + "weighted_flow_matching_loss": 0.025160209834575654, + "weighted_rkd_loss": 0.02041519423946738 + }, + { + "epoch": 4.666860634274077, + "flow_matching_loss": 0.023541827872395515, + "grad_norm": 0.22692430019378662, + "learning_rate": 2.6290397597768803e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009121541073545814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003907613386400044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022150695510208606, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8338883221149445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16040, + "weighted_flow_matching_loss": 0.023541827872395515, + "weighted_rkd_loss": 0.01847125282511115 + }, + { + "epoch": 4.669770148385219, + "flow_matching_loss": 0.02303944490849972, + "grad_norm": 0.2032291144132614, + "learning_rate": 2.6284952896444037e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010003233421593904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004612027830444276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024618494510650634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8336934089660645, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16050, + "weighted_flow_matching_loss": 0.02303944490849972, + "weighted_rkd_loss": 0.02052424065768719 + }, + { + "epoch": 4.672679662496363, + "flow_matching_loss": 0.023372665978968143, + "grad_norm": 0.19866980612277985, + "learning_rate": 2.6279504767048547e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009402027493342758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004199328762479126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023003383725881576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8334983944892883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16060, + "weighted_flow_matching_loss": 0.023372665978968143, + "weighted_rkd_loss": 0.01917327931150794 + }, + { + "epoch": 4.675589176607507, + "flow_matching_loss": 0.024061007425189018, + "grad_norm": 0.2493414431810379, + "learning_rate": 2.6274053211237323e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008801263105124236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003756916685961187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021359443105757235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8333032906055451, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16070, + "weighted_flow_matching_loss": 0.024061007425189018, + "weighted_rkd_loss": 0.01779882637783885 + }, + { + "epoch": 4.67849869071865, + "flow_matching_loss": 0.02484236713498831, + "grad_norm": 0.22092725336551666, + "learning_rate": 2.626859823066641e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009106037765741348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038776012137532235, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022089676558971406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8331080913543701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16080, + "weighted_flow_matching_loss": 0.02484236713498831, + "weighted_rkd_loss": 0.01840301724150777 + }, + { + "epoch": 4.681408204829793, + "flow_matching_loss": 0.02424466107040644, + "grad_norm": 0.25554484128952026, + "learning_rate": 2.6263139826992886e-05, + "loss": 0.0424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008863728633150459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004097459651529789, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021824917010962962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8329128086566925, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16090, + "weighted_flow_matching_loss": 0.02424466107040644, + "weighted_rkd_loss": 0.018178160209208725 + }, + { + "epoch": 4.684317718940937, + "flow_matching_loss": 0.024456286430358888, + "grad_norm": 0.2679002583026886, + "learning_rate": 2.625767800187487e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010099751222878695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004574318462982774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02477382104843855, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8327174246311188, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16100, + "weighted_flow_matching_loss": 0.024456286430358888, + "weighted_rkd_loss": 0.02062970781698823 + }, + { + "epoch": 4.687227233052081, + "flow_matching_loss": 0.022872299980372192, + "grad_norm": 0.22941704094409943, + "learning_rate": 2.6252212756971514e-05, + "loss": 0.0424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009564409358426928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004312059911899269, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02344087865203619, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8325219690799713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16110, + "weighted_flow_matching_loss": 0.022872299980372192, + "weighted_rkd_loss": 0.01951501751318574 + }, + { + "epoch": 4.690136747163224, + "flow_matching_loss": 0.025585980713367464, + "grad_norm": 0.2390533685684204, + "learning_rate": 2.6246744093943022e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009062740160152316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004039862751960754, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02216534297913313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8323263883590698, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16120, + "weighted_flow_matching_loss": 0.025585980713367464, + "weighted_rkd_loss": 0.018448747415095567 + }, + { + "epoch": 4.693046261274367, + "flow_matching_loss": 0.02285037059336901, + "grad_norm": 0.22544007003307343, + "learning_rate": 2.6241272014450625e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008323720004409552, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036940358113497496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02034147549420595, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.832130742073059, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16130, + "weighted_flow_matching_loss": 0.02285037059336901, + "weighted_rkd_loss": 0.01692668404430151 + }, + { + "epoch": 4.69595577538551, + "flow_matching_loss": 0.025352517515420912, + "grad_norm": 0.2357345074415207, + "learning_rate": 2.6235796520156603e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008774691447615623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004053770028986037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02160315290093422, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8319349884986877, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16140, + "weighted_flow_matching_loss": 0.025352517515420912, + "weighted_rkd_loss": 0.01797245377674699 + }, + { + "epoch": 4.698865289496654, + "flow_matching_loss": 0.024166224896907805, + "grad_norm": 0.27334821224212646, + "learning_rate": 2.623031761272426e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009241702407598496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004298529750667512, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022781934402883054, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8317391633987427, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16150, + "weighted_flow_matching_loss": 0.024166224896907805, + "weighted_rkd_loss": 0.018948668055236338 + }, + { + "epoch": 4.7017748036077975, + "flow_matching_loss": 0.024004792608320714, + "grad_norm": 0.2594930827617645, + "learning_rate": 2.6224835293817948e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00946783679537475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00422384082339704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023159514553844927, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8315432369709015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16160, + "weighted_flow_matching_loss": 0.024004792608320714, + "weighted_rkd_loss": 0.019258126523345707 + }, + { + "epoch": 4.704684317718941, + "flow_matching_loss": 0.023877363838255405, + "grad_norm": 0.28760117292404175, + "learning_rate": 2.6219349565103044e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00992382545955479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0046044661896303294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02445211708545685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8313472211360932, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16170, + "weighted_flow_matching_loss": 0.023877363838255405, + "weighted_rkd_loss": 0.020328155905008315 + }, + { + "epoch": 4.707593831830084, + "flow_matching_loss": 0.02534552812576294, + "grad_norm": 0.20008137822151184, + "learning_rate": 2.6213860428245974e-05, + "loss": 0.0441, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009272907907143236, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040746126789599655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022620428539812564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8311511099338531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16180, + "weighted_flow_matching_loss": 0.02534552812576294, + "weighted_rkd_loss": 0.01880107680335641 + }, + { + "epoch": 4.710503345941228, + "flow_matching_loss": 0.024331963248550893, + "grad_norm": 0.2549886107444763, + "learning_rate": 2.620836788491419e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009137162659317254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003970694541931152, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022245019674301147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8309549152851105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16190, + "weighted_flow_matching_loss": 0.024331963248550893, + "weighted_rkd_loss": 0.018484564032405615 + }, + { + "epoch": 4.7134128600523715, + "flow_matching_loss": 0.02334036659449339, + "grad_norm": 0.21612347662448883, + "learning_rate": 2.6202871936776183e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009383747261017561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00406212629750371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022829620540142058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8307586252689362, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16200, + "weighted_flow_matching_loss": 0.02334036659449339, + "weighted_rkd_loss": 0.01896595936268568 + }, + { + "epoch": 4.716322374163514, + "flow_matching_loss": 0.02341005839407444, + "grad_norm": 0.27623626589775085, + "learning_rate": 2.6197372585501477e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0103720523416996, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004663953254930675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025408058241009713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8305622458457946, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16210, + "weighted_flow_matching_loss": 0.02341005839407444, + "weighted_rkd_loss": 0.021103053539991378 + }, + { + "epoch": 4.719231888274658, + "flow_matching_loss": 0.02544729020446539, + "grad_norm": 0.27183014154434204, + "learning_rate": 2.6191869832760626e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009604084445163608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004265862517058849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02347403168678284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.830365777015686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16220, + "weighted_flow_matching_loss": 0.02544729020446539, + "weighted_rkd_loss": 0.019492067210376263 + }, + { + "epoch": 4.722141402385802, + "flow_matching_loss": 0.02569067608565092, + "grad_norm": 0.2451048195362091, + "learning_rate": 2.618636368022523e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009400002844631671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004350139247253537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023150144889950753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8301692128181457, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16230, + "weighted_flow_matching_loss": 0.02569067608565092, + "weighted_rkd_loss": 0.019218547083437443 + }, + { + "epoch": 4.725050916496945, + "flow_matching_loss": 0.023952174559235574, + "grad_norm": 0.2067045271396637, + "learning_rate": 2.6180854129567902e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009599147317931055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004245555284433067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023443849943578243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8299725651741028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16240, + "weighted_flow_matching_loss": 0.023952174559235574, + "weighted_rkd_loss": 0.019457751791924237 + }, + { + "epoch": 4.727960430608088, + "flow_matching_loss": 0.02621483076363802, + "grad_norm": 0.24943670630455017, + "learning_rate": 2.6175341182462303e-05, + "loss": 0.0453, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00941961696371436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004207591223530471, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02304682508111, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8297758221626281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16250, + "weighted_flow_matching_loss": 0.02621483076363802, + "weighted_rkd_loss": 0.019123745430260897 + }, + { + "epoch": 4.730869944719232, + "flow_matching_loss": 0.024158562906086443, + "grad_norm": 0.26863977313041687, + "learning_rate": 2.6169824840583124e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009945450536906719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004556271363981068, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024447172321379183, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.829578983783722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16260, + "weighted_flow_matching_loss": 0.024158562906086443, + "weighted_rkd_loss": 0.02028078744187951 + }, + { + "epoch": 4.733779458830376, + "flow_matching_loss": 0.024154984578490257, + "grad_norm": 0.2804281413555145, + "learning_rate": 2.6164305105606076e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009127399418503046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00398348590824753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022238284721970557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.829382061958313, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16270, + "weighted_flow_matching_loss": 0.024154984578490257, + "weighted_rkd_loss": 0.018443928007036446 + }, + { + "epoch": 4.7366889729415185, + "flow_matching_loss": 0.024859672971069814, + "grad_norm": 0.21921445429325104, + "learning_rate": 2.6158781979207918e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008017243118956685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036559179075993597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019690404180437326, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8291850507259368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16280, + "weighted_flow_matching_loss": 0.024859672971069814, + "weighted_rkd_loss": 0.01632703598588705 + }, + { + "epoch": 4.739598487052662, + "flow_matching_loss": 0.022510454058647156, + "grad_norm": 0.27755293250083923, + "learning_rate": 2.615325546306642e-05, + "loss": 0.0421, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009683048445731401, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004285187483765185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023651284351944923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8289879441261292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16290, + "weighted_flow_matching_loss": 0.022510454058647156, + "weighted_rkd_loss": 0.019606497883796693 + }, + { + "epoch": 4.742508001163806, + "flow_matching_loss": 0.023525376804172993, + "grad_norm": 0.23216408491134644, + "learning_rate": 2.61477255588604e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009488889062777162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004269729088991881, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023247507214546204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8287907540798187, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16300, + "weighted_flow_matching_loss": 0.023525376804172993, + "weighted_rkd_loss": 0.019267329014837742 + }, + { + "epoch": 4.745417515274949, + "flow_matching_loss": 0.023833177611231805, + "grad_norm": 0.23885388672351837, + "learning_rate": 2.614219226826969e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008777955453842879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003951799543574453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021507710590958594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8285934627056122, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16310, + "weighted_flow_matching_loss": 0.023833177611231805, + "weighted_rkd_loss": 0.017821153812110423 + }, + { + "epoch": 4.748327029386092, + "flow_matching_loss": 0.025801135785877706, + "grad_norm": 0.21703821420669556, + "learning_rate": 2.6136655592975166e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009687957679852843, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004480558400973678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02385647390037775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8283960938453674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16320, + "weighted_flow_matching_loss": 0.025801135785877706, + "weighted_rkd_loss": 0.019762679748237134 + }, + { + "epoch": 4.751236543497236, + "flow_matching_loss": 0.023682618513703346, + "grad_norm": 0.25327858328819275, + "learning_rate": 2.613111553465872e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008972844015806914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003925373312085867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021871061250567436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8281986296176911, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16330, + "weighted_flow_matching_loss": 0.023682618513703346, + "weighted_rkd_loss": 0.018113551940768958 + }, + { + "epoch": 4.754146057608379, + "flow_matching_loss": 0.02536178883165121, + "grad_norm": 0.2838883101940155, + "learning_rate": 2.6125572095003276e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009560103621333837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004016769281588495, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023136976547539233, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.828001081943512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16340, + "weighted_flow_matching_loss": 0.02536178883165121, + "weighted_rkd_loss": 0.01915733739733696 + }, + { + "epoch": 4.757055571719523, + "flow_matching_loss": 0.026191908307373522, + "grad_norm": 0.26075631380081177, + "learning_rate": 2.6120025275692777e-05, + "loss": 0.045, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009146337816491722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004390124720521272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02268280033022165, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8278034329414368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16350, + "weighted_flow_matching_loss": 0.026191908307373522, + "weighted_rkd_loss": 0.018776828330010174 + }, + { + "epoch": 4.759965085830666, + "flow_matching_loss": 0.025807429663836957, + "grad_norm": 0.24440591037273407, + "learning_rate": 2.6114475078412212e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008987601194530726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003964619408361614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02193982172757387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8276057004928589, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16360, + "weighted_flow_matching_loss": 0.025807429663836957, + "weighted_rkd_loss": 0.01815749714151025 + }, + { + "epoch": 4.76287459994181, + "flow_matching_loss": 0.024377394281327725, + "grad_norm": 0.24625268578529358, + "learning_rate": 2.6108921504847575e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009479326056316495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004453727998770774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023412379808723925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8274078905582428, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16370, + "weighted_flow_matching_loss": 0.024377394281327725, + "weighted_rkd_loss": 0.01937157204374671 + }, + { + "epoch": 4.765784114052953, + "flow_matching_loss": 0.022890702448785306, + "grad_norm": 0.20136724412441254, + "learning_rate": 2.6103364556685902e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009746411023661495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004356405488215387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023849227465689182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8272099673748017, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16380, + "weighted_flow_matching_loss": 0.022890702448785306, + "weighted_rkd_loss": 0.019728285167366265 + }, + { + "epoch": 4.768693628164097, + "flow_matching_loss": 0.02456074170768261, + "grad_norm": 0.27715539932250977, + "learning_rate": 2.6097804235615242e-05, + "loss": 0.0449, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010085754562169313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004429149301722645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02460065856575966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8270119607448578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16390, + "weighted_flow_matching_loss": 0.02456074170768261, + "weighted_rkd_loss": 0.020344942063093185 + }, + { + "epoch": 4.77160314227524, + "flow_matching_loss": 0.023323452100157738, + "grad_norm": 0.23688073456287384, + "learning_rate": 2.6092240543324676e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008783295191824437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038542613852769136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021420851442962886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8268138706684113, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16400, + "weighted_flow_matching_loss": 0.023323452100157738, + "weighted_rkd_loss": 0.01771110724657774 + }, + { + "epoch": 4.774512656386383, + "flow_matching_loss": 0.02360700573772192, + "grad_norm": 0.2381332367658615, + "learning_rate": 2.6086673481504303e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009488347638398408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003825806430540979, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022802501544356348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8266156852245331, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16410, + "weighted_flow_matching_loss": 0.02360700573772192, + "weighted_rkd_loss": 0.018848862126469614 + }, + { + "epoch": 4.777422170497527, + "flow_matching_loss": 0.0235106335952878, + "grad_norm": 0.20763316750526428, + "learning_rate": 2.608110305184526e-05, + "loss": 0.0432, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009745610039681196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004314705380238593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02380592543631792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8264174163341522, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16420, + "weighted_flow_matching_loss": 0.0235106335952878, + "weighted_rkd_loss": 0.01967366263270378 + }, + { + "epoch": 4.7803316846086705, + "flow_matching_loss": 0.02225065100938082, + "grad_norm": 0.2182702124118805, + "learning_rate": 2.6075529256039687e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008866524649783969, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004019447136670351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021752496436238288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8262190520763397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16430, + "weighted_flow_matching_loss": 0.02225065100938082, + "weighted_rkd_loss": 0.017972452100366354 + }, + { + "epoch": 4.783241198719814, + "flow_matching_loss": 0.022032016701996327, + "grad_norm": 0.2255309373140335, + "learning_rate": 2.6069952095780756e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009273701254278422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003981315181590617, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022528717759996653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.826020622253418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16440, + "weighted_flow_matching_loss": 0.022032016701996327, + "weighted_rkd_loss": 0.018609193805605172 + }, + { + "epoch": 4.786150712830957, + "flow_matching_loss": 0.024310952983796597, + "grad_norm": 0.24514533579349518, + "learning_rate": 2.6064371572762667e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00910398899577558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003965095127932727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022173073142766953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8258220613002777, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16450, + "weighted_flow_matching_loss": 0.024310952983796597, + "weighted_rkd_loss": 0.018311109114438295 + }, + { + "epoch": 4.789060226942101, + "flow_matching_loss": 0.02159520862624049, + "grad_norm": 0.24364270269870758, + "learning_rate": 2.605878768868063e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009143299050629139, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004025564272888004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022312162071466447, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8256234347820282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16460, + "weighted_flow_matching_loss": 0.02159520862624049, + "weighted_rkd_loss": 0.018421286717057228 + }, + { + "epoch": 4.7919697410532445, + "flow_matching_loss": 0.023248793743550777, + "grad_norm": 0.22032111883163452, + "learning_rate": 2.605320044523088e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009044850105419755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003944546170532704, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022034246381372212, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8254247188568116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16470, + "weighted_flow_matching_loss": 0.023248793743550777, + "weighted_rkd_loss": 0.018187751155346633 + }, + { + "epoch": 4.794879255164387, + "flow_matching_loss": 0.024068796448409556, + "grad_norm": 0.23837536573410034, + "learning_rate": 2.6047609844110683e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00933400415815413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004394443007186055, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023062451370060445, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8252259135246277, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16480, + "weighted_flow_matching_loss": 0.024068796448409556, + "weighted_rkd_loss": 0.019031797908246516 + }, + { + "epoch": 4.797788769275531, + "flow_matching_loss": 0.022178039699792863, + "grad_norm": 0.22336804866790771, + "learning_rate": 2.604201588701831e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010159911122173071, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004525862983427942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024845685064792632, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8250270128250122, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16490, + "weighted_flow_matching_loss": 0.022178039699792863, + "weighted_rkd_loss": 0.020498496666550636 + }, + { + "epoch": 4.800698283386675, + "flow_matching_loss": 0.023035365901887418, + "grad_norm": 0.2600865066051483, + "learning_rate": 2.6036418575653057e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009412686759606003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041422718204557896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022967645525932313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8248280227184296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16500, + "weighted_flow_matching_loss": 0.023035365901887418, + "weighted_rkd_loss": 0.018944362085312604 + }, + { + "epoch": 4.8036077974978175, + "flow_matching_loss": 0.024865853041410445, + "grad_norm": 0.21692273020744324, + "learning_rate": 2.603081791171524e-05, + "loss": 0.0438, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009371501579880714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004178976290859282, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022921979241073132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8246289491653442, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16510, + "weighted_flow_matching_loss": 0.024865853041410445, + "weighted_rkd_loss": 0.01890212409198284 + }, + { + "epoch": 4.806517311608961, + "flow_matching_loss": 0.02262820806354284, + "grad_norm": 0.306361585855484, + "learning_rate": 2.602521389690619e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009440376423299313, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004146836302243173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023027588799595832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8244297862052917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16520, + "weighted_flow_matching_loss": 0.02262820806354284, + "weighted_rkd_loss": 0.018984626699239015 + }, + { + "epoch": 4.809426825720105, + "flow_matching_loss": 0.021015338599681854, + "grad_norm": 0.21454152464866638, + "learning_rate": 2.601960653292827e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009371399693191052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00393648671451956, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022679285891354085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.824230533838272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16530, + "weighted_flow_matching_loss": 0.021015338599681854, + "weighted_rkd_loss": 0.018693018797785045 + }, + { + "epoch": 4.812336339831248, + "flow_matching_loss": 0.02433736901730299, + "grad_norm": 0.31988146901130676, + "learning_rate": 2.6013995821484833e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009892035368829965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004551132302731275, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024335203133523463, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8240311980247498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16540, + "weighted_flow_matching_loss": 0.02433736901730299, + "weighted_rkd_loss": 0.02005289774388075 + }, + { + "epoch": 4.8152458539423915, + "flow_matching_loss": 0.022118286602199076, + "grad_norm": 0.22272373735904694, + "learning_rate": 2.600838176428028e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008876956114545465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004093066812492907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021846978925168514, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8238317549228669, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16550, + "weighted_flow_matching_loss": 0.022118286602199076, + "weighted_rkd_loss": 0.01799826817587018 + }, + { + "epoch": 4.818155368053535, + "flow_matching_loss": 0.023526089265942574, + "grad_norm": 0.22688116133213043, + "learning_rate": 2.6002764363020004e-05, + "loss": 0.0424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009330610232427716, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004307877109386027, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022969097644090653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8236322402954102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16560, + "weighted_flow_matching_loss": 0.023526089265942574, + "weighted_rkd_loss": 0.01891804179176688 + }, + { + "epoch": 4.821064882164679, + "flow_matching_loss": 0.025207607448101042, + "grad_norm": 0.2383636236190796, + "learning_rate": 2.5997143619410427e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008667489187791944, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004088311968371272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02142329029738903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8234326422214509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16570, + "weighted_flow_matching_loss": 0.025207607448101042, + "weighted_rkd_loss": 0.01764055211097002 + }, + { + "epoch": 4.823974396275822, + "flow_matching_loss": 0.023580464348196984, + "grad_norm": 0.21399661898612976, + "learning_rate": 2.5991519535158976e-05, + "loss": 0.0411, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008604772621765733, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0040478212758898735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02125736642628908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8232329368591309, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16580, + "weighted_flow_matching_loss": 0.023580464348196984, + "weighted_rkd_loss": 0.017499763891100882 + }, + { + "epoch": 4.826883910386965, + "flow_matching_loss": 0.0255604337900877, + "grad_norm": 0.22446487843990326, + "learning_rate": 2.598589211197411e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008784708473831416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003652710234746337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02122212741523981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8230331540107727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16590, + "weighted_flow_matching_loss": 0.0255604337900877, + "weighted_rkd_loss": 0.017466490808874368 + }, + { + "epoch": 4.829793424498109, + "flow_matching_loss": 0.024543180875480176, + "grad_norm": 0.26899728178977966, + "learning_rate": 2.598026135156528e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008703057281672954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037357572931796313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021141871996223926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8228332817554473, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16600, + "weighted_flow_matching_loss": 0.024543180875480176, + "weighted_rkd_loss": 0.01739616310223937 + }, + { + "epoch": 4.832702938609252, + "flow_matching_loss": 0.02270646598190069, + "grad_norm": 0.22793374955654144, + "learning_rate": 2.5974627255642974e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0095882308203727, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004686207906343043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023862669803202154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8226333200931549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16610, + "weighted_flow_matching_loss": 0.02270646598190069, + "weighted_rkd_loss": 0.019630317203700544 + }, + { + "epoch": 4.835612452720396, + "flow_matching_loss": 0.024152496457099916, + "grad_norm": 0.24049900472164154, + "learning_rate": 2.5968989825918674e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0086952555924654, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037100084591656923, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021100519597530364, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8224332571029663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16620, + "weighted_flow_matching_loss": 0.024152496457099916, + "weighted_rkd_loss": 0.017353804130107165 + }, + { + "epoch": 4.838521966831539, + "flow_matching_loss": 0.0247727420181036, + "grad_norm": 0.27118489146232605, + "learning_rate": 2.5963349064104887e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009070778405293823, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004172510839998722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022314067743718625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8222331225872039, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16630, + "weighted_flow_matching_loss": 0.0247727420181036, + "weighted_rkd_loss": 0.018347281869500874 + }, + { + "epoch": 4.841431480942683, + "flow_matching_loss": 0.02488541789352894, + "grad_norm": 0.24765664339065552, + "learning_rate": 2.5957704971915116e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009507000027224422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004455818608403206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023469818755984306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8220328688621521, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16640, + "weighted_flow_matching_loss": 0.02488541789352894, + "weighted_rkd_loss": 0.019292932562530042 + }, + { + "epoch": 4.844340995053826, + "flow_matching_loss": 0.02241319213062525, + "grad_norm": 0.23770473897457123, + "learning_rate": 2.5952057551063905e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009678375534713268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004345043795183301, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023701795004308225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8218325793743133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16650, + "weighted_flow_matching_loss": 0.02241319213062525, + "weighted_rkd_loss": 0.019478828832507133 + }, + { + "epoch": 4.84725050916497, + "flow_matching_loss": 0.02288006879389286, + "grad_norm": 0.22286537289619446, + "learning_rate": 2.594640680326678e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00895116371102631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004055568622425199, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021957896091043948, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8216321706771851, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16660, + "weighted_flow_matching_loss": 0.02288006879389286, + "weighted_rkd_loss": 0.018041417375206946 + }, + { + "epoch": 4.850160023276112, + "flow_matching_loss": 0.022936680354177952, + "grad_norm": 0.22943836450576782, + "learning_rate": 2.594075273024029e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00881365854293108, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004009751323610544, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02163706813007593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8214316844940186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16670, + "weighted_flow_matching_loss": 0.022936680354177952, + "weighted_rkd_loss": 0.01777330311015248 + }, + { + "epoch": 4.853069537387256, + "flow_matching_loss": 0.022594749741256238, + "grad_norm": 0.19294340908527374, + "learning_rate": 2.5935095333701994e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009060660609975457, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004229117720387876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022350438870489597, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8212311029434204, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16680, + "weighted_flow_matching_loss": 0.022594749741256238, + "weighted_rkd_loss": 0.018354906421154737 + }, + { + "epoch": 4.8559790514984, + "flow_matching_loss": 0.02625818159431219, + "grad_norm": 0.3020685911178589, + "learning_rate": 2.5929434615370462e-05, + "loss": 0.0456, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009593916358426213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004417872079648078, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0236057048663497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8210304260253907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16690, + "weighted_flow_matching_loss": 0.02625818159431219, + "weighted_rkd_loss": 0.019380884524434806 + }, + { + "epoch": 4.8588885656095435, + "flow_matching_loss": 0.02339399829506874, + "grad_norm": 0.21920831501483917, + "learning_rate": 2.592377057696527e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008133330009877681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035781784914433955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0198448384180665, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8208296775817872, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16700, + "weighted_flow_matching_loss": 0.02339399829506874, + "weighted_rkd_loss": 0.016289212461560964 + }, + { + "epoch": 4.861798079720686, + "flow_matching_loss": 0.02282757069915533, + "grad_norm": 0.22050237655639648, + "learning_rate": 2.5918103220206997e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009426724445074797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004246067232452333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023099516145884992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8206288397312165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16710, + "weighted_flow_matching_loss": 0.02282757069915533, + "weighted_rkd_loss": 0.018956142757087947 + }, + { + "epoch": 4.86470759383183, + "flow_matching_loss": 0.024008793756365776, + "grad_norm": 0.27071863412857056, + "learning_rate": 2.5912432546817247e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008597276965156197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038312300108373164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021025784127414227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8204279005527496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16720, + "weighted_flow_matching_loss": 0.024008793756365776, + "weighted_rkd_loss": 0.01725008450448513 + }, + { + "epoch": 4.867617107942974, + "flow_matching_loss": 0.0244474571198225, + "grad_norm": 0.2167385071516037, + "learning_rate": 2.5906758558518613e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00960497404448688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004528858722187579, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02373880669474602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8202268779277802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16730, + "weighted_flow_matching_loss": 0.0244474571198225, + "weighted_rkd_loss": 0.019471189938485622 + }, + { + "epoch": 4.870526622054117, + "flow_matching_loss": 0.02491117548197508, + "grad_norm": 0.21182727813720703, + "learning_rate": 2.5901081257034706e-05, + "loss": 0.0431, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009069257415831089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039929494028911, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022131464257836343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.820025771856308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16740, + "weighted_flow_matching_loss": 0.02491117548197508, + "weighted_rkd_loss": 0.018148475978523492 + }, + { + "epoch": 4.87343613616526, + "flow_matching_loss": 0.022842416912317275, + "grad_norm": 0.22299732267856598, + "learning_rate": 2.5895400644090138e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008644981821998955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003824517223984003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021114481054246424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8198245763778687, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16750, + "weighted_flow_matching_loss": 0.022842416912317275, + "weighted_rkd_loss": 0.017310250923037528 + }, + { + "epoch": 4.876345650276404, + "flow_matching_loss": 0.02389195915311575, + "grad_norm": 0.2322121411561966, + "learning_rate": 2.5889716721410535e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00875177257694304, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004016778920777142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02152032423764467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8196232914924622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16760, + "weighted_flow_matching_loss": 0.02389195915311575, + "weighted_rkd_loss": 0.017638526298105715 + }, + { + "epoch": 4.879255164387548, + "flow_matching_loss": 0.0216894693672657, + "grad_norm": 0.25170207023620605, + "learning_rate": 2.5884029490722515e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00908828703686595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004030044400133192, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022206618450582027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.819421923160553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16770, + "weighted_flow_matching_loss": 0.0216894693672657, + "weighted_rkd_loss": 0.018196620419621466 + }, + { + "epoch": 4.8821646784986905, + "flow_matching_loss": 0.024453817680478097, + "grad_norm": 0.2373252660036087, + "learning_rate": 2.587833895375371e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008492786297574639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037777101853862403, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020763282664120197, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8192204654216766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16780, + "weighted_flow_matching_loss": 0.024453817680478097, + "weighted_rkd_loss": 0.01700973743572831 + }, + { + "epoch": 4.885074192609834, + "flow_matching_loss": 0.025882495380938054, + "grad_norm": 0.23443514108657837, + "learning_rate": 2.587264511223276e-05, + "loss": 0.0444, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009223079076036811, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004160713311284781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0226068714633584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8190189123153686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16790, + "weighted_flow_matching_loss": 0.025882495380938054, + "weighted_rkd_loss": 0.018515404406934978 + }, + { + "epoch": 4.887983706720978, + "flow_matching_loss": 0.02331072520464659, + "grad_norm": 0.242916539311409, + "learning_rate": 2.5866947967889295e-05, + "loss": 0.0427, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00957397660240531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004567853850312531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023715807311236858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8188172817230225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16800, + "weighted_flow_matching_loss": 0.02331072520464659, + "weighted_rkd_loss": 0.01941878441721201 + }, + { + "epoch": 4.890893220832121, + "flow_matching_loss": 0.022093171067535878, + "grad_norm": 0.24623776972293854, + "learning_rate": 2.586124752245397e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009087215689942241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004057187424041331, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022231618873775005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8186155557632446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16810, + "weighted_flow_matching_loss": 0.022093171067535878, + "weighted_rkd_loss": 0.018199173267930746 + }, + { + "epoch": 4.8938027349432645, + "flow_matching_loss": 0.02528779972344637, + "grad_norm": 0.2485160082578659, + "learning_rate": 2.585554377765842e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009126715920865535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004114450071938336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022367881797254085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8184137463569641, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16820, + "weighted_flow_matching_loss": 0.02528779972344637, + "weighted_rkd_loss": 0.01830611387267709 + }, + { + "epoch": 4.896712249054408, + "flow_matching_loss": 0.022283873148262502, + "grad_norm": 0.22577224671840668, + "learning_rate": 2.584983673523529e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00898541221395135, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004208976798690855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022179801389575003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8182118475437165, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16830, + "weighted_flow_matching_loss": 0.022283873148262502, + "weighted_rkd_loss": 0.018147784657776354 + }, + { + "epoch": 4.899621763165551, + "flow_matching_loss": 0.024092743545770644, + "grad_norm": 0.28332918882369995, + "learning_rate": 2.5844126396918236e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009711626544594764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004380950191989541, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023804203420877457, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.818009877204895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16840, + "weighted_flow_matching_loss": 0.024092743545770644, + "weighted_rkd_loss": 0.019472109246999027 + }, + { + "epoch": 4.902531277276695, + "flow_matching_loss": 0.022923637367784976, + "grad_norm": 0.22656668722629547, + "learning_rate": 2.5838412764441905e-05, + "loss": 0.043, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010128196701407433, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004336825083009898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024593218229711056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8178078055381774, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16850, + "weighted_flow_matching_loss": 0.022923637367784976, + "weighted_rkd_loss": 0.020112583879381418 + }, + { + "epoch": 4.905440791387838, + "flow_matching_loss": 0.025249440781772138, + "grad_norm": 0.26817047595977783, + "learning_rate": 2.583269583954195e-05, + "loss": 0.0448, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009817038709297777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00429843224119395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023932509869337083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8176056444644928, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16860, + "weighted_flow_matching_loss": 0.025249440781772138, + "weighted_rkd_loss": 0.01956741716712713 + }, + { + "epoch": 4.908350305498981, + "flow_matching_loss": 0.02382504791021347, + "grad_norm": 0.22614507377147675, + "learning_rate": 2.5826975623955017e-05, + "loss": 0.0423, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009264208003878593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004039859166368842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022568275406956672, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8174033880233764, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16870, + "weighted_flow_matching_loss": 0.02382504791021347, + "weighted_rkd_loss": 0.018447450175881386 + }, + { + "epoch": 4.911259819610125, + "flow_matching_loss": 0.022266579046845435, + "grad_norm": 0.23851297795772552, + "learning_rate": 2.5821252119418762e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009484393196180463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004266121843829751, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023234908096492292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8172010719776154, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16880, + "weighted_flow_matching_loss": 0.022266579046845435, + "weighted_rkd_loss": 0.018987544532865285 + }, + { + "epoch": 4.914169333721269, + "flow_matching_loss": 0.022665855661034583, + "grad_norm": 0.2336612045764923, + "learning_rate": 2.581552532767183e-05, + "loss": 0.0411, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009234371408820153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004085590853355825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022554333508014678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8169986486434937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16890, + "weighted_flow_matching_loss": 0.022665855661034583, + "weighted_rkd_loss": 0.01842689011245966 + }, + { + "epoch": 4.917078847832412, + "flow_matching_loss": 0.02346168868243694, + "grad_norm": 0.2858023941516876, + "learning_rate": 2.580979525045387e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008976038685068488, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004013383993878961, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021965461410582067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8167961418628693, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16900, + "weighted_flow_matching_loss": 0.02346168868243694, + "weighted_rkd_loss": 0.017941272351890802 + }, + { + "epoch": 4.919988361943555, + "flow_matching_loss": 0.024024689570069313, + "grad_norm": 0.2064279168844223, + "learning_rate": 2.580406188950553e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008928639674559235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041583707788959146, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022015650011599063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8165935277938843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16910, + "weighted_flow_matching_loss": 0.024024689570069313, + "weighted_rkd_loss": 0.01797795733436942 + }, + { + "epoch": 4.922897876054699, + "flow_matching_loss": 0.02465870399028063, + "grad_norm": 0.2217634916305542, + "learning_rate": 2.5798325246568453e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008533502975478769, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036430438049137592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02071004994213581, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8163908660411835, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16920, + "weighted_flow_matching_loss": 0.02465870399028063, + "weighted_rkd_loss": 0.016907581966370343 + }, + { + "epoch": 4.925807390165843, + "flow_matching_loss": 0.021939363889396192, + "grad_norm": 0.2214001715183258, + "learning_rate": 2.579258532338528e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009405244840309024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004006703570485115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02281719297170639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8161881029605865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16930, + "weighted_flow_matching_loss": 0.021939363889396192, + "weighted_rkd_loss": 0.01862317193299532 + }, + { + "epoch": 4.928716904276985, + "flow_matching_loss": 0.02417986188083887, + "grad_norm": 0.23857621848583221, + "learning_rate": 2.5786842121699644e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009459750680252909, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003908916283398867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0228284178301692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8159852504730225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16940, + "weighted_flow_matching_loss": 0.02417986188083887, + "weighted_rkd_loss": 0.018627618346363305 + }, + { + "epoch": 4.931626418388129, + "flow_matching_loss": 0.023814916610717773, + "grad_norm": 0.23657578229904175, + "learning_rate": 2.5781095643256186e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009098779829218983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004358785413205624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022556345351040362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8157823085784912, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16950, + "weighted_flow_matching_loss": 0.023814916610717773, + "weighted_rkd_loss": 0.018401000555604695 + }, + { + "epoch": 4.934535932499273, + "flow_matching_loss": 0.022023655101656915, + "grad_norm": 0.21967823803424835, + "learning_rate": 2.577534588980053e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009601586451753973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004274628660641611, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023477801866829396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8155792891979218, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16960, + "weighted_flow_matching_loss": 0.022023655101656915, + "weighted_rkd_loss": 0.019148042984306814 + }, + { + "epoch": 4.9374454466104165, + "flow_matching_loss": 0.02278341930359602, + "grad_norm": 0.24964213371276855, + "learning_rate": 2.5769592863079298e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010194236785173416, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004732207045890391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025120680592954157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8153761684894562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16970, + "weighted_flow_matching_loss": 0.02278341930359602, + "weighted_rkd_loss": 0.020482911076396704 + }, + { + "epoch": 4.940354960721559, + "flow_matching_loss": 0.02378746196627617, + "grad_norm": 0.2756338119506836, + "learning_rate": 2.576383656484011e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009356307657435536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004256536206230521, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022969151102006435, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8151729881763459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16980, + "weighted_flow_matching_loss": 0.02378746196627617, + "weighted_rkd_loss": 0.018723776657134296 + }, + { + "epoch": 4.943264474832703, + "flow_matching_loss": 0.02538830544799566, + "grad_norm": 0.19053305685520172, + "learning_rate": 2.5758076996831572e-05, + "loss": 0.0439, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009268435463309289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00413099923171103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022667869739234448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8149696886539459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 16990, + "weighted_flow_matching_loss": 0.02538830544799566, + "weighted_rkd_loss": 0.018473816569894552 + }, + { + "epoch": 4.946173988943847, + "flow_matching_loss": 0.02264108434319496, + "grad_norm": 0.20465002954006195, + "learning_rate": 2.5752314160803295e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008306272001937032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036209352081641557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020233479142189027, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8147663176059723, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17000, + "weighted_flow_matching_loss": 0.02264108434319496, + "weighted_rkd_loss": 0.01648558219894767 + }, + { + "epoch": 4.94908350305499, + "flow_matching_loss": 0.02365035340189934, + "grad_norm": 0.21931295096874237, + "learning_rate": 2.5746548058505873e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008601739024743437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037030338076874613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020906511601060628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.814562863111496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17010, + "weighted_flow_matching_loss": 0.02365035340189934, + "weighted_rkd_loss": 0.017029615677893163 + }, + { + "epoch": 4.951993017166133, + "flow_matching_loss": 0.02190098222345114, + "grad_norm": 0.19996345043182373, + "learning_rate": 2.57407786916909e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00836285469122231, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003734006220474839, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020459715649485587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8143593192100524, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17020, + "weighted_flow_matching_loss": 0.02190098222345114, + "weighted_rkd_loss": 0.01666150465607643 + }, + { + "epoch": 4.954902531277277, + "flow_matching_loss": 0.02245783470571041, + "grad_norm": 0.2686905562877655, + "learning_rate": 2.573500606211095e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00902290246449411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004029192100279033, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02207499695941806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8141556859016419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17030, + "weighted_flow_matching_loss": 0.02245783470571041, + "weighted_rkd_loss": 0.01797230001538992 + }, + { + "epoch": 4.95781204538842, + "flow_matching_loss": 0.025625751726329325, + "grad_norm": 0.24728041887283325, + "learning_rate": 2.57292301715196e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008619740139693022, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003880807966925204, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021120288409292698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8139519691467285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17040, + "weighted_flow_matching_loss": 0.025625751726329325, + "weighted_rkd_loss": 0.01719091832637787 + }, + { + "epoch": 4.9607215594995635, + "flow_matching_loss": 0.02411222495138645, + "grad_norm": 0.245154470205307, + "learning_rate": 2.572345102167142e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009093547146767378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00402432638220489, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02221142090857029, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8137481808662415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17050, + "weighted_flow_matching_loss": 0.02411222495138645, + "weighted_rkd_loss": 0.018074482213705778 + }, + { + "epoch": 4.963631073610707, + "flow_matching_loss": 0.022459249943494797, + "grad_norm": 0.2598135769367218, + "learning_rate": 2.571766861432195e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009291125973686577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004225859884172678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022808111645281313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8135442793369293, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17060, + "weighted_flow_matching_loss": 0.022459249943494797, + "weighted_rkd_loss": 0.018555467016994953 + }, + { + "epoch": 4.96654058772185, + "flow_matching_loss": 0.024559368565678596, + "grad_norm": 0.21287639439105988, + "learning_rate": 2.5711882951227738e-05, + "loss": 0.0424, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009032470313832164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003824955830350518, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021889896504580973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8133403062820435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17070, + "weighted_flow_matching_loss": 0.024559368565678596, + "weighted_rkd_loss": 0.017803913354873656 + }, + { + "epoch": 4.969450101832994, + "flow_matching_loss": 0.021138927340507506, + "grad_norm": 0.2370966076850891, + "learning_rate": 2.570609403414632e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008067853888496757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003461802937090397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019597510620951653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.813136237859726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17080, + "weighted_flow_matching_loss": 0.021138927340507506, + "weighted_rkd_loss": 0.015935382340103388 + }, + { + "epoch": 4.9723596159441374, + "flow_matching_loss": 0.023313952423632146, + "grad_norm": 0.2518409490585327, + "learning_rate": 2.5700301864836212e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010498478636145591, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004659513314254582, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025656470470130442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8129320979118347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17090, + "weighted_flow_matching_loss": 0.023313952423632146, + "weighted_rkd_loss": 0.020856876485049726 + }, + { + "epoch": 4.975269130055281, + "flow_matching_loss": 0.02295031026005745, + "grad_norm": 0.2427411526441574, + "learning_rate": 2.569450644505692e-05, + "loss": 0.0428, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010046701086685062, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043841297272592785, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024477532505989073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8127278625965119, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17100, + "weighted_flow_matching_loss": 0.02295031026005745, + "weighted_rkd_loss": 0.019893550500273704 + }, + { + "epoch": 4.978178644166424, + "flow_matching_loss": 0.02305846679955721, + "grad_norm": 0.2531702220439911, + "learning_rate": 2.5688707776568943e-05, + "loss": 0.0415, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009292260743677616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004139799764379859, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02272432129830122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8125235497951507, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17110, + "weighted_flow_matching_loss": 0.02305846679955721, + "weighted_rkd_loss": 0.018463916145265102 + }, + { + "epoch": 4.981088158277568, + "flow_matching_loss": 0.022847088519483805, + "grad_norm": 0.2595053017139435, + "learning_rate": 2.5682905861133764e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008695317665114999, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003798224125057459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021188859548419713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8123191297054291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17120, + "weighted_flow_matching_loss": 0.022847088519483805, + "weighted_rkd_loss": 0.01721215071156621 + }, + { + "epoch": 4.983997672388711, + "flow_matching_loss": 0.024989489279687405, + "grad_norm": 0.22152332961559296, + "learning_rate": 2.567710070051385e-05, + "loss": 0.0426, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008896336006000639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003832443500868976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021625115536153316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8121146559715271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17130, + "weighted_flow_matching_loss": 0.024989489279687405, + "weighted_rkd_loss": 0.017562089208513498 + }, + { + "epoch": 4.986907186499854, + "flow_matching_loss": 0.02653862927109003, + "grad_norm": 0.22103260457515717, + "learning_rate": 2.5671292296472652e-05, + "loss": 0.0451, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009278495516628026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004319328418932855, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02287631928920746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.811910080909729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17140, + "weighted_flow_matching_loss": 0.02653862927109003, + "weighted_rkd_loss": 0.018573584966361523 + }, + { + "epoch": 4.989816700610998, + "flow_matching_loss": 0.025370113365352154, + "grad_norm": 0.22499413788318634, + "learning_rate": 2.5665480650774616e-05, + "loss": 0.0436, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009184150910004973, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004040881269611418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022409182973206043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8117054283618927, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17150, + "weighted_flow_matching_loss": 0.025370113365352154, + "weighted_rkd_loss": 0.01818961389362812 + }, + { + "epoch": 4.992726214722142, + "flow_matching_loss": 0.024056308530271055, + "grad_norm": 0.2653127610683441, + "learning_rate": 2.565966576518517e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008910632599145174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004197822324931622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022019087336957455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8115006804466247, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17160, + "weighted_flow_matching_loss": 0.024056308530271055, + "weighted_rkd_loss": 0.017868435755372048 + }, + { + "epoch": 4.995635728833285, + "flow_matching_loss": 0.02440711613744497, + "grad_norm": 0.21743154525756836, + "learning_rate": 2.5653847641470706e-05, + "loss": 0.0429, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009319599624723196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004102451750077308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022741650976240636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8112958490848541, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17170, + "weighted_flow_matching_loss": 0.02440711613744497, + "weighted_rkd_loss": 0.01845035320147872 + }, + { + "epoch": 4.998545242944428, + "flow_matching_loss": 0.022621271759271623, + "grad_norm": 0.18811802566051483, + "learning_rate": 2.564802628139863e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008824880421161651, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003771525784395635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021421286277472973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8110909283161163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17180, + "weighted_flow_matching_loss": 0.022621271759271623, + "weighted_rkd_loss": 0.017374627105891705 + }, + { + "epoch": 5.001454757055572, + "flow_matching_loss": 0.022531115170568227, + "grad_norm": 0.2171848714351654, + "learning_rate": 2.5642201686737318e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008486054232344031, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038213131483644247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02079342184588313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8108859062194824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17190, + "weighted_flow_matching_loss": 0.022531115170568227, + "weighted_rkd_loss": 0.016861153207719325 + }, + { + "epoch": 5.004364271166716, + "flow_matching_loss": 0.02338587511330843, + "grad_norm": 0.2491052895784378, + "learning_rate": 2.563637385925612e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009427868202328682, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004160206275992095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02301594279706478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8106808423995971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17200, + "weighted_flow_matching_loss": 0.02338587511330843, + "weighted_rkd_loss": 0.01865855371579528 + }, + { + "epoch": 5.007273785277858, + "flow_matching_loss": 0.0239449055865407, + "grad_norm": 0.2508564293384552, + "learning_rate": 2.5630542800725388e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007933292305096984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035956438863649965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019462228380143644, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8104756712913513, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17210, + "weighted_flow_matching_loss": 0.0239449055865407, + "weighted_rkd_loss": 0.015773620456457138 + }, + { + "epoch": 5.010183299389002, + "flow_matching_loss": 0.024708293564617633, + "grad_norm": 0.22577188909053802, + "learning_rate": 2.562470851291643e-05, + "loss": 0.0437, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00955006442964077, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004391300375573337, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023491429165005684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8102704167366028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17220, + "weighted_flow_matching_loss": 0.024708293564617633, + "weighted_rkd_loss": 0.01903448048979044 + }, + { + "epoch": 5.013092813500146, + "flow_matching_loss": 0.02300002407282591, + "grad_norm": 0.2384738177061081, + "learning_rate": 2.561887099760156e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008616725075989961, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003750052955001593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020983503200113773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.810065072774887, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17230, + "weighted_flow_matching_loss": 0.02300002407282591, + "weighted_rkd_loss": 0.01699798321351409 + }, + { + "epoch": 5.016002327611289, + "flow_matching_loss": 0.027310055308043958, + "grad_norm": 0.22317370772361755, + "learning_rate": 2.5613030256554054e-05, + "loss": 0.0435, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008262414438650011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003522696834988892, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020047525689005852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8098596453666687, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17240, + "weighted_flow_matching_loss": 0.027310055308043958, + "weighted_rkd_loss": 0.016235634963959456 + }, + { + "epoch": 5.018911841722432, + "flow_matching_loss": 0.02420396525412798, + "grad_norm": 0.24945074319839478, + "learning_rate": 2.5607186291548174e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008530978253111244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00371994785964489, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02078190427273512, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8096541285514831, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17250, + "weighted_flow_matching_loss": 0.02420396525412798, + "weighted_rkd_loss": 0.016826233826577663 + }, + { + "epoch": 5.021821355833576, + "flow_matching_loss": 0.02440042980015278, + "grad_norm": 0.21067772805690765, + "learning_rate": 2.5601339104359167e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008596150064840912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037331458413973452, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02092544622719288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8094485104084015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17260, + "weighted_flow_matching_loss": 0.02440042980015278, + "weighted_rkd_loss": 0.01693811733275652 + }, + { + "epoch": 5.024730869944719, + "flow_matching_loss": 0.024891427531838418, + "grad_norm": 0.2557145953178406, + "learning_rate": 2.559548869676325e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008673257427290082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004004295356571675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02135081011801958, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8092428505420685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17270, + "weighted_flow_matching_loss": 0.024891427531838418, + "weighted_rkd_loss": 0.017277923133224248 + }, + { + "epoch": 5.0276403840558626, + "flow_matching_loss": 0.022098685428500176, + "grad_norm": 0.20084825158119202, + "learning_rate": 2.558963507053763e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008183945203199983, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003667001286521554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020034891832619906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8090370833873749, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17280, + "weighted_flow_matching_loss": 0.022098685428500176, + "weighted_rkd_loss": 0.016208968497812747 + }, + { + "epoch": 5.030549898167006, + "flow_matching_loss": 0.024208352901041508, + "grad_norm": 0.23841628432273865, + "learning_rate": 2.5583778227460476e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008728564530611039, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037611756939440965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021218304708600044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8088312327861786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17290, + "weighted_flow_matching_loss": 0.024208352901041508, + "weighted_rkd_loss": 0.01716199079528451 + }, + { + "epoch": 5.03345941227815, + "flow_matching_loss": 0.02404565066099167, + "grad_norm": 0.19904057681560516, + "learning_rate": 2.5577918169310943e-05, + "loss": 0.0433, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009773792652413248, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004304923163726926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023852508701384067, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8086252987384797, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17300, + "weighted_flow_matching_loss": 0.02404565066099167, + "weighted_rkd_loss": 0.01928776688873768 + }, + { + "epoch": 5.036368926389293, + "flow_matching_loss": 0.02227233238518238, + "grad_norm": 0.2302885353565216, + "learning_rate": 2.5572054897869167e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008661247277632355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004012326919473707, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02133482126519084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8084192752838135, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17310, + "weighted_flow_matching_loss": 0.02227233238518238, + "weighted_rkd_loss": 0.017247491888701914 + }, + { + "epoch": 5.0392784405004365, + "flow_matching_loss": 0.022804592922329902, + "grad_norm": 0.2235785722732544, + "learning_rate": 2.556618841491625e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008369478629902005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036312368931248782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02037019394338131, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8082131624221802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17320, + "weighted_flow_matching_loss": 0.022804592922329902, + "weighted_rkd_loss": 0.016463469248265028 + }, + { + "epoch": 5.04218795461158, + "flow_matching_loss": 0.022045595198869707, + "grad_norm": 0.21860581636428833, + "learning_rate": 2.5560318722234276e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008332293946295977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00367557464633137, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020340162515640258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.808007001876831, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17330, + "weighted_flow_matching_loss": 0.022045595198869707, + "weighted_rkd_loss": 0.016434991639107464 + }, + { + "epoch": 5.045097468722723, + "flow_matching_loss": 0.02490660659968853, + "grad_norm": 0.24230284988880157, + "learning_rate": 2.5554445821606306e-05, + "loss": 0.0452, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010238240007311106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004619496967643499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025095976889133453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8078006982803345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17340, + "weighted_flow_matching_loss": 0.02490660659968853, + "weighted_rkd_loss": 0.020272557623684406 + }, + { + "epoch": 5.048006982833867, + "flow_matching_loss": 0.0246059974655509, + "grad_norm": 0.25128960609436035, + "learning_rate": 2.554856971481637e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008252645423635841, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003796412446536124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020301703363656998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8075943470001221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17350, + "weighted_flow_matching_loss": 0.0246059974655509, + "weighted_rkd_loss": 0.01639551520347595 + }, + { + "epoch": 5.05091649694501, + "flow_matching_loss": 0.02274361215531826, + "grad_norm": 0.19248022139072418, + "learning_rate": 2.554269040364947e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008544519077986479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037160108564421534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020805048942565917, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8073879063129425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17360, + "weighted_flow_matching_loss": 0.02274361215531826, + "weighted_rkd_loss": 0.016797767672687768 + }, + { + "epoch": 5.053826011056153, + "flow_matching_loss": 0.025077789090573788, + "grad_norm": 0.2183360606431961, + "learning_rate": 2.553680788989159e-05, + "loss": 0.0442, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009655051166191697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0043357212096452715, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023645823635160924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8071813762187958, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17370, + "weighted_flow_matching_loss": 0.025077789090573788, + "weighted_rkd_loss": 0.019086404424160718 + }, + { + "epoch": 5.056735525167297, + "flow_matching_loss": 0.02216305136680603, + "grad_norm": 0.2353007048368454, + "learning_rate": 2.5530922175329677e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008904435206204653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004071591072715819, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021880461648106576, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8069747626781464, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17380, + "weighted_flow_matching_loss": 0.02216305136680603, + "weighted_rkd_loss": 0.017657027672976257 + }, + { + "epoch": 5.059645039278441, + "flow_matching_loss": 0.023834277875721455, + "grad_norm": 0.26777517795562744, + "learning_rate": 2.5525033261751667e-05, + "loss": 0.044, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010066457372158766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004881280101835727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025014195032417773, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8067680656909942, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17390, + "weighted_flow_matching_loss": 0.023834277875721455, + "weighted_rkd_loss": 0.020180896762758492 + }, + { + "epoch": 5.0625545533895835, + "flow_matching_loss": 0.022284378111362458, + "grad_norm": 0.2150842696428299, + "learning_rate": 2.5519141150946445e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009333125222474337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004183082235977053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02284933254122734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8065612852573395, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17400, + "weighted_flow_matching_loss": 0.022284378111362458, + "weighted_rkd_loss": 0.018429469689726828 + }, + { + "epoch": 5.065464067500727, + "flow_matching_loss": 0.02246919311583042, + "grad_norm": 0.21106590330600739, + "learning_rate": 2.551324584470389e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008834381168708206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038560183020308614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021524780429899693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.806354421377182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17410, + "weighted_flow_matching_loss": 0.02246919311583042, + "weighted_rkd_loss": 0.01735660955309868 + }, + { + "epoch": 5.068373581611871, + "flow_matching_loss": 0.022053420171141626, + "grad_norm": 0.2555798292160034, + "learning_rate": 2.5507347344814825e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008664632588624955, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003648749436251819, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020978014543652536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8061474800109864, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17420, + "weighted_flow_matching_loss": 0.022053420171141626, + "weighted_rkd_loss": 0.016911369748413563 + }, + { + "epoch": 5.071283095723015, + "flow_matching_loss": 0.024127626977860927, + "grad_norm": 0.22440652549266815, + "learning_rate": 2.5501445653071072e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008651197515428066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003968255198560655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02127065006643534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.805940443277359, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17430, + "weighted_flow_matching_loss": 0.024127626977860927, + "weighted_rkd_loss": 0.017142876144498588 + }, + { + "epoch": 5.074192609834157, + "flow_matching_loss": 0.024244493618607522, + "grad_norm": 0.24446509778499603, + "learning_rate": 2.5495540771265406e-05, + "loss": 0.0418, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008955556293949484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038766534300521014, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02178776599466801, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8057333290576935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17440, + "weighted_flow_matching_loss": 0.024244493618607522, + "weighted_rkd_loss": 0.01755513558164239 + }, + { + "epoch": 5.077102123945301, + "flow_matching_loss": 0.023954248800873757, + "grad_norm": 0.23605936765670776, + "learning_rate": 2.5489632701191565e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00836404412984848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003914800612255931, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020642888732254505, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8055261254310608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17450, + "weighted_flow_matching_loss": 0.023954248800873757, + "weighted_rkd_loss": 0.016628336627036333 + }, + { + "epoch": 5.080011638056445, + "flow_matching_loss": 0.023271752893924712, + "grad_norm": 0.2549858093261719, + "learning_rate": 2.5483721444644276e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008781855693086982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003945876681245864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021509587951004504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8053188502788544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17460, + "weighted_flow_matching_loss": 0.023271752893924712, + "weighted_rkd_loss": 0.017321994155645372 + }, + { + "epoch": 5.082921152167588, + "flow_matching_loss": 0.024935956485569476, + "grad_norm": 0.2319760024547577, + "learning_rate": 2.5477807003419222e-05, + "loss": 0.0446, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009811158105731011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004861324210651219, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024483640491962434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8051114678382874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17470, + "weighted_flow_matching_loss": 0.024935956485569476, + "weighted_rkd_loss": 0.01971211554482579 + }, + { + "epoch": 5.085830666278731, + "flow_matching_loss": 0.02316679898649454, + "grad_norm": 0.24572904407978058, + "learning_rate": 2.547188937931305e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0084288468118757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037325148936361075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02059020847082138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8049040317535401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17480, + "weighted_flow_matching_loss": 0.02316679898649454, + "weighted_rkd_loss": 0.01657314896583557 + }, + { + "epoch": 5.088740180389875, + "flow_matching_loss": 0.022884780913591383, + "grad_norm": 0.23203209042549133, + "learning_rate": 2.5465968574123372e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008988014794886113, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003884836775250733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021860866248607634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8046965003013611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17490, + "weighted_flow_matching_loss": 0.022884780913591383, + "weighted_rkd_loss": 0.017591345962136982 + }, + { + "epoch": 5.091649694501018, + "flow_matching_loss": 0.024508911557495593, + "grad_norm": 0.22127598524093628, + "learning_rate": 2.5460044589648787e-05, + "loss": 0.0425, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009170581866055728, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004047972266562283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02238913606852293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8044888794422149, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17500, + "weighted_flow_matching_loss": 0.024508911557495593, + "weighted_rkd_loss": 0.018011933099478482 + }, + { + "epoch": 5.094559208612162, + "flow_matching_loss": 0.02244962453842163, + "grad_norm": 0.22215041518211365, + "learning_rate": 2.5454117427688833e-05, + "loss": 0.0398, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008792951935902238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039773165248334404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021563220024108886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8042811751365662, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17510, + "weighted_flow_matching_loss": 0.02244962453842163, + "weighted_rkd_loss": 0.017342964373528957 + }, + { + "epoch": 5.097468722723305, + "flow_matching_loss": 0.022745152562856676, + "grad_norm": 0.23019683361053467, + "learning_rate": 2.5448187090044033e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009069428732618689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003943497268483043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022082354687154294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8040733933448792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17520, + "weighted_flow_matching_loss": 0.022745152562856676, + "weighted_rkd_loss": 0.01775581268593669 + }, + { + "epoch": 5.100378236834449, + "flow_matching_loss": 0.023490998335182665, + "grad_norm": 0.2266864776611328, + "learning_rate": 2.544225357851586e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00793421003036201, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034935667645186186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01936198687180877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8038655281066894, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17530, + "weighted_flow_matching_loss": 0.023490998335182665, + "weighted_rkd_loss": 0.015564348921179772 + }, + { + "epoch": 5.103287750945592, + "flow_matching_loss": 0.021801259368658066, + "grad_norm": 0.18596680462360382, + "learning_rate": 2.543631689490676e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00984604712575674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004612387903034687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.024304482340812682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8036575615406036, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17540, + "weighted_flow_matching_loss": 0.021801259368658066, + "weighted_rkd_loss": 0.019532453361898662 + }, + { + "epoch": 5.1061972650567355, + "flow_matching_loss": 0.022846202924847604, + "grad_norm": 0.24770934879779816, + "learning_rate": 2.543037704102015e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009055560873821377, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004165442683734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022276564314961434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8034495353698731, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17550, + "weighted_flow_matching_loss": 0.022846202924847604, + "weighted_rkd_loss": 0.017898003198206426 + }, + { + "epoch": 5.109106779167879, + "flow_matching_loss": 0.021585653349757195, + "grad_norm": 0.25562286376953125, + "learning_rate": 2.5424434018660386e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009440084919333458, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003979119379073382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022859289310872555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8032414138317108, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17560, + "weighted_flow_matching_loss": 0.021585653349757195, + "weighted_rkd_loss": 0.01836145594716072 + }, + { + "epoch": 5.112016293279022, + "flow_matching_loss": 0.02366990055888891, + "grad_norm": 0.22682133316993713, + "learning_rate": 2.541848782963281e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00899235885590315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004031690419651568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022016408573836087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8030332267284394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17570, + "weighted_flow_matching_loss": 0.02366990055888891, + "weighted_rkd_loss": 0.01767983641475439 + }, + { + "epoch": 5.114925807390166, + "flow_matching_loss": 0.023158687725663187, + "grad_norm": 0.22367039322853088, + "learning_rate": 2.5412538475743714e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008829391468316317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037504229927435516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021409205719828607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8028249382972718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17580, + "weighted_flow_matching_loss": 0.023158687725663187, + "weighted_rkd_loss": 0.01718774242326617 + }, + { + "epoch": 5.1178353215013095, + "flow_matching_loss": 0.024089862965047358, + "grad_norm": 0.19140999019145966, + "learning_rate": 2.540658595880036e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008711684495210648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004028051276691258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02145142015069723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8026165664196014, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17590, + "weighted_flow_matching_loss": 0.024089862965047358, + "weighted_rkd_loss": 0.017217225674539805 + }, + { + "epoch": 5.120744835612452, + "flow_matching_loss": 0.021939120814204217, + "grad_norm": 0.23181213438510895, + "learning_rate": 2.540063028061096e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00870424583554268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004032667959108949, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021441159769892693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8024081230163574, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17600, + "weighted_flow_matching_loss": 0.021939120814204217, + "weighted_rkd_loss": 0.017204565182328223 + }, + { + "epoch": 5.123654349723596, + "flow_matching_loss": 0.023453850671648978, + "grad_norm": 0.24468107521533966, + "learning_rate": 2.5394671442984692e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008845825307071209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003951995400711894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02164364606142044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8021995902061463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17610, + "weighted_flow_matching_loss": 0.023453850671648978, + "weighted_rkd_loss": 0.017362600937485696 + }, + { + "epoch": 5.12656386383474, + "flow_matching_loss": 0.022457433864474297, + "grad_norm": 0.2510529160499573, + "learning_rate": 2.5388709447731696e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008596922550350428, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036693196976557373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020863164961338044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8019909739494324, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17620, + "weighted_flow_matching_loss": 0.022457433864474297, + "weighted_rkd_loss": 0.01673205066472292 + }, + { + "epoch": 5.129473377945883, + "flow_matching_loss": 0.023210296221077443, + "grad_norm": 0.2012769877910614, + "learning_rate": 2.5382744296663068e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008704144926741719, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003844266710802913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02125255633145571, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8017822742462158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17630, + "weighted_flow_matching_loss": 0.023210296221077443, + "weighted_rkd_loss": 0.01703994758427143 + }, + { + "epoch": 5.132382892057026, + "flow_matching_loss": 0.02224019728600979, + "grad_norm": 0.20561924576759338, + "learning_rate": 2.5376775991590865e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00824681781232357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037161886924877764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020209824480116368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8015734851360321, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17640, + "weighted_flow_matching_loss": 0.02224019728600979, + "weighted_rkd_loss": 0.016199673898518084 + }, + { + "epoch": 5.13529240616817, + "flow_matching_loss": 0.02508531641215086, + "grad_norm": 0.2027236819267273, + "learning_rate": 2.5370804534328097e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00784429656341672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00339953931979835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01908813267946243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8013646304607391, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17650, + "weighted_flow_matching_loss": 0.02508531641215086, + "weighted_rkd_loss": 0.01529658231884241 + }, + { + "epoch": 5.138201920279314, + "flow_matching_loss": 0.0218027513474226, + "grad_norm": 0.19880013167858124, + "learning_rate": 2.5364829926688736e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00958690014667809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041089030681177975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.023282703198492526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8011556804180145, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17660, + "weighted_flow_matching_loss": 0.0218027513474226, + "weighted_rkd_loss": 0.018653081078082324 + }, + { + "epoch": 5.1411114343904565, + "flow_matching_loss": 0.023867265693843363, + "grad_norm": 0.25435495376586914, + "learning_rate": 2.5358852170487713e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008260464202612639, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003707363991998136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02022829242050648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8009466469287873, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17670, + "weighted_flow_matching_loss": 0.023867265693843363, + "weighted_rkd_loss": 0.016201686952263115 + }, + { + "epoch": 5.1440209485016, + "flow_matching_loss": 0.024343204312026502, + "grad_norm": 0.19577756524085999, + "learning_rate": 2.5352871267540906e-05, + "loss": 0.0402, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008159617241472007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035147820366546513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01983401644974947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8007375419139862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17680, + "weighted_flow_matching_loss": 0.024343204312026502, + "weighted_rkd_loss": 0.015881866123527287 + }, + { + "epoch": 5.146930462612744, + "flow_matching_loss": 0.022754352167248727, + "grad_norm": 0.2000618726015091, + "learning_rate": 2.534688721966516e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007513487106189131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033436892554163933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0183706633746624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8005283534526825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17690, + "weighted_flow_matching_loss": 0.022754352167248727, + "weighted_rkd_loss": 0.014706212747842073 + }, + { + "epoch": 5.149839976723887, + "flow_matching_loss": 0.02172532882541418, + "grad_norm": 0.21418599784374237, + "learning_rate": 2.5340900028678276e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009086890704929828, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003863631607964635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022037412971258163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8003190755844116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17700, + "weighted_flow_matching_loss": 0.02172532882541418, + "weighted_rkd_loss": 0.017636989802122117 + }, + { + "epoch": 5.15274949083503, + "flow_matching_loss": 0.021729161962866783, + "grad_norm": 0.22408133745193481, + "learning_rate": 2.5334909696398998e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008371939742937684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037964319111779333, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02054031128063798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.8001097083091736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17710, + "weighted_flow_matching_loss": 0.021729161962866783, + "weighted_rkd_loss": 0.016434413474053146 + }, + { + "epoch": 5.155659004946174, + "flow_matching_loss": 0.022272896580398084, + "grad_norm": 0.20224983990192413, + "learning_rate": 2.5328916224647027e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008107835613191127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003210661350749433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01942633241415024, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7999002695083618, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17720, + "weighted_flow_matching_loss": 0.022272896580398084, + "weighted_rkd_loss": 0.015539239998906851 + }, + { + "epoch": 5.158568519057318, + "flow_matching_loss": 0.024710582941770552, + "grad_norm": 0.22112110257148743, + "learning_rate": 2.532291961524303e-05, + "loss": 0.0422, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008855891972780227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004182375455275178, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021894159354269506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7996907413005829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17730, + "weighted_flow_matching_loss": 0.024710582941770552, + "weighted_rkd_loss": 0.017508513294160366 + }, + { + "epoch": 5.161478033168461, + "flow_matching_loss": 0.02149718180298805, + "grad_norm": 0.22814032435417175, + "learning_rate": 2.531691987000861e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007681037951260805, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003368754521943629, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018730830401182175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7994811356067657, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17740, + "weighted_flow_matching_loss": 0.02149718180298805, + "weighted_rkd_loss": 0.014975038822740316 + }, + { + "epoch": 5.164387547279604, + "flow_matching_loss": 0.023757705837488173, + "grad_norm": 0.22807972133159637, + "learning_rate": 2.5310916990766343e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008252329006791116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003956331522203982, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02046098969876766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7992714524269104, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17750, + "weighted_flow_matching_loss": 0.023757705837488173, + "weighted_rkd_loss": 0.016353819519281387 + }, + { + "epoch": 5.167297061390748, + "flow_matching_loss": 0.02377532348036766, + "grad_norm": 0.23788222670555115, + "learning_rate": 2.5304910979339734e-05, + "loss": 0.0419, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009249112103134394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004141653678379953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022639877907931805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7990616858005524, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17760, + "weighted_flow_matching_loss": 0.02377532348036766, + "weighted_rkd_loss": 0.018090648390352727 + }, + { + "epoch": 5.170206575501891, + "flow_matching_loss": 0.022907667420804502, + "grad_norm": 0.28587859869003296, + "learning_rate": 2.5298901837553255e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008804294234141707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003742137085646391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021350725553929806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7988518297672271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17770, + "weighted_flow_matching_loss": 0.022907667420804502, + "weighted_rkd_loss": 0.01705600181594491 + }, + { + "epoch": 5.173116089613035, + "flow_matching_loss": 0.023200668580830097, + "grad_norm": 0.23189008235931396, + "learning_rate": 2.5292889567232326e-05, + "loss": 0.0406, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0089012801181525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004030387592501938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02183294780552387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7986419022083282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17780, + "weighted_flow_matching_loss": 0.023200668580830097, + "weighted_rkd_loss": 0.017436678800731898 + }, + { + "epoch": 5.176025603724178, + "flow_matching_loss": 0.023142746090888976, + "grad_norm": 0.20511958003044128, + "learning_rate": 2.528687417020331e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008463913714513183, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003858427982777357, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02078625550493598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7984318852424621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17790, + "weighted_flow_matching_loss": 0.023142746090888976, + "weighted_rkd_loss": 0.01659640995785594 + }, + { + "epoch": 5.178935117835321, + "flow_matching_loss": 0.02466818057000637, + "grad_norm": 0.26184841990470886, + "learning_rate": 2.5280855648293536e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008411862328648568, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036475600209087133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020471284724771978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7982217907905579, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17800, + "weighted_flow_matching_loss": 0.02466818057000637, + "weighted_rkd_loss": 0.016340619418770075 + }, + { + "epoch": 5.181844631946465, + "flow_matching_loss": 0.023558144457638264, + "grad_norm": 0.2225814312696457, + "learning_rate": 2.5274834003331266e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008370588487014174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003849671734496951, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020590848475694656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7980116128921508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17810, + "weighted_flow_matching_loss": 0.023558144457638264, + "weighted_rkd_loss": 0.016431800741702317 + }, + { + "epoch": 5.1847541460576085, + "flow_matching_loss": 0.022625327669084074, + "grad_norm": 0.3783414959907532, + "learning_rate": 2.5268809237145717e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009157832898199558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004139399458654225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02245506551116705, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7978013455867767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17820, + "weighted_flow_matching_loss": 0.022625327669084074, + "weighted_rkd_loss": 0.017914486676454545 + }, + { + "epoch": 5.187663660168752, + "flow_matching_loss": 0.02140819504857063, + "grad_norm": 0.22375032305717468, + "learning_rate": 2.5262781351567052e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009113414585590363, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004115087445825338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02234191671013832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7975910007953644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17830, + "weighted_flow_matching_loss": 0.02140819504857063, + "weighted_rkd_loss": 0.01781972050666809 + }, + { + "epoch": 5.190573174279895, + "flow_matching_loss": 0.02290455438196659, + "grad_norm": 0.2666492462158203, + "learning_rate": 2.525675034842638e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008329584868624807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003771611419506371, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02043078113347292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7973805844783783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17840, + "weighted_flow_matching_loss": 0.02290455438196659, + "weighted_rkd_loss": 0.01629101037979126 + }, + { + "epoch": 5.193482688391039, + "flow_matching_loss": 0.021616287529468536, + "grad_norm": 0.2754433751106262, + "learning_rate": 2.5250716229555774e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008623158512637018, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003962919535115361, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021209236606955527, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7971700727939606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17850, + "weighted_flow_matching_loss": 0.021616287529468536, + "weighted_rkd_loss": 0.01690727910026908 + }, + { + "epoch": 5.1963922025021825, + "flow_matching_loss": 0.021336783468723298, + "grad_norm": 0.18840746581554413, + "learning_rate": 2.524467899678823e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008653782540932298, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038430779008194805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02115064300596714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7969594895839691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17860, + "weighted_flow_matching_loss": 0.021336783468723298, + "weighted_rkd_loss": 0.01685627829283476 + }, + { + "epoch": 5.199301716613325, + "flow_matching_loss": 0.022779695875942706, + "grad_norm": 0.20760367810726166, + "learning_rate": 2.52386386519577e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008294215984642505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037798037519678473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020368235558271407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7967488169670105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17870, + "weighted_flow_matching_loss": 0.022779695875942706, + "weighted_rkd_loss": 0.016228389088064434 + }, + { + "epoch": 5.202211230724469, + "flow_matching_loss": 0.021964528784155845, + "grad_norm": 0.21644651889801025, + "learning_rate": 2.5232595196899086e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008578954823315143, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00374551578424871, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02090342566370964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7965380668640136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17880, + "weighted_flow_matching_loss": 0.021964528784155845, + "weighted_rkd_loss": 0.016650320403277873 + }, + { + "epoch": 5.205120744835613, + "flow_matching_loss": 0.021595644019544124, + "grad_norm": 0.2302667796611786, + "learning_rate": 2.5226548633448224e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008960559964179993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003951184707693756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021872304566204547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7963272392749786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17890, + "weighted_flow_matching_loss": 0.021595644019544124, + "weighted_rkd_loss": 0.01741758417338133 + }, + { + "epoch": 5.2080302589467555, + "flow_matching_loss": 0.021545295603573322, + "grad_norm": 0.19075849652290344, + "learning_rate": 2.5220498963441906e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007890777057036757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034160675946623086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01919762175530195, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7961163341999054, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17900, + "weighted_flow_matching_loss": 0.021545295603573322, + "weighted_rkd_loss": 0.015283550601452589 + }, + { + "epoch": 5.210939773057899, + "flow_matching_loss": 0.02169695943593979, + "grad_norm": 0.183047354221344, + "learning_rate": 2.521444618871786e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008247511787340045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034823231166228653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019977346993982792, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.795905339717865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17910, + "weighted_flow_matching_loss": 0.02169695943593979, + "weighted_rkd_loss": 0.01590014798566699 + }, + { + "epoch": 5.213849287169043, + "flow_matching_loss": 0.022708439826965333, + "grad_norm": 0.2166557013988495, + "learning_rate": 2.5208390311114758e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008285873150452972, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003725925041362643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02029767157509923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7956942617893219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17920, + "weighted_flow_matching_loss": 0.022708439826965333, + "weighted_rkd_loss": 0.016150719579309226 + }, + { + "epoch": 5.216758801280186, + "flow_matching_loss": 0.022210749424993993, + "grad_norm": 0.2409941852092743, + "learning_rate": 2.520233133247221e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008433177229017019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003661052929237485, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02052740752696991, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7954831004142762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17930, + "weighted_flow_matching_loss": 0.022210749424993993, + "weighted_rkd_loss": 0.016329023241996764 + }, + { + "epoch": 5.2196683153913295, + "flow_matching_loss": 0.02294923197478056, + "grad_norm": 0.2848956286907196, + "learning_rate": 2.519626925463079e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008820628467947244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038762365002185106, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021517493203282355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7952718675136566, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17940, + "weighted_flow_matching_loss": 0.02294923197478056, + "weighted_rkd_loss": 0.017112171556800605 + }, + { + "epoch": 5.222577829502473, + "flow_matching_loss": 0.021521932259202004, + "grad_norm": 0.22968818247318268, + "learning_rate": 2.519020407943198e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008966898312792182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038674758514389397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021801272593438625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7950605511665344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17950, + "weighted_flow_matching_loss": 0.021521932259202004, + "weighted_rkd_loss": 0.01733334707096219 + }, + { + "epoch": 5.225487343613617, + "flow_matching_loss": 0.02437053471803665, + "grad_norm": 0.23137883841991425, + "learning_rate": 2.518413580871823e-05, + "loss": 0.0417, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008991415798664092, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003811466135084629, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021794297732412815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7948491513729096, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17960, + "weighted_flow_matching_loss": 0.02437053471803665, + "weighted_rkd_loss": 0.0173231516033411 + }, + { + "epoch": 5.22839685772476, + "flow_matching_loss": 0.022704216465353966, + "grad_norm": 0.2496780902147293, + "learning_rate": 2.5178064444332922e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009092592680826784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004014275944791734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022199461236596107, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7946376740932465, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17970, + "weighted_flow_matching_loss": 0.022704216465353966, + "weighted_rkd_loss": 0.01764043848961592 + }, + { + "epoch": 5.231306371835903, + "flow_matching_loss": 0.022548787109553815, + "grad_norm": 0.24820461869239807, + "learning_rate": 2.5171989988120368e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00788429225794971, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003548761736601591, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019317346066236495, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7944261133670807, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17980, + "weighted_flow_matching_loss": 0.022548787109553815, + "weighted_rkd_loss": 0.015346112847328185 + }, + { + "epoch": 5.234215885947047, + "flow_matching_loss": 0.02377572227269411, + "grad_norm": 0.2351902425289154, + "learning_rate": 2.5165912441925832e-05, + "loss": 0.0414, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009101884020492435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004025448206812143, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022229216434061526, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7942144751548768, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 17990, + "weighted_flow_matching_loss": 0.02377572227269411, + "weighted_rkd_loss": 0.017654789518564938 + }, + { + "epoch": 5.23712540005819, + "flow_matching_loss": 0.022640338540077208, + "grad_norm": 0.1927165687084198, + "learning_rate": 2.515983180759551e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008023612014949322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033332510385662316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01938047483563423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7940027475357055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18000, + "weighted_flow_matching_loss": 0.022640338540077208, + "weighted_rkd_loss": 0.015388167928904295 + }, + { + "epoch": 5.240034914169334, + "flow_matching_loss": 0.023934711702167988, + "grad_norm": 0.2169511914253235, + "learning_rate": 2.515374808697654e-05, + "loss": 0.0445, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.010470757307484747, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004937231866642832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.025878746807575227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7937909483909606, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18010, + "weighted_flow_matching_loss": 0.023934711702167988, + "weighted_rkd_loss": 0.020542167034000157 + }, + { + "epoch": 5.242944428280477, + "flow_matching_loss": 0.022781397961080075, + "grad_norm": 0.1904628872871399, + "learning_rate": 2.5147661281916996e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007907206611707806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00354683434125036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019361247774213552, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7935790657997132, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18020, + "weighted_flow_matching_loss": 0.022781397961080075, + "weighted_rkd_loss": 0.015364779066294432 + }, + { + "epoch": 5.24585394239162, + "flow_matching_loss": 0.02358666732907295, + "grad_norm": 0.20701809227466583, + "learning_rate": 2.5141571394265892e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00798977999947965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003420678176917136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01940023824572563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7933670997619628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18030, + "weighted_flow_matching_loss": 0.02358666732907295, + "weighted_rkd_loss": 0.015391362272202969 + }, + { + "epoch": 5.248763456502764, + "flow_matching_loss": 0.02155309058725834, + "grad_norm": 0.23194454610347748, + "learning_rate": 2.513547842587317e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008736273366957903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003851757454685867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021324304305016993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7931550562381744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18040, + "weighted_flow_matching_loss": 0.02155309058725834, + "weighted_rkd_loss": 0.016913400404155253 + }, + { + "epoch": 5.251672970613908, + "flow_matching_loss": 0.021728016808629035, + "grad_norm": 0.20606698095798492, + "learning_rate": 2.5129382378589708e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008325768634676933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003731781244277954, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020383318699896336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7929429411888123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18050, + "weighted_flow_matching_loss": 0.021728016808629035, + "weighted_rkd_loss": 0.016162863560020925 + }, + { + "epoch": 5.254582484725051, + "flow_matching_loss": 0.024037979170680047, + "grad_norm": 0.20127660036087036, + "learning_rate": 2.512328325426733e-05, + "loss": 0.0408, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008721281122416259, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003746781824156642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021189344115555286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7927307307720184, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18060, + "weighted_flow_matching_loss": 0.024037979170680047, + "weighted_rkd_loss": 0.016797459963709117 + }, + { + "epoch": 5.257491998836194, + "flow_matching_loss": 0.022677868232131006, + "grad_norm": 0.22938275337219238, + "learning_rate": 2.5117181054758798e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007802883489057421, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035475815180689096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019153348449617625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7925184488296508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18070, + "weighted_flow_matching_loss": 0.022677868232131006, + "weighted_rkd_loss": 0.015179375372827052 + }, + { + "epoch": 5.260401512947338, + "flow_matching_loss": 0.022582216560840605, + "grad_norm": 0.196349635720253, + "learning_rate": 2.5111075781917783e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007212080899626017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003219670942053199, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017643832601606846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7923060774803161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18080, + "weighted_flow_matching_loss": 0.022582216560840605, + "weighted_rkd_loss": 0.013979411777108907 + }, + { + "epoch": 5.2633110270584815, + "flow_matching_loss": 0.02310599572956562, + "grad_norm": 0.21869631111621857, + "learning_rate": 2.510496743759892e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008271904475986958, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036218906519934536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020165699627250434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7920936346054077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18090, + "weighted_flow_matching_loss": 0.02310599572956562, + "weighted_rkd_loss": 0.01597325233742595 + }, + { + "epoch": 5.266220541169624, + "flow_matching_loss": 0.02074716314673424, + "grad_norm": 0.25107064843177795, + "learning_rate": 2.509885602365775e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008732331451028585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003729660902172327, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021194323617964982, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7918811142444611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18100, + "weighted_flow_matching_loss": 0.02074716314673424, + "weighted_rkd_loss": 0.0167833567596972 + }, + { + "epoch": 5.269130055280768, + "flow_matching_loss": 0.02456952016800642, + "grad_norm": 0.2224409282207489, + "learning_rate": 2.509274154195076e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008782106917351484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003974731988273561, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02153894565999508, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7916685044765472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18110, + "weighted_flow_matching_loss": 0.02456952016800642, + "weighted_rkd_loss": 0.01705172089859843 + }, + { + "epoch": 5.272039569391912, + "flow_matching_loss": 0.022555418498814105, + "grad_norm": 0.29402583837509155, + "learning_rate": 2.5086623994335383e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008479220001026988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033322163159027695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020290656201541423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7914558231830597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18120, + "weighted_flow_matching_loss": 0.022555418498814105, + "weighted_rkd_loss": 0.016059031430631875 + }, + { + "epoch": 5.274949083503055, + "flow_matching_loss": 0.022431529127061367, + "grad_norm": 0.2660990357398987, + "learning_rate": 2.5080503382669953e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00851333113387227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003635840932838619, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020662502851337194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7912430465221405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18130, + "weighted_flow_matching_loss": 0.022431529127061367, + "weighted_rkd_loss": 0.016349044535309075 + }, + { + "epoch": 5.277858597614198, + "flow_matching_loss": 0.024511811323463917, + "grad_norm": 0.2076573520898819, + "learning_rate": 2.507437970881376e-05, + "loss": 0.0413, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008607044909149408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004026480531319976, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021240570396184922, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7910302162170411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18140, + "weighted_flow_matching_loss": 0.024511811323463917, + "weighted_rkd_loss": 0.016801979392766953 + }, + { + "epoch": 5.280768111725342, + "flow_matching_loss": 0.021480192430317402, + "grad_norm": 0.2942698299884796, + "learning_rate": 2.5068252974627003e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00849228766746819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038997943978756667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020884369499981403, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7908172905445099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18150, + "weighted_flow_matching_loss": 0.021480192430317402, + "weighted_rkd_loss": 0.01651556808501482 + }, + { + "epoch": 5.283677625836486, + "flow_matching_loss": 0.025301047787070274, + "grad_norm": 0.25927382707595825, + "learning_rate": 2.5062123181970834e-05, + "loss": 0.042, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008557847049087287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004017567192204296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021133260894566776, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7906042814254761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18160, + "weighted_flow_matching_loss": 0.025301047787070274, + "weighted_rkd_loss": 0.016707900539040567 + }, + { + "epoch": 5.2865871399476285, + "flow_matching_loss": 0.027527841739356518, + "grad_norm": 0.18562205135822296, + "learning_rate": 2.5055990332707316e-05, + "loss": 0.0434, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008318343525752425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034228562843054535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020059543661773205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7903912007808686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18170, + "weighted_flow_matching_loss": 0.027527841739356518, + "weighted_rkd_loss": 0.01585496384650469 + }, + { + "epoch": 5.289496654058772, + "flow_matching_loss": 0.021559149213135242, + "grad_norm": 0.20161253213882446, + "learning_rate": 2.5049854428699444e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00835974202491343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036657907301560046, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02038527512922883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7901780366897583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18180, + "weighted_flow_matching_loss": 0.021559149213135242, + "weighted_rkd_loss": 0.016108006238937378 + }, + { + "epoch": 5.292406168169916, + "flow_matching_loss": 0.02175204027444124, + "grad_norm": 0.19322167336940765, + "learning_rate": 2.5043715471811158e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008421357162296772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037207589484751225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020563473366200923, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7899647951126099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18190, + "weighted_flow_matching_loss": 0.02175204027444124, + "weighted_rkd_loss": 0.01624444480985403 + }, + { + "epoch": 5.295315682281059, + "flow_matching_loss": 0.023538137413561346, + "grad_norm": 0.18660011887550354, + "learning_rate": 2.5037573463907296e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008181644370779395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00369703252799809, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02006032131612301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7897514700889587, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18200, + "weighted_flow_matching_loss": 0.023538137413561346, + "weighted_rkd_loss": 0.015842737909406424 + }, + { + "epoch": 5.2982251963922025, + "flow_matching_loss": 0.02428104653954506, + "grad_norm": 0.2188146859407425, + "learning_rate": 2.5031428406853637e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008439875300973653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003910700581036508, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02079045120626688, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7895380675792694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18210, + "weighted_flow_matching_loss": 0.02428104653954506, + "weighted_rkd_loss": 0.016414938122034074 + }, + { + "epoch": 5.301134710503346, + "flow_matching_loss": 0.023703180998563767, + "grad_norm": 0.19050049781799316, + "learning_rate": 2.5025280302516897e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0081319950055331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003649381920695305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019913372211158277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7893245875835418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18220, + "weighted_flow_matching_loss": 0.023703180998563767, + "weighted_rkd_loss": 0.015718244574964048 + }, + { + "epoch": 5.304044224614489, + "flow_matching_loss": 0.021979509573429822, + "grad_norm": 0.22174853086471558, + "learning_rate": 2.5019129152764698e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008862546132877468, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036732192151248456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021398311667144298, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7891110241413116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18230, + "weighted_flow_matching_loss": 0.021979509573429822, + "weighted_rkd_loss": 0.01688573360443115 + }, + { + "epoch": 5.306953738725633, + "flow_matching_loss": 0.022874185629189015, + "grad_norm": 0.178588405251503, + "learning_rate": 2.5012974959465612e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00821702079847455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003690559882670641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02012460120022297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7888973951339722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18240, + "weighted_flow_matching_loss": 0.022874185629189015, + "weighted_rkd_loss": 0.015876255836337803 + }, + { + "epoch": 5.309863252836776, + "flow_matching_loss": 0.02283060997724533, + "grad_norm": 0.21986320614814758, + "learning_rate": 2.5006817724489105e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0080805120524019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037001378601416945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01986116198822856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7886836647987365, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18250, + "weighted_flow_matching_loss": 0.02283060997724533, + "weighted_rkd_loss": 0.01566405426710844 + }, + { + "epoch": 5.31277276694792, + "flow_matching_loss": 0.02342719603329897, + "grad_norm": 0.2035953849554062, + "learning_rate": 2.500065744970559e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00882619279436767, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004008415550924837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021660800836980344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7884698688983918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18260, + "weighted_flow_matching_loss": 0.02342719603329897, + "weighted_rkd_loss": 0.01707888813689351 + }, + { + "epoch": 5.315682281059063, + "flow_matching_loss": 0.02155619040131569, + "grad_norm": 0.2676677703857422, + "learning_rate": 2.499449413698639e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009119151113554835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004250650899484754, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022488953173160554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7882559955120086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18270, + "weighted_flow_matching_loss": 0.02155619040131569, + "weighted_rkd_loss": 0.017727001570165157 + }, + { + "epoch": 5.318591795170207, + "flow_matching_loss": 0.022986754402518274, + "grad_norm": 0.1968623399734497, + "learning_rate": 2.4988327788203764e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007573311496526003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003315039724111557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018461662996560334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7880420386791229, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18280, + "weighted_flow_matching_loss": 0.022986754402518274, + "weighted_rkd_loss": 0.01454862579703331 + }, + { + "epoch": 5.32150130928135, + "flow_matching_loss": 0.024260375276207925, + "grad_norm": 0.22867092490196228, + "learning_rate": 2.498215840523088e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008631853992119432, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003872338030487299, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021136045828461646, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7878279983997345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18290, + "weighted_flow_matching_loss": 0.024260375276207925, + "weighted_rkd_loss": 0.016651493217796087 + }, + { + "epoch": 5.324410823392493, + "flow_matching_loss": 0.022379180043935777, + "grad_norm": 0.27039116621017456, + "learning_rate": 2.4975985989941837e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008681991323828697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039003513986244796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021264334116131066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7876138865947724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18300, + "weighted_flow_matching_loss": 0.022379180043935777, + "weighted_rkd_loss": 0.016748017445206643 + }, + { + "epoch": 5.327320337503637, + "flow_matching_loss": 0.021469458006322384, + "grad_norm": 0.24282324314117432, + "learning_rate": 2.4969810544211652e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008588777016848325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038622608641162514, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021039814688265324, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7873996913433075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18310, + "weighted_flow_matching_loss": 0.021469458006322384, + "weighted_rkd_loss": 0.016566752176731824 + }, + { + "epoch": 5.330229851614781, + "flow_matching_loss": 0.024117940664291383, + "grad_norm": 0.17980965971946716, + "learning_rate": 2.4963632069916258e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008401840180158614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003913628286682069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020717308297753333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7871854245662689, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18320, + "weighted_flow_matching_loss": 0.024117940664291383, + "weighted_rkd_loss": 0.01630839901044965 + }, + { + "epoch": 5.333139365725923, + "flow_matching_loss": 0.022312871739268304, + "grad_norm": 0.2103094905614853, + "learning_rate": 2.495745056893252e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00939451577141881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004038133681751788, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022827165201306343, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7869710743427276, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18330, + "weighted_flow_matching_loss": 0.022312871739268304, + "weighted_rkd_loss": 0.017964376229792835 + }, + { + "epoch": 5.336048879837067, + "flow_matching_loss": 0.023485312424600124, + "grad_norm": 0.20914091169834137, + "learning_rate": 2.495126604313821e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008126306394115091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037467841058969496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019999396987259388, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7867566406726837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18340, + "weighted_flow_matching_loss": 0.023485312424600124, + "weighted_rkd_loss": 0.015734650567173956 + }, + { + "epoch": 5.338958393948211, + "flow_matching_loss": 0.02099111583083868, + "grad_norm": 0.25410622358322144, + "learning_rate": 2.4945078494412023e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00782400812022388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034717581002041698, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01911977445706725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7865421175956726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18350, + "weighted_flow_matching_loss": 0.02099111583083868, + "weighted_rkd_loss": 0.015038407780230046 + }, + { + "epoch": 5.3418679080593545, + "flow_matching_loss": 0.022518152743577956, + "grad_norm": 0.22680579125881195, + "learning_rate": 2.493888792463357e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008329132758080959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003701311768963933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020359577238559724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7863275408744812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18360, + "weighted_flow_matching_loss": 0.022518152743577956, + "weighted_rkd_loss": 0.016009240318089722 + }, + { + "epoch": 5.344777422170497, + "flow_matching_loss": 0.020172211900353433, + "grad_norm": 0.19409847259521484, + "learning_rate": 2.4932694335683395e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007489511976018548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003254728252068162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018233752436935903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7861128866672515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18370, + "weighted_flow_matching_loss": 0.020172211900353433, + "weighted_rkd_loss": 0.01433383347466588 + }, + { + "epoch": 5.347686936281641, + "flow_matching_loss": 0.023451692052185537, + "grad_norm": 0.2145187258720398, + "learning_rate": 2.4926497729442932e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074742569122463465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003436059760861099, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01838457379490137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7858981370925904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18380, + "weighted_flow_matching_loss": 0.023451692052185537, + "weighted_rkd_loss": 0.014448382798582316 + }, + { + "epoch": 5.350596450392785, + "flow_matching_loss": 0.02209802661091089, + "grad_norm": 0.2514093518257141, + "learning_rate": 2.4920298107794555e-05, + "loss": 0.0399, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009171932702884078, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004283904936164618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022627770341932772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7856833159923553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18390, + "weighted_flow_matching_loss": 0.02209802661091089, + "weighted_rkd_loss": 0.01777814785018563 + }, + { + "epoch": 5.353505964503928, + "flow_matching_loss": 0.023562074080109595, + "grad_norm": 0.22751305997371674, + "learning_rate": 2.4914095472621544e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008396041067317127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037293145200237633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020521396677941083, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7854684114456176, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18400, + "weighted_flow_matching_loss": 0.023562074080109595, + "weighted_rkd_loss": 0.01611895887181163 + }, + { + "epoch": 5.356415478615071, + "flow_matching_loss": 0.02337545491755009, + "grad_norm": 0.23003937304019928, + "learning_rate": 2.4907889825808093e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009366846550256014, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003953550569713115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022687243483960628, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7852534353733063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18410, + "weighted_flow_matching_loss": 0.02337545491755009, + "weighted_rkd_loss": 0.01781523134559393 + }, + { + "epoch": 5.359324992726215, + "flow_matching_loss": 0.023330249078571796, + "grad_norm": 0.26103487610816956, + "learning_rate": 2.4901681169239314e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008969142660498619, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004173798020929098, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02211208324879408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7850383758544922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18420, + "weighted_flow_matching_loss": 0.023330249078571796, + "weighted_rkd_loss": 0.017358805425465106 + }, + { + "epoch": 5.362234506837358, + "flow_matching_loss": 0.02247688118368387, + "grad_norm": 0.20926876366138458, + "learning_rate": 2.4895469504801236e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00859938864596188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038931188406422733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021091896481812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7848232686519623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18430, + "weighted_flow_matching_loss": 0.02247688118368387, + "weighted_rkd_loss": 0.01655351659283042 + }, + { + "epoch": 5.3651440209485015, + "flow_matching_loss": 0.021863360516726972, + "grad_norm": 0.22655488550662994, + "learning_rate": 2.4889254834380802e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007888365304097533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034938695607706904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019270600471645593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7846080243587494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18440, + "weighted_flow_matching_loss": 0.021863360516726972, + "weighted_rkd_loss": 0.01511975796893239 + }, + { + "epoch": 5.368053535059645, + "flow_matching_loss": 0.021592299453914165, + "grad_norm": 0.21265411376953125, + "learning_rate": 2.488303715986585e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008404317544773221, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003508435608819127, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02031707055866718, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7843927443027496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18450, + "weighted_flow_matching_loss": 0.021592299453914165, + "weighted_rkd_loss": 0.015936606377363206 + }, + { + "epoch": 5.370963049170788, + "flow_matching_loss": 0.022958150878548623, + "grad_norm": 0.234922394156456, + "learning_rate": 2.4876816483145166e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00889277383685112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003916993783786893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02170254122465849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7841773748397827, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18460, + "weighted_flow_matching_loss": 0.022958150878548623, + "weighted_rkd_loss": 0.017018607445061208 + }, + { + "epoch": 5.373872563281932, + "flow_matching_loss": 0.02062912117689848, + "grad_norm": 0.19220255315303802, + "learning_rate": 2.487059280610841e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008541095815598965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003892141836695373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020974333584308624, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7839619219303131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18470, + "weighted_flow_matching_loss": 0.02062912117689848, + "weighted_rkd_loss": 0.016443144623190166 + }, + { + "epoch": 5.3767820773930755, + "flow_matching_loss": 0.02211334053426981, + "grad_norm": 0.22162963449954987, + "learning_rate": 2.4864366130646178e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008498774003237486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038841587025672196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020881706569343805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7837464034557342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18480, + "weighted_flow_matching_loss": 0.02211334053426981, + "weighted_rkd_loss": 0.016365960892289876 + }, + { + "epoch": 5.379691591504219, + "flow_matching_loss": 0.025203806534409522, + "grad_norm": 0.20631512999534607, + "learning_rate": 2.485813645864997e-05, + "loss": 0.0405, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008023692900314927, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003418010286986828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019465396553277968, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7835307955741883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18490, + "weighted_flow_matching_loss": 0.025203806534409522, + "weighted_rkd_loss": 0.015251706074923276 + }, + { + "epoch": 5.382601105615362, + "flow_matching_loss": 0.024989531189203263, + "grad_norm": 0.23220592737197876, + "learning_rate": 2.4851903792012198e-05, + "loss": 0.0416, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008655532589182257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00388983441516757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0212008997797966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.783315122127533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18500, + "weighted_flow_matching_loss": 0.024989531189203263, + "weighted_rkd_loss": 0.016606990713626146 + }, + { + "epoch": 5.385510619726506, + "flow_matching_loss": 0.021900540962815285, + "grad_norm": 0.21897172927856445, + "learning_rate": 2.4845668132626173e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00852081561461091, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004010917223058641, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0210525487549603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.783099353313446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18510, + "weighted_flow_matching_loss": 0.021900540962815285, + "weighted_rkd_loss": 0.01648633452132344 + }, + { + "epoch": 5.388420133837649, + "flow_matching_loss": 0.021744051948189735, + "grad_norm": 0.2212565690279007, + "learning_rate": 2.483942948238613e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007887876685708761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034264825517311692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01920223580673337, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7828835248947144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18520, + "weighted_flow_matching_loss": 0.021744051948189735, + "weighted_rkd_loss": 0.015033034887164832 + }, + { + "epoch": 5.391329647948792, + "flow_matching_loss": 0.023842153139412403, + "grad_norm": 0.22874654829502106, + "learning_rate": 2.4833187843187207e-05, + "loss": 0.0403, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008528951043263077, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003952381387352944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021010283567011356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7826676189899444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18530, + "weighted_flow_matching_loss": 0.023842153139412403, + "weighted_rkd_loss": 0.016444174759089946 + }, + { + "epoch": 5.394239162059936, + "flow_matching_loss": 0.022392526641488074, + "grad_norm": 0.2038070410490036, + "learning_rate": 2.4826943216925448e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007930557196959852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035354400984942912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019396554306149484, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7824516236782074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18540, + "weighted_flow_matching_loss": 0.022392526641488074, + "weighted_rkd_loss": 0.015176923014223576 + }, + { + "epoch": 5.39714867617108, + "flow_matching_loss": 0.022349340096116065, + "grad_norm": 0.18644069135189056, + "learning_rate": 2.4820695605497807e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008487161854282022, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036359924823045732, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020610316284000875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7822355508804322, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18550, + "weighted_flow_matching_loss": 0.022349340096116065, + "weighted_rkd_loss": 0.01612219922244549 + }, + { + "epoch": 5.400058190282223, + "flow_matching_loss": 0.02248156238347292, + "grad_norm": 0.21175885200500488, + "learning_rate": 2.481444501080214e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008074014401063322, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034315077820792792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019579536467790603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7820194125175476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18560, + "weighted_flow_matching_loss": 0.02248156238347292, + "weighted_rkd_loss": 0.015311521757394076 + }, + { + "epoch": 5.402967704393366, + "flow_matching_loss": 0.024758439511060715, + "grad_norm": 0.2104095071554184, + "learning_rate": 2.4808191434737217e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008155412878841162, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003689246694557369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02000007275491953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7818031907081604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18570, + "weighted_flow_matching_loss": 0.024758439511060715, + "weighted_rkd_loss": 0.01563614709302783 + }, + { + "epoch": 5.40587721850451, + "flow_matching_loss": 0.020449301041662694, + "grad_norm": 0.20672081410884857, + "learning_rate": 2.4801934879202696e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007996349642053247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034520547837018968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019444754160940648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7815868973731994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18580, + "weighted_flow_matching_loss": 0.020449301041662694, + "weighted_rkd_loss": 0.01519766841083765 + }, + { + "epoch": 5.408786732615654, + "flow_matching_loss": 0.022161812894046305, + "grad_norm": 0.24501237273216248, + "learning_rate": 2.4795675346099172e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00787326656281948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035684025613591077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01931493580341339, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7813705086708069, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18590, + "weighted_flow_matching_loss": 0.022161812894046305, + "weighted_rkd_loss": 0.015092053916305304 + }, + { + "epoch": 5.411696246726796, + "flow_matching_loss": 0.023178161308169364, + "grad_norm": 0.2183995544910431, + "learning_rate": 2.4789412837328114e-05, + "loss": 0.0386, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008124631363898516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003553651599213481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019802914187312125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7811540603637696, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18600, + "weighted_flow_matching_loss": 0.023178161308169364, + "weighted_rkd_loss": 0.015469047240912914 + }, + { + "epoch": 5.41460576083794, + "flow_matching_loss": 0.023250453919172288, + "grad_norm": 0.19331084191799164, + "learning_rate": 2.4783147354791915e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007846383703872561, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003396937996149063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0190897055901587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7809375286102295, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18610, + "weighted_flow_matching_loss": 0.023250453919172288, + "weighted_rkd_loss": 0.014907851908355951 + }, + { + "epoch": 5.417515274949084, + "flow_matching_loss": 0.025236645713448524, + "grad_norm": 0.19030623137950897, + "learning_rate": 2.4776878900393858e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008335059601813554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00352907560300082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020199194736778735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7807209193706512, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18620, + "weighted_flow_matching_loss": 0.025236645713448524, + "weighted_rkd_loss": 0.015770053956657648 + }, + { + "epoch": 5.420424789060227, + "flow_matching_loss": 0.0219706229865551, + "grad_norm": 0.22240452468395233, + "learning_rate": 2.477060747603813e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008693747920915484, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003822003467939794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021209499426186086, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7805042326450348, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18630, + "weighted_flow_matching_loss": 0.0219706229865551, + "weighted_rkd_loss": 0.016554104071110486 + }, + { + "epoch": 5.42333430317137, + "flow_matching_loss": 0.023204433359205724, + "grad_norm": 0.2212158590555191, + "learning_rate": 2.4764333083629833e-05, + "loss": 0.0397, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00865224413573742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038643666077405213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021168855018913746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7802874684333801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18640, + "weighted_flow_matching_loss": 0.023204433359205724, + "weighted_rkd_loss": 0.01651793336495757 + }, + { + "epoch": 5.426243817282514, + "flow_matching_loss": 0.02131074871867895, + "grad_norm": 0.29788148403167725, + "learning_rate": 2.4758055725074954e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008262904407456518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035367623437196015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020062571112066507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7800706326961517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18650, + "weighted_flow_matching_loss": 0.02131074871867895, + "weighted_rkd_loss": 0.015650218725204466 + }, + { + "epoch": 5.429153331393657, + "flow_matching_loss": 0.020706488005816936, + "grad_norm": 0.20797665417194366, + "learning_rate": 2.4751775402280396e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007765388675034046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032532244455069303, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01878400146961212, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7798537015914917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18660, + "weighted_flow_matching_loss": 0.020706488005816936, + "weighted_rkd_loss": 0.014648793824017048 + }, + { + "epoch": 5.432062845504801, + "flow_matching_loss": 0.022498947754502296, + "grad_norm": 0.19336961209774017, + "learning_rate": 2.4745492117153947e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007558642700314522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034814845072105526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018598769884556533, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7796367228031158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18670, + "weighted_flow_matching_loss": 0.022498947754502296, + "weighted_rkd_loss": 0.014500286057591438 + }, + { + "epoch": 5.434972359615944, + "flow_matching_loss": 0.02262246422469616, + "grad_norm": 0.21374981105327606, + "learning_rate": 2.473920587160431e-05, + "loss": 0.039, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00860711270943284, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003858173033222556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021072398498654367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7794196546077728, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18680, + "weighted_flow_matching_loss": 0.02262246422469616, + "weighted_rkd_loss": 0.0164243271574378 + }, + { + "epoch": 5.437881873727088, + "flow_matching_loss": 0.02202519718557596, + "grad_norm": 0.21971257030963898, + "learning_rate": 2.473291666754108e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00834628939628601, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038370017195120455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020529580488801002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7792025029659271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18690, + "weighted_flow_matching_loss": 0.02202519718557596, + "weighted_rkd_loss": 0.01599685922265053 + }, + { + "epoch": 5.440791387838231, + "flow_matching_loss": 0.022673221305012703, + "grad_norm": 0.2184656411409378, + "learning_rate": 2.4726624506874748e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007764960499480366, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035572286928072573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019087149854749443, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7789852917194366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18700, + "weighted_flow_matching_loss": 0.022673221305012703, + "weighted_rkd_loss": 0.014868610631674528 + }, + { + "epoch": 5.4437009019493745, + "flow_matching_loss": 0.02194875255227089, + "grad_norm": 0.1936979740858078, + "learning_rate": 2.4720329391516708e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007651024265214801, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003298194077797234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01860024221241474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.77876797914505, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18710, + "weighted_flow_matching_loss": 0.02194875255227089, + "weighted_rkd_loss": 0.01448527043685317 + }, + { + "epoch": 5.446610416060518, + "flow_matching_loss": 0.024698251858353615, + "grad_norm": 0.1967453509569168, + "learning_rate": 2.4714031323379248e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008595756348222494, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003768850350752473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02096036318689585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7785506069660186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18720, + "weighted_flow_matching_loss": 0.024698251858353615, + "weighted_rkd_loss": 0.016318810079246758 + }, + { + "epoch": 5.449519930171661, + "flow_matching_loss": 0.023230721801519395, + "grad_norm": 0.2357485145330429, + "learning_rate": 2.4707730304375556e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007784220902249217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034387038787826894, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019007145799696444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7783331513404846, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18730, + "weighted_flow_matching_loss": 0.023230721801519395, + "weighted_rkd_loss": 0.0147939826361835 + }, + { + "epoch": 5.452429444282805, + "flow_matching_loss": 0.02431740313768387, + "grad_norm": 0.20062275230884552, + "learning_rate": 2.4701426336419713e-05, + "loss": 0.0409, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008700559474527835, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038907573791220782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02129187621176243, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7781156361103058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18740, + "weighted_flow_matching_loss": 0.02431740313768387, + "weighted_rkd_loss": 0.016567640472203493 + }, + { + "epoch": 5.4553389583939484, + "flow_matching_loss": 0.021912401728332042, + "grad_norm": 0.205287367105484, + "learning_rate": 2.4695119421426707e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008189230179414152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036684904014691712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020046950690448283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7778980255126953, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18750, + "weighted_flow_matching_loss": 0.021912401728332042, + "weighted_rkd_loss": 0.015594537556171417 + }, + { + "epoch": 5.458248472505091, + "flow_matching_loss": 0.023226252757012844, + "grad_norm": 0.30174848437309265, + "learning_rate": 2.46888095613124e-05, + "loss": 0.041, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009355749888345598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00413432524073869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022845824994146825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7776803493499755, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18760, + "weighted_flow_matching_loss": 0.023226252757012844, + "weighted_rkd_loss": 0.017766692861914634 + }, + { + "epoch": 5.461157986616235, + "flow_matching_loss": 0.024251981265842916, + "grad_norm": 0.2592237889766693, + "learning_rate": 2.468249675799357e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008464281586930156, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037872325628995895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020715795643627643, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7774625897407532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18770, + "weighted_flow_matching_loss": 0.024251981265842916, + "weighted_rkd_loss": 0.01610572263598442 + }, + { + "epoch": 5.464067500727379, + "flow_matching_loss": 0.023273144103586673, + "grad_norm": 0.22253386676311493, + "learning_rate": 2.467618101338787e-05, + "loss": 0.0391, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008349409000948072, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003677996038459241, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020376814156770708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7772447466850281, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18780, + "weighted_flow_matching_loss": 0.023273144103586673, + "weighted_rkd_loss": 0.015837840363383295 + }, + { + "epoch": 5.4669770148385215, + "flow_matching_loss": 0.023281623236835004, + "grad_norm": 0.20129722356796265, + "learning_rate": 2.466986232941387e-05, + "loss": 0.04, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008819517074152827, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003916072170250117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02155510652810335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7770268440246582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18790, + "weighted_flow_matching_loss": 0.023281623236835004, + "weighted_rkd_loss": 0.016748898196965456 + }, + { + "epoch": 5.469886528949665, + "flow_matching_loss": 0.023459598980844022, + "grad_norm": 0.20713557302951813, + "learning_rate": 2.466354070799101e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007932459795847535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034009729279205203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019265892542898656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7768088638782501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18800, + "weighted_flow_matching_loss": 0.023459598980844022, + "weighted_rkd_loss": 0.014965907670557499 + }, + { + "epoch": 5.472796043060809, + "flow_matching_loss": 0.020289889723062515, + "grad_norm": 0.21365971863269806, + "learning_rate": 2.4657216151039634e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00837195017375052, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036450094077736138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020388909988105296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7765908062458038, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18810, + "weighted_flow_matching_loss": 0.020289889723062515, + "weighted_rkd_loss": 0.01583385970443487 + }, + { + "epoch": 5.475705557171953, + "flow_matching_loss": 0.023651325330138206, + "grad_norm": 0.21691708266735077, + "learning_rate": 2.4650888660480976e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008845689380541444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003942265501245856, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02163364440202713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7763726711273193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18820, + "weighted_flow_matching_loss": 0.023651325330138206, + "weighted_rkd_loss": 0.01679582865908742 + }, + { + "epoch": 5.478615071283095, + "flow_matching_loss": 0.021888604573905467, + "grad_norm": 0.20837733149528503, + "learning_rate": 2.464455823823716e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008106592250987888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037158019607886673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019928986206650734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7761544525623322, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18830, + "weighted_flow_matching_loss": 0.021888604573905467, + "weighted_rkd_loss": 0.015467965789139271 + }, + { + "epoch": 5.481524585394239, + "flow_matching_loss": 0.025165139138698576, + "grad_norm": 0.22048266232013702, + "learning_rate": 2.4638224886231196e-05, + "loss": 0.0407, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008222403563559055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00356166095007211, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020006468147039415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7759361684322357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18840, + "weighted_flow_matching_loss": 0.025165139138698576, + "weighted_rkd_loss": 0.015523745864629745 + }, + { + "epoch": 5.484434099505383, + "flow_matching_loss": 0.022909382171928883, + "grad_norm": 0.22130677103996277, + "learning_rate": 2.4631888606387e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007914265291765333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003447719058021903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019276249315589668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.775717806816101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18850, + "weighted_flow_matching_loss": 0.022909382171928883, + "weighted_rkd_loss": 0.014952956605702639 + }, + { + "epoch": 5.487343613616526, + "flow_matching_loss": 0.022774350456893445, + "grad_norm": 0.26196879148483276, + "learning_rate": 2.4625549400629356e-05, + "loss": 0.0396, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008865181868895889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004007571656256914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02173793539404869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7754993677139282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18860, + "weighted_flow_matching_loss": 0.022774350456893445, + "weighted_rkd_loss": 0.01685785558074713 + }, + { + "epoch": 5.490253127727669, + "flow_matching_loss": 0.022618773579597472, + "grad_norm": 0.2163136601448059, + "learning_rate": 2.4619207270883958e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0087608955334872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003957110946066678, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02147890217602253, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7752808511257172, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18870, + "weighted_flow_matching_loss": 0.022618773579597472, + "weighted_rkd_loss": 0.016652241256088017 + }, + { + "epoch": 5.493162641838813, + "flow_matching_loss": 0.0220781235024333, + "grad_norm": 0.19870272278785706, + "learning_rate": 2.461286221907737e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007653513457626104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033333380473777653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01864036489278078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7750622570514679, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18880, + "weighted_flow_matching_loss": 0.0220781235024333, + "weighted_rkd_loss": 0.01444741040468216 + }, + { + "epoch": 5.496072155949957, + "flow_matching_loss": 0.02329641245305538, + "grad_norm": 0.19278550148010254, + "learning_rate": 2.460651424713705e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008227541670203209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0041646625148132445, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02061974573880434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7748435974121094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18890, + "weighted_flow_matching_loss": 0.02329641245305538, + "weighted_rkd_loss": 0.015977186895906926 + }, + { + "epoch": 5.4989816700611, + "flow_matching_loss": 0.02043846044689417, + "grad_norm": 0.2397058755159378, + "learning_rate": 2.4600163356991347e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00788648328743875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003204180090688169, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01897714687511325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7746248543262482, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18900, + "weighted_flow_matching_loss": 0.02043846044689417, + "weighted_rkd_loss": 0.014700100291520357 + }, + { + "epoch": 5.501891184172243, + "flow_matching_loss": 0.023295652121305466, + "grad_norm": 0.24113695323467255, + "learning_rate": 2.4593809550569498e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008159473305568099, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037225881358608605, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02004153486341238, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7744060337543488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18910, + "weighted_flow_matching_loss": 0.023295652121305466, + "weighted_rkd_loss": 0.015520235523581505 + }, + { + "epoch": 5.504800698283387, + "flow_matching_loss": 0.02201029509305954, + "grad_norm": 0.18792064487934113, + "learning_rate": 2.4587452829801614e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007757719326764345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003636683500371873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01915212208405137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7741871416568756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18920, + "weighted_flow_matching_loss": 0.02201029509305954, + "weighted_rkd_loss": 0.014827396254986525 + }, + { + "epoch": 5.50771021239453, + "flow_matching_loss": 0.025302915647625924, + "grad_norm": 0.19563132524490356, + "learning_rate": 2.45810931966187e-05, + "loss": 0.0412, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008417881885543465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037420547800138594, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020577818714082242, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7739681780338288, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18930, + "weighted_flow_matching_loss": 0.025302915647625924, + "weighted_rkd_loss": 0.01592659056186676 + }, + { + "epoch": 5.5106197265056736, + "flow_matching_loss": 0.02281213542446494, + "grad_norm": 0.19248761236667633, + "learning_rate": 2.457473065295265e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008061890304088593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034632406197488306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019587021227926014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7737491369247437, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18940, + "weighted_flow_matching_loss": 0.02281213542446494, + "weighted_rkd_loss": 0.015155417378991842 + }, + { + "epoch": 5.513529240616817, + "flow_matching_loss": 0.022752862982451917, + "grad_norm": 0.24781325459480286, + "learning_rate": 2.4568365200736232e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008688622200861574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003829312534071505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021206556633114815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7735300183296203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18950, + "weighted_flow_matching_loss": 0.022752862982451917, + "weighted_rkd_loss": 0.016403965931385756 + }, + { + "epoch": 5.51643875472796, + "flow_matching_loss": 0.022519285045564174, + "grad_norm": 0.21574679017066956, + "learning_rate": 2.4561996841903104e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008198678027838469, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036934451665729283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020090801175683737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7733108282089234, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18960, + "weighted_flow_matching_loss": 0.022519285045564174, + "weighted_rkd_loss": 0.015536436904221773 + }, + { + "epoch": 5.519348268839104, + "flow_matching_loss": 0.019970852322876452, + "grad_norm": 0.1955079734325409, + "learning_rate": 2.455562557838781e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008052899921312928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003628011606633663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019733811356127264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7730915546417236, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18970, + "weighted_flow_matching_loss": 0.019970852322876452, + "weighted_rkd_loss": 0.015256031882017851 + }, + { + "epoch": 5.5222577829502475, + "flow_matching_loss": 0.021615130268037318, + "grad_norm": 0.22013632953166962, + "learning_rate": 2.4549251412125762e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007938190130516887, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034723928198218345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01934877298772335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7728722095489502, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18980, + "weighted_flow_matching_loss": 0.021615130268037318, + "weighted_rkd_loss": 0.01495414013043046 + }, + { + "epoch": 5.52516729706139, + "flow_matching_loss": 0.02172722052782774, + "grad_norm": 0.21310196816921234, + "learning_rate": 2.454287434505327e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007838386856019498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003625576593913138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019302350468933582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.772652804851532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 18990, + "weighted_flow_matching_loss": 0.02172722052782774, + "weighted_rkd_loss": 0.014913977775722741 + }, + { + "epoch": 5.528076811172534, + "flow_matching_loss": 0.02281438261270523, + "grad_norm": 0.18192598223686218, + "learning_rate": 2.4536494379107514e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072446238249540326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032102872151881456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01769953463226557, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7724333107471466, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19000, + "weighted_flow_matching_loss": 0.02281438261270523, + "weighted_rkd_loss": 0.013671777676790953 + }, + { + "epoch": 5.530986325283678, + "flow_matching_loss": 0.0227881470695138, + "grad_norm": 0.17555038630962372, + "learning_rate": 2.453011151622657e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007746722642332315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003349126526154578, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018842571787536143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.772213751077652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19010, + "weighted_flow_matching_loss": 0.0227881470695138, + "weighted_rkd_loss": 0.014550571236759425 + }, + { + "epoch": 5.533895839394821, + "flow_matching_loss": 0.020350064896047116, + "grad_norm": 0.18366169929504395, + "learning_rate": 2.452372575834937e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007721766317263246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034676389768719673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018911171704530716, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7719941079616547, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19020, + "weighted_flow_matching_loss": 0.020350064896047116, + "weighted_rkd_loss": 0.014599371515214444 + }, + { + "epoch": 5.536805353505964, + "flow_matching_loss": 0.021617510914802553, + "grad_norm": 0.18247021734714508, + "learning_rate": 2.4517337107415742e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0084802798461169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003945529344491661, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02090608887374401, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7717743933200836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19030, + "weighted_flow_matching_loss": 0.021617510914802553, + "weighted_rkd_loss": 0.01613481668755412 + }, + { + "epoch": 5.539714867617108, + "flow_matching_loss": 0.021982036530971527, + "grad_norm": 0.21164491772651672, + "learning_rate": 2.4510945565366397e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008229203196242452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036573421210050585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02011574860662222, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7715546011924743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19040, + "weighted_flow_matching_loss": 0.021982036530971527, + "weighted_rkd_loss": 0.015520504955202342 + }, + { + "epoch": 5.542624381728252, + "flow_matching_loss": 0.0226899653673172, + "grad_norm": 0.4438873827457428, + "learning_rate": 2.4504551134142905e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007878580316901206, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035133648896589877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019270525500178336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7713347375392914, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19050, + "weighted_flow_matching_loss": 0.0226899653673172, + "weighted_rkd_loss": 0.014863957371562719 + }, + { + "epoch": 5.5455338958393945, + "flow_matching_loss": 0.02301832176744938, + "grad_norm": 0.20844405889511108, + "learning_rate": 2.4498153815687738e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008163472078740597, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003649591514840722, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019976535625755788, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7711147964000702, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19060, + "weighted_flow_matching_loss": 0.02301832176744938, + "weighted_rkd_loss": 0.015404181461781264 + }, + { + "epoch": 5.548443409950538, + "flow_matching_loss": 0.022006539814174175, + "grad_norm": 0.19186754524707794, + "learning_rate": 2.4491753611944224e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00778772053308785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035356330219656227, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019111074134707452, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7708947837352753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19070, + "weighted_flow_matching_loss": 0.022006539814174175, + "weighted_rkd_loss": 0.014732592087239028 + }, + { + "epoch": 5.551352924061682, + "flow_matching_loss": 0.022024841606616975, + "grad_norm": 0.23512554168701172, + "learning_rate": 2.4485350524856577e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008201715163886547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035961598390713335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019999589957296848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7706746995449066, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19080, + "weighted_flow_matching_loss": 0.022024841606616975, + "weighted_rkd_loss": 0.015413204114884138 + }, + { + "epoch": 5.554262438172826, + "flow_matching_loss": 0.019794352538883685, + "grad_norm": 0.1765081286430359, + "learning_rate": 2.4478944556369886e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008004594268277288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035136751597747208, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019522863812744616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7704545438289643, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19090, + "weighted_flow_matching_loss": 0.019794352538883685, + "weighted_rkd_loss": 0.015041521936655044 + }, + { + "epoch": 5.557171952283968, + "flow_matching_loss": 0.021077523473650217, + "grad_norm": 0.23309288918972015, + "learning_rate": 2.4472535708430116e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008562569366768002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003578097024001181, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02070323582738638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7702343046665192, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19100, + "weighted_flow_matching_loss": 0.021077523473650217, + "weighted_rkd_loss": 0.01594628281891346 + }, + { + "epoch": 5.560081466395112, + "flow_matching_loss": 0.02113640308380127, + "grad_norm": 0.2292931079864502, + "learning_rate": 2.4466123982984103e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008640471193939448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003774084686301649, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021055027097463607, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7700139939785003, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19110, + "weighted_flow_matching_loss": 0.02113640308380127, + "weighted_rkd_loss": 0.01621265523135662 + }, + { + "epoch": 5.562990980506255, + "flow_matching_loss": 0.02366008944809437, + "grad_norm": 0.1969892531633377, + "learning_rate": 2.445970938197957e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008039103541523217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035075327847152948, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019585739821195602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7697936117649078, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19120, + "weighted_flow_matching_loss": 0.02366008944809437, + "weighted_rkd_loss": 0.015076965559273958 + }, + { + "epoch": 5.565900494617399, + "flow_matching_loss": 0.021710250154137613, + "grad_norm": 0.2604992687702179, + "learning_rate": 2.4453291907365092e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00842317850328982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038619652390480042, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020708322525024414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7695731520652771, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19130, + "weighted_flow_matching_loss": 0.021710250154137613, + "weighted_rkd_loss": 0.015936437249183654 + }, + { + "epoch": 5.568810008728542, + "flow_matching_loss": 0.022829634509980677, + "grad_norm": 0.21807630360126495, + "learning_rate": 2.444687156109013e-05, + "loss": 0.0384, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008230478642508388, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003766519669443369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02022747676819563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7693526148796082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19140, + "weighted_flow_matching_loss": 0.022829634509980677, + "weighted_rkd_loss": 0.015562070906162262 + }, + { + "epoch": 5.571719522839686, + "flow_matching_loss": 0.02129395343363285, + "grad_norm": 0.2062900811433792, + "learning_rate": 2.444044834510502e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008576901303604246, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003690334944985807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020844137668609618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7691320180892944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19150, + "weighted_flow_matching_loss": 0.02129395343363285, + "weighted_rkd_loss": 0.016031962260603903 + }, + { + "epoch": 5.574629036950829, + "flow_matching_loss": 0.019505090638995172, + "grad_norm": 0.18846870958805084, + "learning_rate": 2.443402226136096e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008627129718661308, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038048578426241875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02105911746621132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7689113438129425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19160, + "weighted_flow_matching_loss": 0.019505090638995172, + "weighted_rkd_loss": 0.01619259649887681 + }, + { + "epoch": 5.577538551061973, + "flow_matching_loss": 0.022542891465127467, + "grad_norm": 0.18713323771953583, + "learning_rate": 2.442759331181003e-05, + "loss": 0.0385, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008494835766032337, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037522477796301246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02074191942811012, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7686905860900879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19170, + "weighted_flow_matching_loss": 0.022542891465127467, + "weighted_rkd_loss": 0.01594420289620757 + }, + { + "epoch": 5.580448065173116, + "flow_matching_loss": 0.02048707250505686, + "grad_norm": 0.23169443011283875, + "learning_rate": 2.4421161498405176e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00771920420229435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003234249772503972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01867265794426203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.768469762802124, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19180, + "weighted_flow_matching_loss": 0.02048707250505686, + "weighted_rkd_loss": 0.014349369704723359 + }, + { + "epoch": 5.583357579284259, + "flow_matching_loss": 0.022366117499768735, + "grad_norm": 0.22617462277412415, + "learning_rate": 2.4414726823100207e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008414017967879772, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037643172079697253, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02059235293418169, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7682488560676575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19190, + "weighted_flow_matching_loss": 0.022366117499768735, + "weighted_rkd_loss": 0.01581999696791172 + }, + { + "epoch": 5.586267093395403, + "flow_matching_loss": 0.02298932820558548, + "grad_norm": 0.24310091137886047, + "learning_rate": 2.4408289287849813e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007414619484916329, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032886374508962037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018117876537144185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7680278897285462, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19200, + "weighted_flow_matching_loss": 0.02298932820558548, + "weighted_rkd_loss": 0.013914873544126748 + }, + { + "epoch": 5.5891766075065465, + "flow_matching_loss": 0.022124798968434334, + "grad_norm": 0.24288497865200043, + "learning_rate": 2.4401848894609543e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008507492812350392, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035325732082128526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020547558926045893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7678068459033967, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19210, + "weighted_flow_matching_loss": 0.022124798968434334, + "weighted_rkd_loss": 0.015776429884135725 + }, + { + "epoch": 5.59208612161769, + "flow_matching_loss": 0.021842608228325844, + "grad_norm": 0.2166466861963272, + "learning_rate": 2.439540564533582e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.009154518879950047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004410619125701487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02271965704858303, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7675857245922089, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19220, + "weighted_flow_matching_loss": 0.021842608228325844, + "weighted_rkd_loss": 0.01743931444361806 + }, + { + "epoch": 5.594995635728833, + "flow_matching_loss": 0.024314704537391662, + "grad_norm": 0.2213447242975235, + "learning_rate": 2.4388959541985935e-05, + "loss": 0.0401, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008393340790644289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003798723570071161, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02058540526777506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7673645257949829, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19230, + "weighted_flow_matching_loss": 0.024314704537391662, + "weighted_rkd_loss": 0.015796511713415385 + }, + { + "epoch": 5.597905149839977, + "flow_matching_loss": 0.02209543585777283, + "grad_norm": 0.19979768991470337, + "learning_rate": 2.438251058651804e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007969328854233026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033796072006225587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019318265095353126, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7671432614326477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19240, + "weighted_flow_matching_loss": 0.02209543585777283, + "weighted_rkd_loss": 0.014819906279444694 + }, + { + "epoch": 5.6008146639511205, + "flow_matching_loss": 0.022638265043497086, + "grad_norm": 0.19385607540607452, + "learning_rate": 2.437605878089116e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007584774633869529, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003375366562977433, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018544915691018106, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7669219195842742, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19250, + "weighted_flow_matching_loss": 0.022638265043497086, + "weighted_rkd_loss": 0.014222539961338043 + }, + { + "epoch": 5.603724178062263, + "flow_matching_loss": 0.022607576567679644, + "grad_norm": 0.17521731555461884, + "learning_rate": 2.4369604127065183e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008902244362980127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039947993820533155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021799288038164378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7667005062103271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19260, + "weighted_flow_matching_loss": 0.022607576567679644, + "weighted_rkd_loss": 0.016713539510965346 + }, + { + "epoch": 5.606633692173407, + "flow_matching_loss": 0.022945021837949754, + "grad_norm": 0.18811100721359253, + "learning_rate": 2.4363146627000862e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007620130525901913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033470896771177648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018587350938469173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7664790213108063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19270, + "weighted_flow_matching_loss": 0.022945021837949754, + "weighted_rkd_loss": 0.014246920589357615 + }, + { + "epoch": 5.609543206284551, + "flow_matching_loss": 0.023219514824450015, + "grad_norm": 0.19467920064926147, + "learning_rate": 2.435668628265982e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00770610710605979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003386821853928268, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018799036275595427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7662574589252472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19280, + "weighted_flow_matching_loss": 0.023219514824450015, + "weighted_rkd_loss": 0.01440481673926115 + }, + { + "epoch": 5.6124527203956935, + "flow_matching_loss": 0.020333915203809737, + "grad_norm": 0.22304703295230865, + "learning_rate": 2.4350223096004533e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008594605373218656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00397642448078841, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02116563534364104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7660358309745788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19290, + "weighted_flow_matching_loss": 0.020333915203809737, + "weighted_rkd_loss": 0.016213532909750937 + }, + { + "epoch": 5.615362234506837, + "flow_matching_loss": 0.02176089324057102, + "grad_norm": 0.23227518796920776, + "learning_rate": 2.4343757068998343e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0085737319663167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004006444476544857, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02115390859544277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7658141255378723, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19300, + "weighted_flow_matching_loss": 0.02176089324057102, + "weighted_rkd_loss": 0.01620000135153532 + }, + { + "epoch": 5.618271748617981, + "flow_matching_loss": 0.022405608929693697, + "grad_norm": 0.2353588044643402, + "learning_rate": 2.4337288203605465e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008203248819336295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036273711360991, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020033868867903946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7655923426151275, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19310, + "weighted_flow_matching_loss": 0.022405608929693697, + "weighted_rkd_loss": 0.01533770365640521 + }, + { + "epoch": 5.621181262729124, + "flow_matching_loss": 0.022126591019332408, + "grad_norm": 0.19690227508544922, + "learning_rate": 2.433081650179097e-05, + "loss": 0.0381, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00849742079153657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003820863412693143, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020815705135464668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7653704941272735, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19320, + "weighted_flow_matching_loss": 0.022126591019332408, + "weighted_rkd_loss": 0.015931736771017314 + }, + { + "epoch": 5.6240907768402675, + "flow_matching_loss": 0.021003457903862, + "grad_norm": 0.2449297159910202, + "learning_rate": 2.4324341965520787e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007642389088869095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033573161344975233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01864209435880184, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7651485621929168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19330, + "weighted_flow_matching_loss": 0.021003457903862, + "weighted_rkd_loss": 0.014263961650431155 + }, + { + "epoch": 5.627000290951411, + "flow_matching_loss": 0.022681705839931964, + "grad_norm": 0.17989768087863922, + "learning_rate": 2.4317864596761706e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007773234182968736, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034178716596215963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018964340351521967, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7649265825748444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19340, + "weighted_flow_matching_loss": 0.022681705839931964, + "weighted_rkd_loss": 0.01450634617358446 + }, + { + "epoch": 5.629909805062555, + "flow_matching_loss": 0.0211013937368989, + "grad_norm": 0.18403343856334686, + "learning_rate": 2.4311384397481387e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007614406151697039, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032613562420010568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018490168638527392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7647045016288757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19350, + "weighted_flow_matching_loss": 0.0211013937368989, + "weighted_rkd_loss": 0.014139486383646727 + }, + { + "epoch": 5.632819319173698, + "flow_matching_loss": 0.02169441506266594, + "grad_norm": 0.19825567305088043, + "learning_rate": 2.4304901369648344e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008478666376322507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035328611033037306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020490193832665683, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7644823610782623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19360, + "weighted_flow_matching_loss": 0.02169441506266594, + "weighted_rkd_loss": 0.01566438181325793 + }, + { + "epoch": 5.635728833284841, + "flow_matching_loss": 0.0202125689946115, + "grad_norm": 0.21531078219413757, + "learning_rate": 2.429841551523194e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008414277574047446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036463546799495815, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02047490980476141, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7642601430416107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19370, + "weighted_flow_matching_loss": 0.0202125689946115, + "weighted_rkd_loss": 0.015648205298930407 + }, + { + "epoch": 5.638638347395985, + "flow_matching_loss": 0.022472681291401387, + "grad_norm": 0.22378137707710266, + "learning_rate": 2.4291926836202416e-05, + "loss": 0.0392, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008965592179447413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0039455503923818466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021876734495162965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7640378773212433, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19380, + "weighted_flow_matching_loss": 0.022472681291401387, + "weighted_rkd_loss": 0.016714700032025576 + }, + { + "epoch": 5.641547861507128, + "flow_matching_loss": 0.02146578636020422, + "grad_norm": 0.21763984858989716, + "learning_rate": 2.428543533453086e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007840942963957787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036128197563812135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019294705893844365, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7638155043125152, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19390, + "weighted_flow_matching_loss": 0.02146578636020422, + "weighted_rkd_loss": 0.014737671427428722 + }, + { + "epoch": 5.644457375618272, + "flow_matching_loss": 0.021061462070792913, + "grad_norm": 0.17888763546943665, + "learning_rate": 2.4278941012189215e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00767441987991333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003320483444258571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018669323157519103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.763593065738678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19400, + "weighted_flow_matching_loss": 0.021061462070792913, + "weighted_rkd_loss": 0.014255770109593868 + }, + { + "epoch": 5.647366889729415, + "flow_matching_loss": 0.022096603363752364, + "grad_norm": 0.1987563520669937, + "learning_rate": 2.427244387115029e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007825395232066512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003527467022649944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019178257137537003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.763370567560196, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19410, + "weighted_flow_matching_loss": 0.022096603363752364, + "weighted_rkd_loss": 0.0146401378326118 + }, + { + "epoch": 5.650276403840559, + "flow_matching_loss": 0.022225392423570155, + "grad_norm": 0.2198517918586731, + "learning_rate": 2.4265943913387738e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008030393207445741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003467760724015534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01952854711562395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7631479799747467, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19420, + "weighted_flow_matching_loss": 0.022225392423570155, + "weighted_rkd_loss": 0.014903232175856828 + }, + { + "epoch": 5.653185917951702, + "flow_matching_loss": 0.022118538059294222, + "grad_norm": 0.24824455380439758, + "learning_rate": 2.425944114087608e-05, + "loss": 0.0389, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008951669093221426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004117942904122174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.022021281253546476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7629253387451171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19430, + "weighted_flow_matching_loss": 0.022118538059294222, + "weighted_rkd_loss": 0.016800650767982005 + }, + { + "epoch": 5.656095432062846, + "flow_matching_loss": 0.022631442733109, + "grad_norm": 0.2194591909646988, + "learning_rate": 2.4252935555590684e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00783454766497016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037225983804091813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019391693640500307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7627026259899139, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19440, + "weighted_flow_matching_loss": 0.022631442733109, + "weighted_rkd_loss": 0.014790174458175898 + }, + { + "epoch": 5.659004946173989, + "flow_matching_loss": 0.020322808623313905, + "grad_norm": 0.17422257363796234, + "learning_rate": 2.4246427159507772e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008001460321247577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003655483317561448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01965840421617031, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7624798238277435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19450, + "weighted_flow_matching_loss": 0.020322808623313905, + "weighted_rkd_loss": 0.014989169128239156 + }, + { + "epoch": 5.661914460285132, + "flow_matching_loss": 0.019972536806017162, + "grad_norm": 0.25732719898223877, + "learning_rate": 2.4239915954604433e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076744181104004385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003373788413591683, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018722624704241752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7622569739818573, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19460, + "weighted_flow_matching_loss": 0.019972536806017162, + "weighted_rkd_loss": 0.01427143970504403 + }, + { + "epoch": 5.664823974396276, + "flow_matching_loss": 0.023472917266190052, + "grad_norm": 0.22431287169456482, + "learning_rate": 2.4233401942858595e-05, + "loss": 0.0394, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008551508886739612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003757265210151672, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020860282983630896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7620340287685394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19470, + "weighted_flow_matching_loss": 0.023472917266190052, + "weighted_rkd_loss": 0.015896349120885132 + }, + { + "epoch": 5.6677334885074195, + "flow_matching_loss": 0.02060252781957388, + "grad_norm": 0.1710829734802246, + "learning_rate": 2.4226885126249033e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008122809045016766, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035331199644133447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01977873770520091, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7618110239505768, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19480, + "weighted_flow_matching_loss": 0.02060252781957388, + "weighted_rkd_loss": 0.015067811030894518 + }, + { + "epoch": 5.670643002618562, + "flow_matching_loss": 0.02270900271832943, + "grad_norm": 0.2663765847682953, + "learning_rate": 2.4220365506755394e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008030989672988653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00370614652056247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01976812593638897, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.761587941646576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19490, + "weighted_flow_matching_loss": 0.02270900271832943, + "weighted_rkd_loss": 0.015055099036544561 + }, + { + "epoch": 5.673552516729706, + "flow_matching_loss": 0.021167760342359544, + "grad_norm": 0.18624675273895264, + "learning_rate": 2.4213843086358166e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007285561552271247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032997621921822427, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017870885599404574, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7613647818565369, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19500, + "weighted_flow_matching_loss": 0.021167760342359544, + "weighted_rkd_loss": 0.013606219179928302 + }, + { + "epoch": 5.67646203084085, + "flow_matching_loss": 0.020273224078118803, + "grad_norm": 0.21673984825611115, + "learning_rate": 2.4207317867038684e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008331097010523082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038710638182237744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02053325781598687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.761141562461853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19510, + "weighted_flow_matching_loss": 0.020273224078118803, + "weighted_rkd_loss": 0.015628634858876465 + }, + { + "epoch": 5.679371544951993, + "flow_matching_loss": 0.020178500190377237, + "grad_norm": 0.209200918674469, + "learning_rate": 2.4200789850779137e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007290975330397487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003225869336165488, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01780781978741288, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7609182715415954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19520, + "weighted_flow_matching_loss": 0.020178500190377237, + "weighted_rkd_loss": 0.013550320826470851 + }, + { + "epoch": 5.682281059063136, + "flow_matching_loss": 0.021222117356956005, + "grad_norm": 0.26634666323661804, + "learning_rate": 2.419425903956257e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008030081586912274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003567878878675401, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019628041703253985, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7606949090957642, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19530, + "weighted_flow_matching_loss": 0.021222117356956005, + "weighted_rkd_loss": 0.014930875319987536 + }, + { + "epoch": 5.68519057317428, + "flow_matching_loss": 0.02198485042899847, + "grad_norm": 0.20301410555839539, + "learning_rate": 2.4187725435372865e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008224444137886167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036112438421696423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020060132071375848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7604714691638946, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19540, + "weighted_flow_matching_loss": 0.02198485042899847, + "weighted_rkd_loss": 0.0152551282197237 + }, + { + "epoch": 5.688100087285424, + "flow_matching_loss": 0.022440985031425952, + "grad_norm": 0.21994778513908386, + "learning_rate": 2.418118904019476e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007524979673326016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003241442469879985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018291401863098144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7602479696273804, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19550, + "weighted_flow_matching_loss": 0.022440985031425952, + "weighted_rkd_loss": 0.013905973359942436 + }, + { + "epoch": 5.6910096013965665, + "flow_matching_loss": 0.022462536580860615, + "grad_norm": 0.2238919883966446, + "learning_rate": 2.4174649856013842e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008042317628860474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036250366130843757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01970967184752226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7600243747234344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19560, + "weighted_flow_matching_loss": 0.022462536580860615, + "weighted_rkd_loss": 0.014979769010096788 + }, + { + "epoch": 5.69391911550771, + "flow_matching_loss": 0.02210095841437578, + "grad_norm": 0.25196561217308044, + "learning_rate": 2.416810788481654e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007860629446804523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036087348125874996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019329993799328805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7598007261753082, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19570, + "weighted_flow_matching_loss": 0.02210095841437578, + "weighted_rkd_loss": 0.014686975628137589 + }, + { + "epoch": 5.696828629618854, + "flow_matching_loss": 0.022509424202144145, + "grad_norm": 0.22514119744300842, + "learning_rate": 2.416156312859013e-05, + "loss": 0.0388, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008784854225814343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038573341909795998, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021427042596042157, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7595770001411438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19580, + "weighted_flow_matching_loss": 0.022509424202144145, + "weighted_rkd_loss": 0.016275527235120536 + }, + { + "epoch": 5.699738143729997, + "flow_matching_loss": 0.021731524914503097, + "grad_norm": 0.1967538297176361, + "learning_rate": 2.4155015589322742e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007695308234542608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035276984097436072, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018918314762413502, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7593532145023346, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19590, + "weighted_flow_matching_loss": 0.021731524914503097, + "weighted_rkd_loss": 0.014365669433027506 + }, + { + "epoch": 5.7026476578411405, + "flow_matching_loss": 0.022678390704095365, + "grad_norm": 0.22067797183990479, + "learning_rate": 2.4148465269003338e-05, + "loss": 0.038, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008352390863001347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035225859144702555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0202273678034544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7591293394565582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19600, + "weighted_flow_matching_loss": 0.022678390704095365, + "weighted_rkd_loss": 0.015355344209820033 + }, + { + "epoch": 5.705557171952284, + "flow_matching_loss": 0.02229987978935242, + "grad_norm": 0.18868021667003632, + "learning_rate": 2.4141912169621742e-05, + "loss": 0.0377, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008438280411064624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003400827432051301, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020277388393878937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7589054107666016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19610, + "weighted_flow_matching_loss": 0.02229987978935242, + "weighted_rkd_loss": 0.015388683415949345 + }, + { + "epoch": 5.708466686063428, + "flow_matching_loss": 0.021529543027281762, + "grad_norm": 0.20034213364124298, + "learning_rate": 2.4135356293168602e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008663066010922194, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037291253218427302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02105525732040405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7586813986301422, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19620, + "weighted_flow_matching_loss": 0.021529543027281762, + "weighted_rkd_loss": 0.015974268782883883 + }, + { + "epoch": 5.711376200174571, + "flow_matching_loss": 0.023064996302127837, + "grad_norm": 0.19133062660694122, + "learning_rate": 2.4128797641635427e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007459300523623824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032516690203920006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018170269951224328, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7584573030471802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19630, + "weighted_flow_matching_loss": 0.023064996302127837, + "weighted_rkd_loss": 0.013781390339136123 + }, + { + "epoch": 5.714285714285714, + "flow_matching_loss": 0.021942598931491374, + "grad_norm": 0.21779242157936096, + "learning_rate": 2.412223621701456e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007990107219666243, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003218638035468757, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019198852218687536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7582331717014312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19640, + "weighted_flow_matching_loss": 0.021942598931491374, + "weighted_rkd_loss": 0.014557203464210033 + }, + { + "epoch": 5.717195228396857, + "flow_matching_loss": 0.023283697478473187, + "grad_norm": 0.17374809086322784, + "learning_rate": 2.4115672021299193e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007934867683798075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003278143238276243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019147878512740135, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7580089569091797, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19650, + "weighted_flow_matching_loss": 0.023283697478473187, + "weighted_rkd_loss": 0.014514299388974906 + }, + { + "epoch": 5.720104742508001, + "flow_matching_loss": 0.02353608924895525, + "grad_norm": 0.23614740371704102, + "learning_rate": 2.4109105056483345e-05, + "loss": 0.0404, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008950962405651807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.004329485655762255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02223141025751829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7577846586704254, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19660, + "weighted_flow_matching_loss": 0.02353608924895525, + "weighted_rkd_loss": 0.01684654103592038 + }, + { + "epoch": 5.723014256619145, + "flow_matching_loss": 0.021905521303415297, + "grad_norm": 0.1808859258890152, + "learning_rate": 2.4102535324561898e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007603083783760667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00332285191398114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018529019318521023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7575603067874909, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19670, + "weighted_flow_matching_loss": 0.021905521303415297, + "weighted_rkd_loss": 0.014036939013749361 + }, + { + "epoch": 5.725923770730288, + "flow_matching_loss": 0.022831321321427824, + "grad_norm": 0.16924238204956055, + "learning_rate": 2.4095962827530556e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008610710641369224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003778863837942481, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021000285167247057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7573358654975891, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19680, + "weighted_flow_matching_loss": 0.022831321321427824, + "weighted_rkd_loss": 0.015904291812330485 + }, + { + "epoch": 5.728833284841431, + "flow_matching_loss": 0.02208588309586048, + "grad_norm": 0.22568760812282562, + "learning_rate": 2.408938756738587e-05, + "loss": 0.0378, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00854822196997702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003714790567755699, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020811234600842, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7571113646030426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19690, + "weighted_flow_matching_loss": 0.02208588309586048, + "weighted_rkd_loss": 0.01575650740414858 + }, + { + "epoch": 5.731742798952575, + "flow_matching_loss": 0.02239637579768896, + "grad_norm": 0.2677585184574127, + "learning_rate": 2.4082809546125235e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00820966213941574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035299867624416946, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01994931111112237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7568867862224579, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19700, + "weighted_flow_matching_loss": 0.02239637579768896, + "weighted_rkd_loss": 0.015099436603486538 + }, + { + "epoch": 5.734652313063719, + "flow_matching_loss": 0.020772461593151093, + "grad_norm": 0.19112934172153473, + "learning_rate": 2.4076228765746876e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007953960541635751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003457198734395206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019365119561553002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7566621601581573, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19710, + "weighted_flow_matching_loss": 0.020772461593151093, + "weighted_rkd_loss": 0.014652858395129443 + }, + { + "epoch": 5.737561827174861, + "flow_matching_loss": 0.02061183676123619, + "grad_norm": 0.233080193400383, + "learning_rate": 2.4069645228249864e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007797629805281758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033551965141668917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01895045628771186, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7564374387264252, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19720, + "weighted_flow_matching_loss": 0.02061183676123619, + "weighted_rkd_loss": 0.014334853924810886 + }, + { + "epoch": 5.740471341286005, + "flow_matching_loss": 0.022530146315693854, + "grad_norm": 0.22054077684879303, + "learning_rate": 2.40630589356341e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007272833120077848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003282203455455601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01782786976546049, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7562126517295837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19730, + "weighted_flow_matching_loss": 0.022530146315693854, + "weighted_rkd_loss": 0.01348168570548296 + }, + { + "epoch": 5.743380855397149, + "flow_matching_loss": 0.021802037488669157, + "grad_norm": 0.192035511136055, + "learning_rate": 2.4056469889900327e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008033522590994834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003421393618918955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01948843877762556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.755987799167633, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19740, + "weighted_flow_matching_loss": 0.021802037488669157, + "weighted_rkd_loss": 0.014733089692890644 + }, + { + "epoch": 5.7462903695082925, + "flow_matching_loss": 0.022218183614313602, + "grad_norm": 0.18269284069538116, + "learning_rate": 2.404987809305012e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008123244578018784, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037909553386271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02003744449466467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7557628691196442, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19750, + "weighted_flow_matching_loss": 0.022218183614313602, + "weighted_rkd_loss": 0.015143559593707324 + }, + { + "epoch": 5.749199883619435, + "flow_matching_loss": 0.02141941515728831, + "grad_norm": 0.16873681545257568, + "learning_rate": 2.4043283547085903e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007871738588437437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003398381266742945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019141858257353307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7555378794670105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19760, + "weighted_flow_matching_loss": 0.02141941515728831, + "weighted_rkd_loss": 0.014462408982217312 + }, + { + "epoch": 5.752109397730579, + "flow_matching_loss": 0.021438808366656303, + "grad_norm": 0.211658775806427, + "learning_rate": 2.403668625401092e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00812261519022286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003547299187630415, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01979252975434065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7553128123283386, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19770, + "weighted_flow_matching_loss": 0.021438808366656303, + "weighted_rkd_loss": 0.014949662704020738 + }, + { + "epoch": 5.755018911841723, + "flow_matching_loss": 0.022171317413449288, + "grad_norm": 0.22421212494373322, + "learning_rate": 2.4030086215829247e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076829144731163975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033639976754784584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018729826621711253, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7550876796245575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19780, + "weighted_flow_matching_loss": 0.022171317413449288, + "weighted_rkd_loss": 0.0141426264308393 + }, + { + "epoch": 5.757928425952866, + "flow_matching_loss": 0.02153978142887354, + "grad_norm": 0.19891537725925446, + "learning_rate": 2.4023483434545806e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008054666733369232, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035389929777011274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019648326188325883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7548624753952027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19790, + "weighted_flow_matching_loss": 0.02153978142887354, + "weighted_rkd_loss": 0.01483179060742259 + }, + { + "epoch": 5.760837940064009, + "flow_matching_loss": 0.02383688520640135, + "grad_norm": 0.20470857620239258, + "learning_rate": 2.4016877912166356e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007115353969857097, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031243885634467004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017355096712708473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7546371817588806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19800, + "weighted_flow_matching_loss": 0.02383688520640135, + "weighted_rkd_loss": 0.013096848502755165 + }, + { + "epoch": 5.763747454175153, + "flow_matching_loss": 0.022817962896078826, + "grad_norm": 0.20606017112731934, + "learning_rate": 2.401026965069747e-05, + "loss": 0.0387, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00859678634442389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0038470457773655655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021040618419647217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7544118523597717, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19810, + "weighted_flow_matching_loss": 0.022817962896078826, + "weighted_rkd_loss": 0.015873241983354092 + }, + { + "epoch": 5.766656968286296, + "flow_matching_loss": 0.022915918938815595, + "grad_norm": 0.19532093405723572, + "learning_rate": 2.4003658652146564e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007293311227113008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030928505584597588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017679473385214806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7541864395141602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19820, + "weighted_flow_matching_loss": 0.022915918938815595, + "weighted_rkd_loss": 0.01333371214568615 + }, + { + "epoch": 5.7695664823974395, + "flow_matching_loss": 0.021769660525023937, + "grad_norm": 0.2028866559267044, + "learning_rate": 2.3997044918521896e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008259581262245774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003373665921390057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01989282835274935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7539609551429749, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19830, + "weighted_flow_matching_loss": 0.021769660525023937, + "weighted_rkd_loss": 0.01499849809333682 + }, + { + "epoch": 5.772475996508583, + "flow_matching_loss": 0.022402596659958363, + "grad_norm": 0.19815227389335632, + "learning_rate": 2.3990428451832524e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007504415279254317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003181739756837487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018190570268779994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7537353992462158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19840, + "weighted_flow_matching_loss": 0.022402596659958363, + "weighted_rkd_loss": 0.013710945192724467 + }, + { + "epoch": 5.775385510619726, + "flow_matching_loss": 0.021014434285461903, + "grad_norm": 0.22588099539279938, + "learning_rate": 2.398380925408837e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076683245599269865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003290615417063236, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018627264723181724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.753509783744812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19850, + "weighted_flow_matching_loss": 0.021014434285461903, + "weighted_rkd_loss": 0.014035762287676334 + }, + { + "epoch": 5.77829502473087, + "flow_matching_loss": 0.021671270579099657, + "grad_norm": 0.1954067051410675, + "learning_rate": 2.3977187327300174e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007233710447326303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031403800239786506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017607800848782062, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.75328409075737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19860, + "weighted_flow_matching_loss": 0.021671270579099657, + "weighted_rkd_loss": 0.013263772893697023 + }, + { + "epoch": 5.7812045388420135, + "flow_matching_loss": 0.022634055558592082, + "grad_norm": 0.2481154501438141, + "learning_rate": 2.397056267347949e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008292063232511282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003717146161943674, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020301272440701724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7530583202838897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19870, + "weighted_flow_matching_loss": 0.022634055558592082, + "weighted_rkd_loss": 0.015287902113050223 + }, + { + "epoch": 5.784114052953157, + "flow_matching_loss": 0.02223490960896015, + "grad_norm": 0.21235151588916779, + "learning_rate": 2.3963935294638725e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006956683425232768, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002890244545415044, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016803611442446707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7528325021266937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19880, + "weighted_flow_matching_loss": 0.02223490960896015, + "weighted_rkd_loss": 0.012650281190872192 + }, + { + "epoch": 5.7870235670643, + "flow_matching_loss": 0.0216168662533164, + "grad_norm": 0.22924789786338806, + "learning_rate": 2.395730519279109e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00762092755176127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003359267604537308, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018601122498512267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.752606600522995, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19890, + "weighted_flow_matching_loss": 0.0216168662533164, + "weighted_rkd_loss": 0.013999368250370025 + }, + { + "epoch": 5.789933081175444, + "flow_matching_loss": 0.021453370712697505, + "grad_norm": 0.1960206925868988, + "learning_rate": 2.3950672369950646e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007240303652361035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003043960710056126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01752456808462739, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.752380621433258, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19900, + "weighted_flow_matching_loss": 0.021453370712697505, + "weighted_rkd_loss": 0.013185194879770278 + }, + { + "epoch": 5.792842595286587, + "flow_matching_loss": 0.022349401749670504, + "grad_norm": 0.2720269560813904, + "learning_rate": 2.3944036828132266e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008079049084335566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035306657431647182, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01968876402825117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7521545827388764, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19910, + "weighted_flow_matching_loss": 0.022349401749670504, + "weighted_rkd_loss": 0.01480900039896369 + }, + { + "epoch": 5.79575210939773, + "flow_matching_loss": 0.02211491670459509, + "grad_norm": 0.18583133816719055, + "learning_rate": 2.3937398569351647e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007194208446890116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032229425618425013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017611359339207412, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7519284784793854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19920, + "weighted_flow_matching_loss": 0.02211491670459509, + "weighted_rkd_loss": 0.013242605980485678 + }, + { + "epoch": 5.798661623508874, + "flow_matching_loss": 0.0201772166416049, + "grad_norm": 0.2273213416337967, + "learning_rate": 2.3930757595625326e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007725839642807842, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032200086396187546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018671687971800566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7517023086547852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19930, + "weighted_flow_matching_loss": 0.0201772166416049, + "weighted_rkd_loss": 0.014035377930849791 + }, + { + "epoch": 5.801571137620018, + "flow_matching_loss": 0.02046709358692169, + "grad_norm": 0.17761245369911194, + "learning_rate": 2.3924113908970645e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007388030784204603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032120966352522373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017988158203661442, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7514760553836822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19940, + "weighted_flow_matching_loss": 0.02046709358692169, + "weighted_rkd_loss": 0.013517727144062519 + }, + { + "epoch": 5.804480651731161, + "flow_matching_loss": 0.021339476481080054, + "grad_norm": 0.20722058415412903, + "learning_rate": 2.391746751140579e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007378701493144036, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032086637103930114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01796606685966253, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7512497425079345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19950, + "weighted_flow_matching_loss": 0.021339476481080054, + "weighted_rkd_loss": 0.0134969268925488 + }, + { + "epoch": 5.807390165842304, + "flow_matching_loss": 0.0228892819955945, + "grad_norm": 0.2636241316795349, + "learning_rate": 2.3910818404949756e-05, + "loss": 0.0393, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008809814602136612, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0042327088303864, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02185233812779188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7510233521461487, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19960, + "weighted_flow_matching_loss": 0.0228892819955945, + "weighted_rkd_loss": 0.016411678027361633 + }, + { + "epoch": 5.810299679953448, + "flow_matching_loss": 0.022627967409789562, + "grad_norm": 0.22317656874656677, + "learning_rate": 2.3904166591622367e-05, + "loss": 0.0372, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00797261856496334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035173411713913085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019462578278034926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.750796914100647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19970, + "weighted_flow_matching_loss": 0.022627967409789562, + "weighted_rkd_loss": 0.014612461626529693 + }, + { + "epoch": 5.813209194064592, + "flow_matching_loss": 0.022213488817214966, + "grad_norm": 0.20318366587162018, + "learning_rate": 2.3897512073444267e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008339876867830753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035161178559064863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020195871591567993, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7505703806877136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19980, + "weighted_flow_matching_loss": 0.022213488817214966, + "weighted_rkd_loss": 0.015158386155962943 + }, + { + "epoch": 5.816118708175734, + "flow_matching_loss": 0.023492151126265525, + "grad_norm": 0.1871214509010315, + "learning_rate": 2.389085485243693e-05, + "loss": 0.0379, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007857165625318885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035055536776781083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019219885021448134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7503437936306, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 19990, + "weighted_flow_matching_loss": 0.023492151126265525, + "weighted_rkd_loss": 0.01442153351381421 + }, + { + "epoch": 5.819028222286878, + "flow_matching_loss": 0.022710474021732808, + "grad_norm": 0.2089865654706955, + "learning_rate": 2.3884194930622632e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007836295384913683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003414998156949878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019087588973343374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7501171350479126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20000, + "weighted_flow_matching_loss": 0.022710474021732808, + "weighted_rkd_loss": 0.014317905157804489 + }, + { + "epoch": 5.821937736398022, + "flow_matching_loss": 0.023972402140498162, + "grad_norm": 0.236797034740448, + "learning_rate": 2.3877532310024488e-05, + "loss": 0.0382, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007783155934885144, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033746396191418173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01894095158204436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7498904049396515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20010, + "weighted_flow_matching_loss": 0.023972402140498162, + "weighted_rkd_loss": 0.01420367881655693 + }, + { + "epoch": 5.824847250509165, + "flow_matching_loss": 0.022826254647225142, + "grad_norm": 0.20944111049175262, + "learning_rate": 2.3870866992666428e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007994729094207287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003505060006864369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01949451807886362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7496636092662812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20020, + "weighted_flow_matching_loss": 0.022826254647225142, + "weighted_rkd_loss": 0.01461436776444316 + }, + { + "epoch": 5.827756764620308, + "flow_matching_loss": 0.02045508800074458, + "grad_norm": 0.22637858986854553, + "learning_rate": 2.3864198980573196e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008097452344372869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003601398877799511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019796303939074278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7494367361068726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20030, + "weighted_flow_matching_loss": 0.02045508800074458, + "weighted_rkd_loss": 0.014836100488901138 + }, + { + "epoch": 5.830666278731452, + "flow_matching_loss": 0.02186627686023712, + "grad_norm": 0.20990540087223053, + "learning_rate": 2.385752827577036e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008421729458495975, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00380732377525419, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020650782622396946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7492098093032837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20040, + "weighted_flow_matching_loss": 0.02186627686023712, + "weighted_rkd_loss": 0.015471777319908142 + }, + { + "epoch": 5.833575792842595, + "flow_matching_loss": 0.022632733173668384, + "grad_norm": 0.22568298876285553, + "learning_rate": 2.38508548802843e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007473819563165307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032873706659302115, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01823500981554389, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7489828109741211, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20050, + "weighted_flow_matching_loss": 0.022632733173668384, + "weighted_rkd_loss": 0.013657709583640099 + }, + { + "epoch": 5.836485306953739, + "flow_matching_loss": 0.02069179657846689, + "grad_norm": 0.2468022257089615, + "learning_rate": 2.3844178796142226e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008229935960844158, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036024078959599136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020062279887497425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7487557232379913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20060, + "weighted_flow_matching_loss": 0.02069179657846689, + "weighted_rkd_loss": 0.015021731331944465 + }, + { + "epoch": 5.839394821064882, + "flow_matching_loss": 0.020676549710333347, + "grad_norm": 0.2743586003780365, + "learning_rate": 2.383750002537215e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00778540400788188, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032838111044839026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018854619562625886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7485286056995392, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20070, + "weighted_flow_matching_loss": 0.020676549710333347, + "weighted_rkd_loss": 0.014113092795014381 + }, + { + "epoch": 5.842304335176026, + "flow_matching_loss": 0.021449005603790282, + "grad_norm": 0.22440025210380554, + "learning_rate": 2.383081857000291e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008289149962365627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003884399961680174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020462699979543687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7483013927936554, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20080, + "weighted_flow_matching_loss": 0.021449005603790282, + "weighted_rkd_loss": 0.015312182437628507 + }, + { + "epoch": 5.845213849287169, + "flow_matching_loss": 0.02201110925525427, + "grad_norm": 0.1932203322649002, + "learning_rate": 2.3824134432064145e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0078002054709941145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033577936934307217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018958204612135886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7480741202831268, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20090, + "weighted_flow_matching_loss": 0.02201110925525427, + "weighted_rkd_loss": 0.01418223576620221 + }, + { + "epoch": 5.8481233633983125, + "flow_matching_loss": 0.020922796055674552, + "grad_norm": 0.30025628209114075, + "learning_rate": 2.3817447613586332e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00753213744610548, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003452797792851925, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018517072685062884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.747846782207489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20100, + "weighted_flow_matching_loss": 0.020922796055674552, + "weighted_rkd_loss": 0.013847963884472847 + }, + { + "epoch": 5.851032877509456, + "flow_matching_loss": 0.022177997790277005, + "grad_norm": 0.21875064074993134, + "learning_rate": 2.381075811660074e-05, + "loss": 0.0375, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00832999860867858, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037714581238105895, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020431455317884685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.747619366645813, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20110, + "weighted_flow_matching_loss": 0.022177997790277005, + "weighted_rkd_loss": 0.015275002829730511 + }, + { + "epoch": 5.853942391620599, + "flow_matching_loss": 0.023273734003305437, + "grad_norm": 0.1804378777742386, + "learning_rate": 2.3804065943139473e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007645753305405378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031080666230991483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018399573117494582, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7473918855190277, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20120, + "weighted_flow_matching_loss": 0.023273734003305437, + "weighted_rkd_loss": 0.013751833420246839 + }, + { + "epoch": 5.856851905731743, + "flow_matching_loss": 0.019512569159269334, + "grad_norm": 0.2238827794790268, + "learning_rate": 2.379737109523543e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075978631619364025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003333648410625756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018529374711215497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7471643447875976, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20130, + "weighted_flow_matching_loss": 0.019512569159269334, + "weighted_rkd_loss": 0.0138444684445858 + }, + { + "epoch": 5.8597614198428865, + "flow_matching_loss": 0.0211762560531497, + "grad_norm": 0.21682347357273102, + "learning_rate": 2.3790673574922324e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007629245659336448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032985234865918756, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018557014875113964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7469367384910583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20140, + "weighted_flow_matching_loss": 0.0211762560531497, + "weighted_rkd_loss": 0.013860849104821682 + }, + { + "epoch": 5.86267093395403, + "flow_matching_loss": 0.02302105873823166, + "grad_norm": 0.2270888239145279, + "learning_rate": 2.3783973384234695e-05, + "loss": 0.0383, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008473292179405689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003536213142797351, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020482797548174857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7467090487480164, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20150, + "weighted_flow_matching_loss": 0.02302105873823166, + "weighted_rkd_loss": 0.01529460409656167 + }, + { + "epoch": 5.865580448065173, + "flow_matching_loss": 0.02095547057688236, + "grad_norm": 0.18654097616672516, + "learning_rate": 2.3777270525207876e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007165213162079453, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030728761572390794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01740330262109637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.746481305360794, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20160, + "weighted_flow_matching_loss": 0.02095547057688236, + "weighted_rkd_loss": 0.012991317454725504 + }, + { + "epoch": 5.868489962176317, + "flow_matching_loss": 0.02139845658093691, + "grad_norm": 0.1943708062171936, + "learning_rate": 2.3770564999878024e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006907813297584653, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029222679790109397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01673789480701089, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.746253490447998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20170, + "weighted_flow_matching_loss": 0.02139845658093691, + "weighted_rkd_loss": 0.012490724958479405 + }, + { + "epoch": 5.8713994762874595, + "flow_matching_loss": 0.02170031759887934, + "grad_norm": 0.20733574032783508, + "learning_rate": 2.37638568102821e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007794235879555344, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034137214999645947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019002193119376898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7460256040096283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20180, + "weighted_flow_matching_loss": 0.02170031759887934, + "weighted_rkd_loss": 0.01417615581303835 + }, + { + "epoch": 5.874308990398603, + "flow_matching_loss": 0.021600902173668148, + "grad_norm": 0.15930035710334778, + "learning_rate": 2.3757145958457868e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006859168736264109, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002984604611992836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01670294236391783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7457976579666138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20190, + "weighted_flow_matching_loss": 0.021600902173668148, + "weighted_rkd_loss": 0.012457042839378119 + }, + { + "epoch": 5.877218504509747, + "flow_matching_loss": 0.020458504743874072, + "grad_norm": 0.2059718519449234, + "learning_rate": 2.375043244644392e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008330300077795982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003804047731682658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020464648306369782, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7455696344375611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20200, + "weighted_flow_matching_loss": 0.020458504743874072, + "weighted_rkd_loss": 0.015257739555090665 + }, + { + "epoch": 5.880128018620891, + "flow_matching_loss": 0.02164372457191348, + "grad_norm": 0.21521610021591187, + "learning_rate": 2.374371627627963e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007441653171554208, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032047244952991606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0180880312807858, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7453415513038635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20210, + "weighted_flow_matching_loss": 0.02164372457191348, + "weighted_rkd_loss": 0.013481756485998631 + }, + { + "epoch": 5.8830375327320334, + "flow_matching_loss": 0.02137458398938179, + "grad_norm": 0.23221458494663239, + "learning_rate": 2.37369974500052e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008862219983711839, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003708694176748395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.021433134190738203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7451133906841279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20220, + "weighted_flow_matching_loss": 0.02137458398938179, + "weighted_rkd_loss": 0.015969956759363413 + }, + { + "epoch": 5.885947046843177, + "flow_matching_loss": 0.02032565139234066, + "grad_norm": 0.19003070890903473, + "learning_rate": 2.373027596966164e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007698220713064074, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003458424611017108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018854865804314613, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7448851764202118, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20230, + "weighted_flow_matching_loss": 0.02032565139234066, + "weighted_rkd_loss": 0.014044710155576467 + }, + { + "epoch": 5.888856560954321, + "flow_matching_loss": 0.020994689874351026, + "grad_norm": 0.2111099362373352, + "learning_rate": 2.3723551837290744e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007440800406038761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032474500592797996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01812905129045248, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.744656890630722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20240, + "weighted_flow_matching_loss": 0.020994689874351026, + "weighted_rkd_loss": 0.013499897345900536 + }, + { + "epoch": 5.891766075065464, + "flow_matching_loss": 0.022443260997533798, + "grad_norm": 0.2175777703523636, + "learning_rate": 2.3716825054935133e-05, + "loss": 0.0374, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008278163615614176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003597674937918782, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.02015400193631649, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.744428527355194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20250, + "weighted_flow_matching_loss": 0.022443260997533798, + "weighted_rkd_loss": 0.015003235172480344 + }, + { + "epoch": 5.894675589176607, + "flow_matching_loss": 0.020767354313284157, + "grad_norm": 0.23974180221557617, + "learning_rate": 2.371009562463822e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007514042872935533, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032317652134224773, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018259850703179835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7442001044750214, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20260, + "weighted_flow_matching_loss": 0.020767354313284157, + "weighted_rkd_loss": 0.01358906114473939 + }, + { + "epoch": 5.897585103287751, + "flow_matching_loss": 0.02203256096690893, + "grad_norm": 0.19032692909240723, + "learning_rate": 2.3703363548444238e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007385336374863982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003227434353902936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017998107243329287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7439716219902038, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20270, + "weighted_flow_matching_loss": 0.02203256096690893, + "weighted_rkd_loss": 0.013390162959694862 + }, + { + "epoch": 5.900494617398895, + "flow_matching_loss": 0.024217881634831427, + "grad_norm": 0.23805676400661469, + "learning_rate": 2.3696628828398208e-05, + "loss": 0.0395, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008362953178584575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003861447819508612, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0205873541533947, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7437430620193481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20280, + "weighted_flow_matching_loss": 0.024217881634831427, + "weighted_rkd_loss": 0.015311675425618888 + }, + { + "epoch": 5.903404131510038, + "flow_matching_loss": 0.022201311215758322, + "grad_norm": 0.24457137286663055, + "learning_rate": 2.368989146654596e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00830084211193025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037350859958678482, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020336769893765448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7435144424438477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20290, + "weighted_flow_matching_loss": 0.022201311215758322, + "weighted_rkd_loss": 0.015120580419898033 + }, + { + "epoch": 5.906313645621181, + "flow_matching_loss": 0.019972415827214717, + "grad_norm": 0.20238378643989563, + "learning_rate": 2.368315146493413e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007522965501993895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035251139895990493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018571045063436032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7432857573032379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20300, + "weighted_flow_matching_loss": 0.019972415827214717, + "weighted_rkd_loss": 0.013803640846163034 + }, + { + "epoch": 5.909223159732325, + "flow_matching_loss": 0.02226103115826845, + "grad_norm": 0.18254642188549042, + "learning_rate": 2.367640882561015e-05, + "loss": 0.0367, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007933600293472409, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035026630386710165, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019369863625615834, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7430569946765899, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20310, + "weighted_flow_matching_loss": 0.02226103115826845, + "weighted_rkd_loss": 0.014393042400479316 + }, + { + "epoch": 5.912132673843468, + "flow_matching_loss": 0.02258387580513954, + "grad_norm": 0.22569391131401062, + "learning_rate": 2.3669663550622255e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077344354707747694, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003347554081119597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018816425185650586, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7428281903266907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20320, + "weighted_flow_matching_loss": 0.02258387580513954, + "weighted_rkd_loss": 0.013977244310081006 + }, + { + "epoch": 5.915042187954612, + "flow_matching_loss": 0.022495236061513425, + "grad_norm": 0.20095348358154297, + "learning_rate": 2.3662915642019482e-05, + "loss": 0.0376, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008371456060558558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035588515223935247, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020301763899624348, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7425992906093597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20330, + "weighted_flow_matching_loss": 0.022495236061513425, + "weighted_rkd_loss": 0.015075959917157888 + }, + { + "epoch": 5.917951702065755, + "flow_matching_loss": 0.021701168455183505, + "grad_norm": 0.15963245928287506, + "learning_rate": 2.365616510185167e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007660315837711096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035016425186768173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01882227435708046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7423703372478485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20340, + "weighted_flow_matching_loss": 0.021701168455183505, + "weighted_rkd_loss": 0.013973121251910924 + }, + { + "epoch": 5.920861216176898, + "flow_matching_loss": 0.02289236355572939, + "grad_norm": 0.22487133741378784, + "learning_rate": 2.364941193216945e-05, + "loss": 0.0373, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007970808260142802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003474669181741774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019416285771876575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7421413123607635, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20350, + "weighted_flow_matching_loss": 0.02289236355572939, + "weighted_rkd_loss": 0.01440962366759777 + }, + { + "epoch": 5.923770730288042, + "flow_matching_loss": 0.021410900354385375, + "grad_norm": 0.1729166954755783, + "learning_rate": 2.364265613502426e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007487972546368837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003328589117154479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018304534535855056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7419122219085693, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20360, + "weighted_flow_matching_loss": 0.021410900354385375, + "weighted_rkd_loss": 0.013580505270510912 + }, + { + "epoch": 5.9266802443991855, + "flow_matching_loss": 0.020794909819960593, + "grad_norm": 0.18961335718631744, + "learning_rate": 2.3635897712468333e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00788453477434814, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003451762651093304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019220832176506518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7416830658912659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20370, + "weighted_flow_matching_loss": 0.020794909819960593, + "weighted_rkd_loss": 0.014255684148520232 + }, + { + "epoch": 5.929589758510328, + "flow_matching_loss": 0.02115774005651474, + "grad_norm": 0.2055976688861847, + "learning_rate": 2.36291366665547e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072316228412091735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030682444805279372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017531490232795478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7414538443088532, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20380, + "weighted_flow_matching_loss": 0.02115774005651474, + "weighted_rkd_loss": 0.012998773530125619 + }, + { + "epoch": 5.932499272621472, + "flow_matching_loss": 0.020928390324115753, + "grad_norm": 0.19511736929416656, + "learning_rate": 2.362237299933719e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007929823594167829, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034368981840088965, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019296545535326004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7412245571613312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20390, + "weighted_flow_matching_loss": 0.020928390324115753, + "weighted_rkd_loss": 0.01430314714089036 + }, + { + "epoch": 5.935408786732616, + "flow_matching_loss": 0.021262344717979432, + "grad_norm": 0.21904702484607697, + "learning_rate": 2.3615606712870414e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007701802207157016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033748931251466273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01877849753946066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7409952104091644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20400, + "weighted_flow_matching_loss": 0.021262344717979432, + "weighted_rkd_loss": 0.01391468346118927 + }, + { + "epoch": 5.9383183008437594, + "flow_matching_loss": 0.022419271059334278, + "grad_norm": 0.20447982847690582, + "learning_rate": 2.3608837809209807e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007404054887592793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003062336309812963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01787044620141387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.740765780210495, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20410, + "weighted_flow_matching_loss": 0.022419271059334278, + "weighted_rkd_loss": 0.013237777817994355 + }, + { + "epoch": 5.941227814954902, + "flow_matching_loss": 0.021124495193362236, + "grad_norm": 0.17793963849544525, + "learning_rate": 2.360206629041157e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072667705826461315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034965373342856765, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018030078522861002, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7405363023281097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20420, + "weighted_flow_matching_loss": 0.021124495193362236, + "weighted_rkd_loss": 0.013351924531161786 + }, + { + "epoch": 5.944137329066046, + "flow_matching_loss": 0.02063941825181246, + "grad_norm": 0.19989940524101257, + "learning_rate": 2.359529215853273e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007772002508863807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034253364661708473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01896934164687991, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7403067469596862, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20430, + "weighted_flow_matching_loss": 0.02063941825181246, + "weighted_rkd_loss": 0.014043040666729211 + }, + { + "epoch": 5.94704684317719, + "flow_matching_loss": 0.0228162195533514, + "grad_norm": 0.23042133450508118, + "learning_rate": 2.358851541563107e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007709909370169044, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034236594568938018, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018843477964401244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7400771319866181, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20440, + "weighted_flow_matching_loss": 0.0228162195533514, + "weighted_rkd_loss": 0.013945632986724377 + }, + { + "epoch": 5.9499563572883325, + "flow_matching_loss": 0.02110069589689374, + "grad_norm": 0.18700549006462097, + "learning_rate": 2.3581736063765193e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007472277618944645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003296027239412069, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01824058247730136, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7398474514484406, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20450, + "weighted_flow_matching_loss": 0.02110069589689374, + "weighted_rkd_loss": 0.013495348393917084 + }, + { + "epoch": 5.952865871399476, + "flow_matching_loss": 0.02153353877365589, + "grad_norm": 0.26572155952453613, + "learning_rate": 2.3574954104994492e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007095769979059696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003127919603139162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017319459468126297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7396177053451538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20460, + "weighted_flow_matching_loss": 0.02153353877365589, + "weighted_rkd_loss": 0.01280974904075265 + }, + { + "epoch": 5.95577538551062, + "flow_matching_loss": 0.02129007214680314, + "grad_norm": 0.1741139441728592, + "learning_rate": 2.3568169541379137e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008061627857387066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035537197487428784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01967697562649846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7393878936767578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20470, + "weighted_flow_matching_loss": 0.02129007214680314, + "weighted_rkd_loss": 0.014548982214182615 + }, + { + "epoch": 5.958684899621764, + "flow_matching_loss": 0.019777664728462697, + "grad_norm": 0.22446152567863464, + "learning_rate": 2.356138237498011e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007386954082176089, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003230623132549226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018004531413316725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7391580045223236, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20480, + "weighted_flow_matching_loss": 0.019777664728462697, + "weighted_rkd_loss": 0.013308058492839336 + }, + { + "epoch": 5.961594413732906, + "flow_matching_loss": 0.02109465803951025, + "grad_norm": 0.1685689240694046, + "learning_rate": 2.3554592607859166e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008090482745319604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034266294213011863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01960759488865733, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7389280557632446, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20490, + "weighted_flow_matching_loss": 0.02109465803951025, + "weighted_rkd_loss": 0.014488656353205443 + }, + { + "epoch": 5.96450392784405, + "flow_matching_loss": 0.02099911533296108, + "grad_norm": 0.25209492444992065, + "learning_rate": 2.3547800242078857e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007885619811713696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003517100494354963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019288340117782354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7386980593204499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20500, + "weighted_flow_matching_loss": 0.02099911533296108, + "weighted_rkd_loss": 0.014248123485594988 + }, + { + "epoch": 5.967413441955194, + "flow_matching_loss": 0.021280237287282944, + "grad_norm": 0.18103620409965515, + "learning_rate": 2.3541005279702528e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007638905849307776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033348251366987825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01861263681203127, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7384679794311524, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20510, + "weighted_flow_matching_loss": 0.021280237287282944, + "weighted_rkd_loss": 0.013744874019175768 + }, + { + "epoch": 5.970322956066337, + "flow_matching_loss": 0.022082566656172274, + "grad_norm": 0.20729291439056396, + "learning_rate": 2.3534207722794307e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007978576142340899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003496809909120202, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01945396214723587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7382378339767456, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20520, + "weighted_flow_matching_loss": 0.022082566656172274, + "weighted_rkd_loss": 0.014361713826656342 + }, + { + "epoch": 5.97323247017748, + "flow_matching_loss": 0.021210609562695028, + "grad_norm": 0.20498166978359222, + "learning_rate": 2.352740757341911e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007130858954042196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028546213638037444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017116339411586523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7380076289176941, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20530, + "weighted_flow_matching_loss": 0.021210609562695028, + "weighted_rkd_loss": 0.012632057536393404 + }, + { + "epoch": 5.976141984288624, + "flow_matching_loss": 0.022492024302482604, + "grad_norm": 0.19338497519493103, + "learning_rate": 2.3520604833642645e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007631750032305718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034305243054404853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018694024346768857, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7377773523330688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20540, + "weighted_flow_matching_loss": 0.022492024302482604, + "weighted_rkd_loss": 0.013792014494538308 + }, + { + "epoch": 5.979051498399767, + "flow_matching_loss": 0.020623210165649653, + "grad_norm": 0.24266113340854645, + "learning_rate": 2.3513799505531407e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007641369616612792, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033712673932313917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018654006347060204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7375470101833344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20550, + "weighted_flow_matching_loss": 0.020623210165649653, + "weighted_rkd_loss": 0.013758266530930995 + }, + { + "epoch": 5.981961012510911, + "flow_matching_loss": 0.02246137671172619, + "grad_norm": 0.20359580218791962, + "learning_rate": 2.350699159115268e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007860422553494572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003420950146391988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019141795672476293, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7373166143894195, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20560, + "weighted_flow_matching_loss": 0.02246137671172619, + "weighted_rkd_loss": 0.014113665651530027 + }, + { + "epoch": 5.984870526622054, + "flow_matching_loss": 0.02251790463924408, + "grad_norm": 0.23485159873962402, + "learning_rate": 2.350018109257451e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00796267087571323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034683569800108673, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019393698778003454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.737086147069931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20570, + "weighted_flow_matching_loss": 0.02251790463924408, + "weighted_rkd_loss": 0.014294881466776132 + }, + { + "epoch": 5.987780040733197, + "flow_matching_loss": 0.020547594502568246, + "grad_norm": 0.21383152902126312, + "learning_rate": 2.3493368011865766e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00827793162316084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035900500370189548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020145913306623697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7368556141853333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20580, + "weighted_flow_matching_loss": 0.020547594502568246, + "weighted_rkd_loss": 0.014844568260014058 + }, + { + "epoch": 5.990689554844341, + "flow_matching_loss": 0.020348888728767632, + "grad_norm": 0.2336784303188324, + "learning_rate": 2.348655235109607e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00797261856496334, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034524769056588412, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01939771417528391, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7366250157356262, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20590, + "weighted_flow_matching_loss": 0.020348888728767632, + "weighted_rkd_loss": 0.014288695342838765 + }, + { + "epoch": 5.9935990689554846, + "flow_matching_loss": 0.020705951564013957, + "grad_norm": 0.20481809973716736, + "learning_rate": 2.3479734112335845e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00803267559967935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035806097788736225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01964596062898636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7363943576812744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20600, + "weighted_flow_matching_loss": 0.020705951564013957, + "weighted_rkd_loss": 0.014467168413102627 + }, + { + "epoch": 5.996508583066628, + "flow_matching_loss": 0.020708880852907896, + "grad_norm": 0.16244804859161377, + "learning_rate": 2.347291329765629e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00746184759773314, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00326370766852051, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018187403213232756, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7361636221408844, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20610, + "weighted_flow_matching_loss": 0.020708880852907896, + "weighted_rkd_loss": 0.013388983067125082 + }, + { + "epoch": 5.999418097177771, + "flow_matching_loss": 0.020969359576702117, + "grad_norm": 0.1893804520368576, + "learning_rate": 2.3466089909129396e-05, + "loss": 0.036, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008331997785717249, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037238119868561624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020387807860970497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7359328269958496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20620, + "weighted_flow_matching_loss": 0.020969359576702117, + "weighted_rkd_loss": 0.015004113130271435 + }, + { + "epoch": 6.002327611288915, + "flow_matching_loss": 0.021079814247786998, + "grad_norm": 0.18188802897930145, + "learning_rate": 2.3459263948827916e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071600352879613634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00316686094738543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017486931569874287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7357019782066345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20630, + "weighted_flow_matching_loss": 0.021079814247786998, + "weighted_rkd_loss": 0.012865124549716712 + }, + { + "epoch": 6.0052371254000585, + "flow_matching_loss": 0.020822915155440568, + "grad_norm": 0.2322741150856018, + "learning_rate": 2.3452435418825404e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007632883125916123, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003307805536314845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01857357155531645, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7354710519313812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20640, + "weighted_flow_matching_loss": 0.020822915155440568, + "weighted_rkd_loss": 0.01366020506247878 + }, + { + "epoch": 6.008146639511201, + "flow_matching_loss": 0.022170011699199677, + "grad_norm": 0.18707957863807678, + "learning_rate": 2.344560432119618e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007369061559438705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032442432129755616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017982366401702164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7352400600910187, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20650, + "weighted_flow_matching_loss": 0.022170011699199677, + "weighted_rkd_loss": 0.013221298716962338 + }, + { + "epoch": 6.011056153622345, + "flow_matching_loss": 0.021069298312067986, + "grad_norm": 0.1950387805700302, + "learning_rate": 2.3438770658015368e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006716775847598911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002840662794187665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016274214535951615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7350090205669403, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20660, + "weighted_flow_matching_loss": 0.021069298312067986, + "weighted_rkd_loss": 0.011961596086621284 + }, + { + "epoch": 6.013965667733489, + "flow_matching_loss": 0.02267466839402914, + "grad_norm": 0.20446418225765228, + "learning_rate": 2.3431934431358835e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008024457935243845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035230773501098156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019571992941200732, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7347778975963593, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20670, + "weighted_flow_matching_loss": 0.02267466839402914, + "weighted_rkd_loss": 0.014380982890725136 + }, + { + "epoch": 6.0168751818446315, + "flow_matching_loss": 0.01945654898881912, + "grad_norm": 0.20554119348526, + "learning_rate": 2.342509564330325e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006928625050932169, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028980149887502193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016755265090614556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7345467209815979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20680, + "weighted_flow_matching_loss": 0.01945654898881912, + "weighted_rkd_loss": 0.012307464983314276 + }, + { + "epoch": 6.019784695955775, + "flow_matching_loss": 0.02282909080386162, + "grad_norm": 0.22788085043430328, + "learning_rate": 2.3418254295926068e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007635345309972763, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033432552590966225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018613945785909893, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7343154549598694, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20690, + "weighted_flow_matching_loss": 0.02282909080386162, + "weighted_rkd_loss": 0.013668409921228886 + }, + { + "epoch": 6.022694210066919, + "flow_matching_loss": 0.020742679573595525, + "grad_norm": 0.20521461963653564, + "learning_rate": 2.34114103913055e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007623415766283869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035267993807792664, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01877363072708249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7340841710567474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20700, + "weighted_flow_matching_loss": 0.020742679573595525, + "weighted_rkd_loss": 0.013781471736729145 + }, + { + "epoch": 6.025603724178063, + "flow_matching_loss": 0.021407859958708288, + "grad_norm": 0.20424073934555054, + "learning_rate": 2.3404563931520545e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008233065903186797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036533509846776724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020119482465088368, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7338527977466583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20710, + "weighted_flow_matching_loss": 0.021407859958708288, + "weighted_rkd_loss": 0.014764711074531079 + }, + { + "epoch": 6.0285132382892055, + "flow_matching_loss": 0.020129221118986608, + "grad_norm": 0.21701332926750183, + "learning_rate": 2.3397714918650982e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007367906393483281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031916154315695167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017927428148686884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.73362135887146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20720, + "weighted_flow_matching_loss": 0.020129221118986608, + "weighted_rkd_loss": 0.013151958398520947 + }, + { + "epoch": 6.031422752400349, + "flow_matching_loss": 0.021912353485822676, + "grad_norm": 0.18499040603637695, + "learning_rate": 2.3390863354777346e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073158056009560825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031821322161704304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017813743464648722, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7333898603916168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20730, + "weighted_flow_matching_loss": 0.021912353485822676, + "weighted_rkd_loss": 0.013064493145793676 + }, + { + "epoch": 6.034332266511493, + "flow_matching_loss": 0.02057616077363491, + "grad_norm": 0.22522981464862823, + "learning_rate": 2.338400924198098e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007688865531235933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033394058234989645, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018717136606574057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7331582963466644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20740, + "weighted_flow_matching_loss": 0.02057616077363491, + "weighted_rkd_loss": 0.013722644932568074 + }, + { + "epoch": 6.037241780622636, + "flow_matching_loss": 0.021673476696014403, + "grad_norm": 0.22551585733890533, + "learning_rate": 2.3377152582343963e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077648207545280455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003395026363432407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018924667686223983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7329266667366028, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20750, + "weighted_flow_matching_loss": 0.021673476696014403, + "weighted_rkd_loss": 0.01387038230895996 + }, + { + "epoch": 6.040151294733779, + "flow_matching_loss": 0.021867396868765356, + "grad_norm": 0.1875399649143219, + "learning_rate": 2.3370293377949184e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007608856726437807, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00356561578810215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018783329334110022, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7326949775218964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20760, + "weighted_flow_matching_loss": 0.021867396868765356, + "weighted_rkd_loss": 0.01376251494511962 + }, + { + "epoch": 6.043060808844923, + "flow_matching_loss": 0.021138694323599338, + "grad_norm": 0.23442798852920532, + "learning_rate": 2.3363431630880282e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007786583807319403, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003283811779692769, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018856979440897703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7324632227420806, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20770, + "weighted_flow_matching_loss": 0.021138694323599338, + "weighted_rkd_loss": 0.013812000676989556 + }, + { + "epoch": 6.045970322956066, + "flow_matching_loss": 0.021394300926476716, + "grad_norm": 0.20304381847381592, + "learning_rate": 2.3356567343221676e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008223747182637454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003598990780301392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020046485215425493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7322314083576202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20780, + "weighted_flow_matching_loss": 0.021394300926476716, + "weighted_rkd_loss": 0.01467856988310814 + }, + { + "epoch": 6.04887983706721, + "flow_matching_loss": 0.019806998129934073, + "grad_norm": 0.21087954938411713, + "learning_rate": 2.3349700517058548e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00689491331577301, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028163727838546038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01660619955509901, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.731999522447586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20790, + "weighted_flow_matching_loss": 0.019806998129934073, + "weighted_rkd_loss": 0.012155694048851728 + }, + { + "epoch": 6.051789351178353, + "flow_matching_loss": 0.02095209639519453, + "grad_norm": 0.23544399440288544, + "learning_rate": 2.3342831154476866e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007542006205767393, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003355824667960405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018439837172627448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7317675709724426, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20800, + "weighted_flow_matching_loss": 0.02095209639519453, + "weighted_rkd_loss": 0.013493495341390372 + }, + { + "epoch": 6.054698865289497, + "flow_matching_loss": 0.020573695190250874, + "grad_norm": 0.19816510379314423, + "learning_rate": 2.3335959257563362e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006871909555047751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032011901959776877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016945009306073188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7315355598926544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20810, + "weighted_flow_matching_loss": 0.020573695190250874, + "weighted_rkd_loss": 0.012395893130451441 + }, + { + "epoch": 6.05760837940064, + "flow_matching_loss": 0.021296411007642745, + "grad_norm": 0.20182465016841888, + "learning_rate": 2.332908482840553e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007389778364449739, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003263884666375816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018043441325426103, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.731303483247757, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20820, + "weighted_flow_matching_loss": 0.021296411007642745, + "weighted_rkd_loss": 0.013195190206170082 + }, + { + "epoch": 6.060517893511784, + "flow_matching_loss": 0.020737181790173052, + "grad_norm": 0.20479482412338257, + "learning_rate": 2.3322207869091648e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008335463283583523, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036079062381759288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020278832595795392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7310713529586792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20830, + "weighted_flow_matching_loss": 0.020737181790173052, + "weighted_rkd_loss": 0.014825377613306046 + }, + { + "epoch": 6.063427407622927, + "flow_matching_loss": 0.02110304348170757, + "grad_norm": 0.18024638295173645, + "learning_rate": 2.3315328381710743e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007191150309517979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032125994097441436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017594899982213974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7308391451835632, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20840, + "weighted_flow_matching_loss": 0.02110304348170757, + "weighted_rkd_loss": 0.01285903798416257 + }, + { + "epoch": 6.06633692173407, + "flow_matching_loss": 0.021431459300220012, + "grad_norm": 0.22265207767486572, + "learning_rate": 2.3308446368352634e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007589287962764502, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033291904954239728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01850776644423604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.730606883764267, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20850, + "weighted_flow_matching_loss": 0.021431459300220012, + "weighted_rkd_loss": 0.013521834462881088 + }, + { + "epoch": 6.069246435845214, + "flow_matching_loss": 0.021997526101768018, + "grad_norm": 0.17566877603530884, + "learning_rate": 2.3301561831107888e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007844718452543021, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003412092151120305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019101529475301506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.730374550819397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20860, + "weighted_flow_matching_loss": 0.021997526101768018, + "weighted_rkd_loss": 0.013951190840452909 + }, + { + "epoch": 6.0721559499563575, + "flow_matching_loss": 0.02158992290496826, + "grad_norm": 0.21272724866867065, + "learning_rate": 2.3294674772067853e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007816743617877365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003475257079117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019108744338154793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7301421701908112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20870, + "weighted_flow_matching_loss": 0.02158992290496826, + "weighted_rkd_loss": 0.01395203834399581 + }, + { + "epoch": 6.0750654640675, + "flow_matching_loss": 0.019302230048924684, + "grad_norm": 0.18855884671211243, + "learning_rate": 2.3287785193324628e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007854562252759933, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003368259477429092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019077383913099766, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7299097180366516, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20880, + "weighted_flow_matching_loss": 0.019302230048924684, + "weighted_rkd_loss": 0.013924817740917205 + }, + { + "epoch": 6.077974978178644, + "flow_matching_loss": 0.023208300210535528, + "grad_norm": 0.21692702174186707, + "learning_rate": 2.3280893096971087e-05, + "loss": 0.0368, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007630089111626148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003354020114056766, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018614198081195354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7296771943569184, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20890, + "weighted_flow_matching_loss": 0.023208300210535528, + "weighted_rkd_loss": 0.013582311663776636 + }, + { + "epoch": 6.080884492289788, + "flow_matching_loss": 0.021149091050028802, + "grad_norm": 0.21441024541854858, + "learning_rate": 2.3273998485100867e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007869006786495446, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034448892110958695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01918290266767144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7294446229934692, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20900, + "weighted_flow_matching_loss": 0.021149091050028802, + "weighted_rkd_loss": 0.013992747012525796 + }, + { + "epoch": 6.083794006400931, + "flow_matching_loss": 0.022104874812066556, + "grad_norm": 0.23320716619491577, + "learning_rate": 2.326710135980837e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0074336086865514515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033649759367108344, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018232193123549224, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.729211974143982, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20910, + "weighted_flow_matching_loss": 0.022104874812066556, + "weighted_rkd_loss": 0.013295140489935875 + }, + { + "epoch": 6.086703520512074, + "flow_matching_loss": 0.0225098704919219, + "grad_norm": 0.20207658410072327, + "learning_rate": 2.326020172318876e-05, + "loss": 0.0357, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007443715538829565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003256559371948242, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018143990356475116, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7289792597293854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20920, + "weighted_flow_matching_loss": 0.0225098704919219, + "weighted_rkd_loss": 0.013226653542369604 + }, + { + "epoch": 6.089613034623218, + "flow_matching_loss": 0.019737131893634796, + "grad_norm": 0.19255231320858002, + "learning_rate": 2.3253299577337966e-05, + "loss": 0.0326, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007206977438181638, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00321265603415668, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017626610957086085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.728746497631073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20930, + "weighted_flow_matching_loss": 0.019737131893634796, + "weighted_rkd_loss": 0.012845307774841785 + }, + { + "epoch": 6.092522548734362, + "flow_matching_loss": 0.022705693356692792, + "grad_norm": 0.199962317943573, + "learning_rate": 2.324639492435267e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007809242000803351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034553628880530595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019073846749961376, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7285136759281159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20940, + "weighted_flow_matching_loss": 0.022705693356692792, + "weighted_rkd_loss": 0.013895551674067974 + }, + { + "epoch": 6.0954320628455045, + "flow_matching_loss": 0.02271758746355772, + "grad_norm": 0.20323921740055084, + "learning_rate": 2.3239487766330336e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075735487043857574, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032496339408680797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018396731466054916, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7282807767391205, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20950, + "weighted_flow_matching_loss": 0.02271758746355772, + "weighted_rkd_loss": 0.01339797331020236 + }, + { + "epoch": 6.098341576956648, + "flow_matching_loss": 0.02111369576305151, + "grad_norm": 0.20198652148246765, + "learning_rate": 2.323257810536917e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077605992089957, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035976682556793094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019118866696953774, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7280478119850159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20960, + "weighted_flow_matching_loss": 0.02111369576305151, + "weighted_rkd_loss": 0.013919494114816189 + }, + { + "epoch": 6.101251091067792, + "flow_matching_loss": 0.019988146238029005, + "grad_norm": 0.20262473821640015, + "learning_rate": 2.3225665943568138e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007248448953032493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003015621192753315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017512519005686046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7278148233890533, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20970, + "weighted_flow_matching_loss": 0.019988146238029005, + "weighted_rkd_loss": 0.012745886761695146 + }, + { + "epoch": 6.104160605178935, + "flow_matching_loss": 0.021725063771009447, + "grad_norm": 0.22983916103839874, + "learning_rate": 2.3218751283026984e-05, + "loss": 0.0366, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008280738163739443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003839214867912233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020400691125541926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7275817215442657, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20980, + "weighted_flow_matching_loss": 0.021725063771009447, + "weighted_rkd_loss": 0.014843135420233012 + }, + { + "epoch": 6.1070701192900785, + "flow_matching_loss": 0.0206991545855999, + "grad_norm": 0.20629359781742096, + "learning_rate": 2.321183412584619e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00772659694775939, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033150996780022977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018768293689936398, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7273485779762268, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 20990, + "weighted_flow_matching_loss": 0.0206991545855999, + "weighted_rkd_loss": 0.013651077076792717 + }, + { + "epoch": 6.109979633401222, + "flow_matching_loss": 0.023216885142028333, + "grad_norm": 0.1907455325126648, + "learning_rate": 2.320491447412701e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073994822334498165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003423808957450092, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01822277335450053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7271153688430786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21000, + "weighted_flow_matching_loss": 0.023216885142028333, + "weighted_rkd_loss": 0.013250153791159392 + }, + { + "epoch": 6.112889147512365, + "flow_matching_loss": 0.022035343945026396, + "grad_norm": 0.17230896651744843, + "learning_rate": 2.3197992329971455e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00755981095135212, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031634859507903455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018283107690513133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7268821060657501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21010, + "weighted_flow_matching_loss": 0.022035343945026396, + "weighted_rkd_loss": 0.013289692904800176 + }, + { + "epoch": 6.115798661623509, + "flow_matching_loss": 0.019317458756268023, + "grad_norm": 0.18455173075199127, + "learning_rate": 2.3191067695482285e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008008623030036688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035757629200816156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019593008887022735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7266487777233124, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21020, + "weighted_flow_matching_loss": 0.019317458756268023, + "weighted_rkd_loss": 0.014237214718014002 + }, + { + "epoch": 6.118708175734652, + "flow_matching_loss": 0.0212233429774642, + "grad_norm": 0.20901159942150116, + "learning_rate": 2.3184140572763018e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073602828197181225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032624749932438136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01798304049298167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7264153778553009, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21030, + "weighted_flow_matching_loss": 0.0212233429774642, + "weighted_rkd_loss": 0.01306316452100873 + }, + { + "epoch": 6.121617689845796, + "flow_matching_loss": 0.020304698310792447, + "grad_norm": 0.19569851458072662, + "learning_rate": 2.3177210963917937e-05, + "loss": 0.0353, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00835374491289258, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003901960188522935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020609450154006483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7261819303035736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21040, + "weighted_flow_matching_loss": 0.020304698310792447, + "weighted_rkd_loss": 0.01496623121201992 + }, + { + "epoch": 6.124527203956939, + "flow_matching_loss": 0.023029662482440472, + "grad_norm": 0.18750768899917603, + "learning_rate": 2.317027887105207e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007449780590832233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033695194637402893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01826908057555556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7259484112262726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21050, + "weighted_flow_matching_loss": 0.023029662482440472, + "weighted_rkd_loss": 0.013262513745576143 + }, + { + "epoch": 6.127436718068083, + "flow_matching_loss": 0.021734202280640602, + "grad_norm": 0.20220890641212463, + "learning_rate": 2.3163344296271206e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006976970378309489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031220697099342944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01707601035013795, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7257148385047912, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21060, + "weighted_flow_matching_loss": 0.021734202280640602, + "weighted_rkd_loss": 0.012392247375100852 + }, + { + "epoch": 6.130346232179226, + "flow_matching_loss": 0.020805337373167275, + "grad_norm": 0.1856408566236496, + "learning_rate": 2.3156407241681888e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007239036681130529, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030261413659900425, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017504214867949487, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7254812061786652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21070, + "weighted_flow_matching_loss": 0.020805337373167275, + "weighted_rkd_loss": 0.012699000909924507 + }, + { + "epoch": 6.133255746290369, + "flow_matching_loss": 0.020421875361353158, + "grad_norm": 0.17534460127353668, + "learning_rate": 2.3149467709391407e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007986170332878828, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0036345207365229724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019606861658394335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7252474844455719, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21080, + "weighted_flow_matching_loss": 0.020421875361353158, + "weighted_rkd_loss": 0.01421983577311039 + }, + { + "epoch": 6.136165260401513, + "flow_matching_loss": 0.020030657574534417, + "grad_norm": 0.1975301206111908, + "learning_rate": 2.3142525701507807e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006813737843185663, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028591500828042626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01648662593215704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7250137329101562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21090, + "weighted_flow_matching_loss": 0.020030657574534417, + "weighted_rkd_loss": 0.01195302177220583 + }, + { + "epoch": 6.139074774512657, + "flow_matching_loss": 0.022740100510418415, + "grad_norm": 0.20781391859054565, + "learning_rate": 2.3135581220139893e-05, + "loss": 0.0369, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0080220069270581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003561339364387095, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01960535328835249, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7247798979282379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21100, + "weighted_flow_matching_loss": 0.022740100510418415, + "weighted_rkd_loss": 0.014209500327706338 + }, + { + "epoch": 6.141984288623799, + "flow_matching_loss": 0.021760992612689732, + "grad_norm": 0.18580879271030426, + "learning_rate": 2.3128634267397213e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007152026891708374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003194355173036456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01749840918928385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7245460152626038, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21110, + "weighted_flow_matching_loss": 0.021760992612689732, + "weighted_rkd_loss": 0.012678395677357911 + }, + { + "epoch": 6.144893802734943, + "flow_matching_loss": 0.020314007811248303, + "grad_norm": 0.20818905532360077, + "learning_rate": 2.3121684845390065e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071328938473016025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003091590199619532, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017357377894222738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7243120610713959, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21120, + "weighted_flow_matching_loss": 0.020314007811248303, + "weighted_rkd_loss": 0.012572102341800928 + }, + { + "epoch": 6.147803316846087, + "flow_matching_loss": 0.021716740727424622, + "grad_norm": 0.19202397763729095, + "learning_rate": 2.3114732956229496e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007071792241185903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003004433074966073, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017148017510771753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7240780591964722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21130, + "weighted_flow_matching_loss": 0.021716740727424622, + "weighted_rkd_loss": 0.012416402902454137 + }, + { + "epoch": 6.1507128309572305, + "flow_matching_loss": 0.020288599468767644, + "grad_norm": 0.17874813079833984, + "learning_rate": 2.3107778602027318e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007307735737413168, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031025668373331427, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0177180384285748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7238439857959748, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21140, + "weighted_flow_matching_loss": 0.020288599468767644, + "weighted_rkd_loss": 0.012825135607272387 + }, + { + "epoch": 6.153622345068373, + "flow_matching_loss": 0.020598590094596147, + "grad_norm": 0.22232256829738617, + "learning_rate": 2.3100821784896072e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072461002971977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003212556359358132, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01770475711673498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.723609858751297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21150, + "weighted_flow_matching_loss": 0.020598590094596147, + "weighted_rkd_loss": 0.01281128404662013 + }, + { + "epoch": 6.156531859179517, + "flow_matching_loss": 0.023270326666533947, + "grad_norm": 0.18984192609786987, + "learning_rate": 2.309386250694906e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006450576242059469, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002771979640237987, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01567313224077225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7233756482601166, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21160, + "weighted_flow_matching_loss": 0.023270326666533947, + "weighted_rkd_loss": 0.011337488424032927 + }, + { + "epoch": 6.159441373290661, + "flow_matching_loss": 0.019558481127023696, + "grad_norm": 0.16036967933177948, + "learning_rate": 2.3086900770300325e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007149128615856171, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003285456052981317, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017583713214844466, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7231414020061493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21170, + "weighted_flow_matching_loss": 0.019558481127023696, + "weighted_rkd_loss": 0.012715456448495388 + }, + { + "epoch": 6.162350887401804, + "flow_matching_loss": 0.022398629412055015, + "grad_norm": 0.18060140311717987, + "learning_rate": 2.3079936577064656e-05, + "loss": 0.0362, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076939532533288005, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037449989002197983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019132905453443528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7229070842266083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21180, + "weighted_flow_matching_loss": 0.022398629412055015, + "weighted_rkd_loss": 0.01383137172088027 + }, + { + "epoch": 6.165260401512947, + "flow_matching_loss": 0.02118658199906349, + "grad_norm": 0.2356155514717102, + "learning_rate": 2.3072969929357593e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007024897867813707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030410862527787685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017090881988406182, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7226727128028869, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21190, + "weighted_flow_matching_loss": 0.02118658199906349, + "weighted_rkd_loss": 0.012351059261709452 + }, + { + "epoch": 6.168169915624091, + "flow_matching_loss": 0.021327303536236287, + "grad_norm": 0.1681404858827591, + "learning_rate": 2.3066000829295422e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007300341641530395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033242228673771024, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017924906220287085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7224382698535919, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21200, + "weighted_flow_matching_loss": 0.021327303536236287, + "weighted_rkd_loss": 0.01294974535703659 + }, + { + "epoch": 6.171079429735234, + "flow_matching_loss": 0.022350717149674893, + "grad_norm": 0.258489727973938, + "learning_rate": 2.3059029278995167e-05, + "loss": 0.0359, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007689902139827609, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003320454340428114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01870025843381882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7222037613391876, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21210, + "weighted_flow_matching_loss": 0.022350717149674893, + "weighted_rkd_loss": 0.013505323510617017 + }, + { + "epoch": 6.1739889438463775, + "flow_matching_loss": 0.021043758001178504, + "grad_norm": 0.22423139214515686, + "learning_rate": 2.3052055280574605e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006198371481150389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002762914006598294, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015159656852483749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7219692051410675, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21220, + "weighted_flow_matching_loss": 0.021043758001178504, + "weighted_rkd_loss": 0.010944806784391404 + }, + { + "epoch": 6.176898457957521, + "flow_matching_loss": 0.02112435232847929, + "grad_norm": 0.15828000009059906, + "learning_rate": 2.3045078836152254e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076780580449849365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003260584664531052, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018616700638085605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7217345893383026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21230, + "weighted_flow_matching_loss": 0.02112435232847929, + "weighted_rkd_loss": 0.01343644754961133 + }, + { + "epoch": 6.179807972068665, + "flow_matching_loss": 0.022822161205112934, + "grad_norm": 0.21211767196655273, + "learning_rate": 2.3038099947847367e-05, + "loss": 0.0363, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00760826556943357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034653039183467627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018681835290044545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.721499902009964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21240, + "weighted_flow_matching_loss": 0.022822161205112934, + "weighted_rkd_loss": 0.01347895162180066 + }, + { + "epoch": 6.182717486179808, + "flow_matching_loss": 0.022280612960457802, + "grad_norm": 0.23954704403877258, + "learning_rate": 2.3031118617779946e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007554331654682755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003234653943218291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018343317322432995, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7212651669979095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21250, + "weighted_flow_matching_loss": 0.022280612960457802, + "weighted_rkd_loss": 0.013230392523109914 + }, + { + "epoch": 6.1856270002909515, + "flow_matching_loss": 0.02113884501159191, + "grad_norm": 0.21692907810211182, + "learning_rate": 2.3024134848070744e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007551245205104351, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034379676450043916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018540458008646964, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7210303544998169, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21260, + "weighted_flow_matching_loss": 0.02113884501159191, + "weighted_rkd_loss": 0.013368146307766438 + }, + { + "epoch": 6.188536514402095, + "flow_matching_loss": 0.021295342687517406, + "grad_norm": 0.18933585286140442, + "learning_rate": 2.301714864084124e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006995680788531899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031317177694290877, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0171230792067945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7207954943180084, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21270, + "weighted_flow_matching_loss": 0.021295342687517406, + "weighted_rkd_loss": 0.012342233210802078 + }, + { + "epoch": 6.191446028513238, + "flow_matching_loss": 0.019556928146630525, + "grad_norm": 0.153569757938385, + "learning_rate": 2.3010159998213654e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006845182552933693, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029741102131083607, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016664475575089453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7205605685710907, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21280, + "weighted_flow_matching_loss": 0.019556928146630525, + "weighted_rkd_loss": 0.012007876439020038 + }, + { + "epoch": 6.194355542624382, + "flow_matching_loss": 0.02136457562446594, + "grad_norm": 0.19071944057941437, + "learning_rate": 2.300316892231096e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007485205680131913, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003161766310222447, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018132177647203207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7203255832195282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21290, + "weighted_flow_matching_loss": 0.02136457562446594, + "weighted_rkd_loss": 0.013061085436493158 + }, + { + "epoch": 6.197265056735525, + "flow_matching_loss": 0.022045744955539702, + "grad_norm": 0.21310269832611084, + "learning_rate": 2.299617541525686e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007592619583010674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003613868076354265, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01879910733550787, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7200905382633209, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21300, + "weighted_flow_matching_loss": 0.022045744955539702, + "weighted_rkd_loss": 0.01353710675612092 + }, + { + "epoch": 6.200174570846668, + "flow_matching_loss": 0.022922829166054724, + "grad_norm": 0.20288048684597015, + "learning_rate": 2.2989179479175794e-05, + "loss": 0.0355, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00718704704195261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031547795282676817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017528873682022095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7198554277420044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21310, + "weighted_flow_matching_loss": 0.022922829166054724, + "weighted_rkd_loss": 0.012618389260023833 + }, + { + "epoch": 6.203084084957812, + "flow_matching_loss": 0.021115580387413502, + "grad_norm": 0.20163701474666595, + "learning_rate": 2.2982181116192944e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00756939323619008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032822141656652093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018421000428497793, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.719620269536972, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21320, + "weighted_flow_matching_loss": 0.021115580387413502, + "weighted_rkd_loss": 0.013256232626736164 + }, + { + "epoch": 6.205993599068956, + "flow_matching_loss": 0.019002122431993486, + "grad_norm": 0.18683335185050964, + "learning_rate": 2.297518032843423e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073279760312289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030302714090794324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017686223704367877, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7193850398063659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21330, + "weighted_flow_matching_loss": 0.019002122431993486, + "weighted_rkd_loss": 0.01272311545908451 + }, + { + "epoch": 6.208903113180099, + "flow_matching_loss": 0.02165610883384943, + "grad_norm": 0.17719970643520355, + "learning_rate": 2.2968177118026305e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00817317827604711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0037376416381448506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.020083998329937458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7191497504711151, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21340, + "weighted_flow_matching_loss": 0.02165610883384943, + "weighted_rkd_loss": 0.014443487487733364 + }, + { + "epoch": 6.211812627291242, + "flow_matching_loss": 0.01946404278278351, + "grad_norm": 0.24848082661628723, + "learning_rate": 2.296117148709656e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006908959802240133, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003188660088926554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017006579786539078, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7189144015312194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21350, + "weighted_flow_matching_loss": 0.01946404278278351, + "weighted_rkd_loss": 0.012226242013275624 + }, + { + "epoch": 6.214722141402386, + "flow_matching_loss": 0.02130206534639001, + "grad_norm": 0.16558389365673065, + "learning_rate": 2.295416343777312e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007339443732053042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003380981693044305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01805986911058426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7186789810657501, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21360, + "weighted_flow_matching_loss": 0.02130206534639001, + "weighted_rkd_loss": 0.012979235965758562 + }, + { + "epoch": 6.21763165551353, + "flow_matching_loss": 0.019985891692340375, + "grad_norm": 0.21909044682979584, + "learning_rate": 2.2947152972184838e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075900641735643145, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003355545550584793, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01853567371144891, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7184435307979584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21370, + "weighted_flow_matching_loss": 0.019985891692340375, + "weighted_rkd_loss": 0.013316718023270368 + }, + { + "epoch": 6.220541169624672, + "flow_matching_loss": 0.020610930398106575, + "grad_norm": 0.1698908805847168, + "learning_rate": 2.294014009246132e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008025791449472307, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003726896340958774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019778479263186453, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7182080030441285, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21380, + "weighted_flow_matching_loss": 0.020610930398106575, + "weighted_rkd_loss": 0.014205383230000734 + }, + { + "epoch": 6.223450683735816, + "flow_matching_loss": 0.022627591341733932, + "grad_norm": 0.2010974884033203, + "learning_rate": 2.293312480073289e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007903106603771448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034277127124369144, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01923392601311207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7179724156856537, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21390, + "weighted_flow_matching_loss": 0.022627591341733932, + "weighted_rkd_loss": 0.013809393718838691 + }, + { + "epoch": 6.22636019784696, + "flow_matching_loss": 0.01922550955787301, + "grad_norm": 0.18036112189292908, + "learning_rate": 2.2926107099130607e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0075681338552385565, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003126278193667531, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018262546230107545, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7177367687225342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21400, + "weighted_flow_matching_loss": 0.01922550955787301, + "weighted_rkd_loss": 0.013107632752507925 + }, + { + "epoch": 6.229269711958103, + "flow_matching_loss": 0.022208596765995025, + "grad_norm": 0.20897839963436127, + "learning_rate": 2.2919086989786267e-05, + "loss": 0.0364, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008031206717714668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003672439884394407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019734853133559226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7175010740756989, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21410, + "weighted_flow_matching_loss": 0.022208596765995025, + "weighted_rkd_loss": 0.014159923885017633 + }, + { + "epoch": 6.232179226069246, + "flow_matching_loss": 0.021676244959235192, + "grad_norm": 0.18313199281692505, + "learning_rate": 2.2912064474832384e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00716527751646936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031041130423545837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017434668261557817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7172652959823609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21420, + "weighted_flow_matching_loss": 0.021676244959235192, + "weighted_rkd_loss": 0.012505351286381482 + }, + { + "epoch": 6.23508874018039, + "flow_matching_loss": 0.021291187684983014, + "grad_norm": 0.2114933729171753, + "learning_rate": 2.2905039556402223e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007772010238841176, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034380804281681776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018982101138681172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7170294761657715, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21430, + "weighted_flow_matching_loss": 0.021291187684983014, + "weighted_rkd_loss": 0.013610680028796196 + }, + { + "epoch": 6.237998254291533, + "flow_matching_loss": 0.01936420639976859, + "grad_norm": 0.1940133422613144, + "learning_rate": 2.2898012236629766e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007030994351953268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030170698184520004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01707905838266015, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7167935907840729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21440, + "weighted_flow_matching_loss": 0.01936420639976859, + "weighted_rkd_loss": 0.012242194358259439 + }, + { + "epoch": 6.240907768402677, + "flow_matching_loss": 0.021695029549300672, + "grad_norm": 0.18429215252399445, + "learning_rate": 2.2890982517649728e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007388277817517519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035961389541625975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01837269440293312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7165576457977295, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21450, + "weighted_flow_matching_loss": 0.021695029549300672, + "weighted_rkd_loss": 0.013165078777819871 + }, + { + "epoch": 6.24381728251382, + "flow_matching_loss": 0.01845543272793293, + "grad_norm": 0.1808357834815979, + "learning_rate": 2.2883950401597548e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006840883754193783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031882710289210083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016870038583874702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7163216471672058, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21460, + "weighted_flow_matching_loss": 0.01845543272793293, + "weighted_rkd_loss": 0.012084291316568851 + }, + { + "epoch": 6.246726796624964, + "flow_matching_loss": 0.020572335459291935, + "grad_norm": 0.18091481924057007, + "learning_rate": 2.2876915890609396e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007102082157507539, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003177084610797465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017381248623132707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7160855829715729, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21470, + "weighted_flow_matching_loss": 0.020572335459291935, + "weighted_rkd_loss": 0.01244647316634655 + }, + { + "epoch": 6.249636310736107, + "flow_matching_loss": 0.022034774348139764, + "grad_norm": 0.20541803538799286, + "learning_rate": 2.2869878986822182e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0076336934696882965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003390907729044557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018658294714987277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7158494532108307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21480, + "weighted_flow_matching_loss": 0.022034774348139764, + "weighted_rkd_loss": 0.013356485031545162 + }, + { + "epoch": 6.2525458248472505, + "flow_matching_loss": 0.02342815976589918, + "grad_norm": 0.23562511801719666, + "learning_rate": 2.286283969237352e-05, + "loss": 0.037, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0077209191396832464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035735937068238854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01901543205603957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7156132757663727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21490, + "weighted_flow_matching_loss": 0.02342815976589918, + "weighted_rkd_loss": 0.013607497047632932 + }, + { + "epoch": 6.255455338958394, + "flow_matching_loss": 0.02171125030145049, + "grad_norm": 0.18665863573551178, + "learning_rate": 2.2855798009401768e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007637407351285219, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003417233726941049, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018692048639059065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7153770327568054, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21500, + "weighted_flow_matching_loss": 0.02171125030145049, + "weighted_rkd_loss": 0.013371850643306971 + }, + { + "epoch": 6.258364853069537, + "flow_matching_loss": 0.022439856082201004, + "grad_norm": 0.1961977779865265, + "learning_rate": 2.2848753940045996e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007381144631654024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032979298615828157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018060218915343283, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7151407301425934, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21510, + "weighted_flow_matching_loss": 0.022439856082201004, + "weighted_rkd_loss": 0.012915526703000068 + }, + { + "epoch": 6.261274367180681, + "flow_matching_loss": 0.020176092535257338, + "grad_norm": 0.21303990483283997, + "learning_rate": 2.284170748644602e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007067774841561914, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029986910754814743, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01713424064218998, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7149043798446655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21520, + "weighted_flow_matching_loss": 0.020176092535257338, + "weighted_rkd_loss": 0.012249392736703158 + }, + { + "epoch": 6.2641838812918245, + "flow_matching_loss": 0.0203250371851027, + "grad_norm": 0.18850569427013397, + "learning_rate": 2.283465865074235e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007458247430622578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034144913079217076, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01833098642528057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7146679580211639, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21530, + "weighted_flow_matching_loss": 0.0203250371851027, + "weighted_rkd_loss": 0.013100563175976276 + }, + { + "epoch": 6.267093395402968, + "flow_matching_loss": 0.021599544025957585, + "grad_norm": 0.2256695181131363, + "learning_rate": 2.2827607435076245e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007707563880831003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003371993824839592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01878712121397257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7144314885139466, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21540, + "weighted_flow_matching_loss": 0.021599544025957585, + "weighted_rkd_loss": 0.013422087300568819 + }, + { + "epoch": 6.270002909514111, + "flow_matching_loss": 0.020724038034677504, + "grad_norm": 0.26514333486557007, + "learning_rate": 2.282055384158968e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007037601573392749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030778497573919596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017153052613139153, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7141949534416199, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21550, + "weighted_flow_matching_loss": 0.020724038034677504, + "weighted_rkd_loss": 0.012250462640076875 + }, + { + "epoch": 6.272912423625255, + "flow_matching_loss": 0.022126204892992975, + "grad_norm": 0.19322887063026428, + "learning_rate": 2.2813497872425348e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007206061668694019, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030147423502057792, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017426866013556717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.713958352804184, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21560, + "weighted_flow_matching_loss": 0.022126204892992975, + "weighted_rkd_loss": 0.012442095763981342 + }, + { + "epoch": 6.275821937736398, + "flow_matching_loss": 0.02154563311487436, + "grad_norm": 0.19270311295986176, + "learning_rate": 2.2806439529726662e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069078001193702224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029713925905525684, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016786992829293013, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7137217044830322, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21570, + "weighted_flow_matching_loss": 0.02154563311487436, + "weighted_rkd_loss": 0.011981224827468395 + }, + { + "epoch": 6.278731451847541, + "flow_matching_loss": 0.02221035845577717, + "grad_norm": 0.2653876841068268, + "learning_rate": 2.2799378815637766e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073567593935877085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032116848975419997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0179252034984529, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7134849905967713, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21580, + "weighted_flow_matching_loss": 0.02221035845577717, + "weighted_rkd_loss": 0.012789303809404374 + }, + { + "epoch": 6.281640965958685, + "flow_matching_loss": 0.022044479846954346, + "grad_norm": 0.18056190013885498, + "learning_rate": 2.2792315732303517e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007406158419325948, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033632273552939296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018175544403493404, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.71324822306633, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21590, + "weighted_flow_matching_loss": 0.022044479846954346, + "weighted_rkd_loss": 0.012963810004293919 + }, + { + "epoch": 6.284550480069829, + "flow_matching_loss": 0.02096132282167673, + "grad_norm": 0.18813049793243408, + "learning_rate": 2.2785250281869492e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068416035734117035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029054533690214155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016588660329580306, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7130113959312439, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21600, + "weighted_flow_matching_loss": 0.02096132282167673, + "weighted_rkd_loss": 0.011827843450009822 + }, + { + "epoch": 6.2874599941809715, + "flow_matching_loss": 0.022373863868415354, + "grad_norm": 0.19718943536281586, + "learning_rate": 2.277818246648199e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007404353423044086, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003407640475779772, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018216347321867943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7127745091915131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21610, + "weighted_flow_matching_loss": 0.022373863868415354, + "weighted_rkd_loss": 0.012984034512192011 + }, + { + "epoch": 6.290369508292115, + "flow_matching_loss": 0.020506780222058298, + "grad_norm": 0.18283571302890778, + "learning_rate": 2.2771112288288025e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006930632051080465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003022901830263436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01688416600227356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7125375628471374, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21620, + "weighted_flow_matching_loss": 0.020506780222058298, + "weighted_rkd_loss": 0.012030699010938405 + }, + { + "epoch": 6.293279022403259, + "flow_matching_loss": 0.021919185295701027, + "grad_norm": 0.19498251378536224, + "learning_rate": 2.276403974943533e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007000700710341335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031023929594084622, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01710379431024194, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.712300568819046, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21630, + "weighted_flow_matching_loss": 0.021919185295701027, + "weighted_rkd_loss": 0.012183082476258277 + }, + { + "epoch": 6.296188536514402, + "flow_matching_loss": 0.021169118583202362, + "grad_norm": 0.2451135516166687, + "learning_rate": 2.2756964852072357e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008065292751416564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034765424206852913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019607127644121647, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7120635032653808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21640, + "weighted_flow_matching_loss": 0.021169118583202362, + "weighted_rkd_loss": 0.013961491640657187 + }, + { + "epoch": 6.299098050625545, + "flow_matching_loss": 0.02046170551329851, + "grad_norm": 0.24266017973423004, + "learning_rate": 2.2749887598348273e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006726437108591199, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002919777249917388, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0163726513274014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7118263721466065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21650, + "weighted_flow_matching_loss": 0.02046170551329851, + "weighted_rkd_loss": 0.011654369533061981 + }, + { + "epoch": 6.302007564736689, + "flow_matching_loss": 0.02207350768148899, + "grad_norm": 0.1840728372335434, + "learning_rate": 2.2742807990412963e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072398898657411335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031238694209605455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017603649012744426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7115892171859741, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21660, + "weighted_flow_matching_loss": 0.02207350768148899, + "weighted_rkd_loss": 0.012526565883308648 + }, + { + "epoch": 6.304917078847833, + "flow_matching_loss": 0.02018175143748522, + "grad_norm": 0.19379961490631104, + "learning_rate": 2.2735726030417022e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007352005690336227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031928775599226354, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017896888963878154, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7113519728183746, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21670, + "weighted_flow_matching_loss": 0.02018175143748522, + "weighted_rkd_loss": 0.012730939872562885 + }, + { + "epoch": 6.307826592958976, + "flow_matching_loss": 0.019059680216014386, + "grad_norm": 0.164076566696167, + "learning_rate": 2.272864172051176e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006901951134204864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029612624086439608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01676516467705369, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7111146748065948, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21680, + "weighted_flow_matching_loss": 0.019059680216014386, + "weighted_rkd_loss": 0.01192200966179371 + }, + { + "epoch": 6.310736107070119, + "flow_matching_loss": 0.020378199592232703, + "grad_norm": 0.14828607439994812, + "learning_rate": 2.272155506284921e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006588027207180857, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002836179663427174, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016012234054505824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7108773231506348, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21690, + "weighted_flow_matching_loss": 0.020378199592232703, + "weighted_rkd_loss": 0.01138286767527461 + }, + { + "epoch": 6.313645621181263, + "flow_matching_loss": 0.01996348574757576, + "grad_norm": 0.15989521145820618, + "learning_rate": 2.2714466059582105e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007606905605643988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032925354083999992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018506346549838782, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7106399178504944, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21700, + "weighted_flow_matching_loss": 0.01996348574757576, + "weighted_rkd_loss": 0.01315133753232658 + }, + { + "epoch": 6.316555135292406, + "flow_matching_loss": 0.0203517678193748, + "grad_norm": 0.18234822154045105, + "learning_rate": 2.2707374712863902e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006651626573875547, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029806850245222448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016283938195556402, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7104024469852448, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21710, + "weighted_flow_matching_loss": 0.0203517678193748, + "weighted_rkd_loss": 0.011568226292729378 + }, + { + "epoch": 6.31946464940355, + "flow_matching_loss": 0.020604982040822506, + "grad_norm": 0.21512070298194885, + "learning_rate": 2.270028102484876e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007106919912621379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002965572685934603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01717941239476204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7101649284362793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21720, + "weighted_flow_matching_loss": 0.020604982040822506, + "weighted_rkd_loss": 0.012200231477618218 + }, + { + "epoch": 6.322374163514693, + "flow_matching_loss": 0.020533730927854776, + "grad_norm": 0.1589382290840149, + "learning_rate": 2.2693184997691558e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006714565213769674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002809261926449835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01623839242383838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7099273383617402, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21730, + "weighted_flow_matching_loss": 0.020533730927854776, + "weighted_rkd_loss": 0.011528157535940409 + }, + { + "epoch": 6.325283677625836, + "flow_matching_loss": 0.020968216564506294, + "grad_norm": 0.2107182890176773, + "learning_rate": 2.268608663354788e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071749120019376274, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003320919768884778, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017670744005590676, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7096897065639496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21740, + "weighted_flow_matching_loss": 0.020968216564506294, + "weighted_rkd_loss": 0.01254069795832038 + }, + { + "epoch": 6.32819319173698, + "flow_matching_loss": 0.0217145343311131, + "grad_norm": 0.1773904263973236, + "learning_rate": 2.2678985934574022e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00713520753197372, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003076911112293601, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017347326409071682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7094520032405853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21750, + "weighted_flow_matching_loss": 0.0217145343311131, + "weighted_rkd_loss": 0.01230723662301898 + }, + { + "epoch": 6.3311027058481235, + "flow_matching_loss": 0.020835553109645844, + "grad_norm": 0.17602095007896423, + "learning_rate": 2.267188290292699e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007568425871431828, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032708055572584272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018407657369971277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7092142403125763, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21760, + "weighted_flow_matching_loss": 0.020835553109645844, + "weighted_rkd_loss": 0.013055028952658176 + }, + { + "epoch": 6.334012219959266, + "flow_matching_loss": 0.024238541536033154, + "grad_norm": 0.19659534096717834, + "learning_rate": 2.2664777540764496e-05, + "loss": 0.0371, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007464699679985642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032445435179397466, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018173942808061838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7089764177799225, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21770, + "weighted_flow_matching_loss": 0.024238541536033154, + "weighted_rkd_loss": 0.012884904071688652 + }, + { + "epoch": 6.33692173407041, + "flow_matching_loss": 0.020250053517520426, + "grad_norm": 0.21576283872127533, + "learning_rate": 2.265766985024496e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006999276205897331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029393233591690658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016937875654548407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7087385714054107, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21780, + "weighted_flow_matching_loss": 0.020250053517520426, + "weighted_rkd_loss": 0.01200446579605341 + }, + { + "epoch": 6.339831248181554, + "flow_matching_loss": 0.022618422284722328, + "grad_norm": 0.17667117714881897, + "learning_rate": 2.2650559833527516e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007228905195370316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003137639514170587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017595450021326543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7085006475448609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21790, + "weighted_flow_matching_loss": 0.022618422284722328, + "weighted_rkd_loss": 0.012466390058398247 + }, + { + "epoch": 6.3427407622926975, + "flow_matching_loss": 0.02031027227640152, + "grad_norm": 0.2027687132358551, + "learning_rate": 2.2643447492771998e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007942900946363806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035635847132653, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019449387025088073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7082626640796661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21800, + "weighted_flow_matching_loss": 0.02031027227640152, + "weighted_rkd_loss": 0.013775138184428215 + }, + { + "epoch": 6.34565027640384, + "flow_matching_loss": 0.019504634849727153, + "grad_norm": 0.21785873174667358, + "learning_rate": 2.2636332830138943e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006707060011103749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030250659678131343, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016439186036586763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7080246210098267, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21810, + "weighted_flow_matching_loss": 0.019504634849727153, + "weighted_rkd_loss": 0.011639304831624031 + }, + { + "epoch": 6.348559790514984, + "flow_matching_loss": 0.021746819280087946, + "grad_norm": 0.1986209899187088, + "learning_rate": 2.2629215847789596e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007353995181620121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032612062292173504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01796919647604227, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7077865302562714, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21820, + "weighted_flow_matching_loss": 0.021746819280087946, + "weighted_rkd_loss": 0.0127183279953897 + }, + { + "epoch": 6.351469304626128, + "flow_matching_loss": 0.020414932072162627, + "grad_norm": 0.1603057086467743, + "learning_rate": 2.2622096547885912e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007321450253948569, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031164549058303238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01775935534387827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7075483679771424, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21830, + "weighted_flow_matching_loss": 0.020414932072162627, + "weighted_rkd_loss": 0.01256567407399416 + }, + { + "epoch": 6.3543788187372705, + "flow_matching_loss": 0.020624277368187904, + "grad_norm": 0.20018437504768372, + "learning_rate": 2.261497493259055e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007143585290759802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031716267578303812, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01745879715308547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7073101580142975, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21840, + "weighted_flow_matching_loss": 0.020624277368187904, + "weighted_rkd_loss": 0.012348737753927707 + }, + { + "epoch": 6.357288332848414, + "flow_matching_loss": 0.02228486128151417, + "grad_norm": 0.20163635909557343, + "learning_rate": 2.2607851004066863e-05, + "loss": 0.0356, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00775719890370965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033046534284949304, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018819051142781974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7070718884468079, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21850, + "weighted_flow_matching_loss": 0.02228486128151417, + "weighted_rkd_loss": 0.013306551054120064 + }, + { + "epoch": 6.360197846959558, + "flow_matching_loss": 0.020672392658889295, + "grad_norm": 0.19795648753643036, + "learning_rate": 2.2600724764478915e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007491133967414498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031166524393484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01809892039746046, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.706833565235138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21860, + "weighted_flow_matching_loss": 0.020672392658889295, + "weighted_rkd_loss": 0.012792913243174553 + }, + { + "epoch": 6.363107361070702, + "flow_matching_loss": 0.022574487142264843, + "grad_norm": 0.1904587596654892, + "learning_rate": 2.2593596215991473e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006719736196100712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030153945088386537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016454867087304592, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7065951824188232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21870, + "weighted_flow_matching_loss": 0.022574487142264843, + "weighted_rkd_loss": 0.011626903805881739 + }, + { + "epoch": 6.3660168751818444, + "flow_matching_loss": 0.021609846130013467, + "grad_norm": 0.18346287310123444, + "learning_rate": 2.2586465360769996e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007740780198946595, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003191968146711588, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018673528637737035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7063567459583282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21880, + "weighted_flow_matching_loss": 0.021609846130013467, + "weighted_rkd_loss": 0.013190306862816215 + }, + { + "epoch": 6.368926389292988, + "flow_matching_loss": 0.02294114986434579, + "grad_norm": 0.28433388471603394, + "learning_rate": 2.257933220098065e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007135522086173296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003085478954017162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01735652294009924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.706118243932724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21890, + "weighted_flow_matching_loss": 0.02294114986434579, + "weighted_rkd_loss": 0.012255616951733828 + }, + { + "epoch": 6.371835903404132, + "flow_matching_loss": 0.02071453761309385, + "grad_norm": 0.21011491119861603, + "learning_rate": 2.2572196738790298e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007921709958463908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003473373898304999, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019316793978214265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7058796942234039, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21900, + "weighted_flow_matching_loss": 0.02071453761309385, + "weighted_rkd_loss": 0.013635301776230336 + }, + { + "epoch": 6.374745417515275, + "flow_matching_loss": 0.02125411219894886, + "grad_norm": 0.23860041797161102, + "learning_rate": 2.2565058976366517e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007232309412211179, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032813356956467032, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017745954170823097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7056410908699036, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21910, + "weighted_flow_matching_loss": 0.02125411219894886, + "weighted_rkd_loss": 0.012522099819034339 + }, + { + "epoch": 6.377654931626418, + "flow_matching_loss": 0.02256295131519437, + "grad_norm": 0.25808924436569214, + "learning_rate": 2.2557918915877558e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007612671051174402, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0035037015564739703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018729043938219548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7054024279117584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21920, + "weighted_flow_matching_loss": 0.02256295131519437, + "weighted_rkd_loss": 0.01321153100579977 + }, + { + "epoch": 6.380564445737562, + "flow_matching_loss": 0.021712579671293496, + "grad_norm": 0.17698127031326294, + "learning_rate": 2.2550776559492392e-05, + "loss": 0.0348, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007624934054911136, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032466039061546325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01849647192284465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7051637053489686, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21930, + "weighted_flow_matching_loss": 0.021712579671293496, + "weighted_rkd_loss": 0.013042966835200787 + }, + { + "epoch": 6.383473959848705, + "flow_matching_loss": 0.020188437588512898, + "grad_norm": 0.1896638423204422, + "learning_rate": 2.254363190938067e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007732791267335415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033472424605861307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018812824971973896, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7049249231815338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21940, + "weighted_flow_matching_loss": 0.020188437588512898, + "weighted_rkd_loss": 0.013261678069829941 + }, + { + "epoch": 6.386383473959849, + "flow_matching_loss": 0.020191051065921783, + "grad_norm": 0.25807881355285645, + "learning_rate": 2.2536484967712753e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071399535052478315, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030895209638401866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017369428277015687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7046860873699188, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21950, + "weighted_flow_matching_loss": 0.020191051065921783, + "weighted_rkd_loss": 0.012239854037761688 + }, + { + "epoch": 6.389292988070992, + "flow_matching_loss": 0.02203264646232128, + "grad_norm": 0.178810715675354, + "learning_rate": 2.252933573665969e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007140917982906103, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033596324967220424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017641468346118926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.704447203874588, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21960, + "weighted_flow_matching_loss": 0.02203264646232128, + "weighted_rkd_loss": 0.012427625898271798 + }, + { + "epoch": 6.392202502182135, + "flow_matching_loss": 0.020485398918390275, + "grad_norm": 0.20919573307037354, + "learning_rate": 2.2522184218393234e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007521458109840751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031764660961925983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0182193823158741, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7042082607746124, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21970, + "weighted_flow_matching_loss": 0.020485398918390275, + "weighted_rkd_loss": 0.012830263655632734 + }, + { + "epoch": 6.395112016293279, + "flow_matching_loss": 0.020315589383244514, + "grad_norm": 0.18211820721626282, + "learning_rate": 2.2515030415085822e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007289989665150642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003213766682893038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01779374582692981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.703969269990921, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21980, + "weighted_flow_matching_loss": 0.020315589383244514, + "weighted_rkd_loss": 0.012526310421526432 + }, + { + "epoch": 6.398021530404423, + "flow_matching_loss": 0.019027662742882968, + "grad_norm": 0.1772608458995819, + "learning_rate": 2.2507874328910592e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00705712684430182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003028375213034451, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017142629064619542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7037301957607269, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 21990, + "weighted_flow_matching_loss": 0.019027662742882968, + "weighted_rkd_loss": 0.012063848692923785 + }, + { + "epoch": 6.400931044515566, + "flow_matching_loss": 0.021545948646962642, + "grad_norm": 0.24673035740852356, + "learning_rate": 2.250071596204138e-05, + "loss": 0.035, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007915455289185047, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033631597412750124, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019194070249795914, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.703491085767746, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22000, + "weighted_flow_matching_loss": 0.021545948646962642, + "weighted_rkd_loss": 0.013502760324627161 + }, + { + "epoch": 6.403840558626709, + "flow_matching_loss": 0.021289705671370028, + "grad_norm": 0.24346989393234253, + "learning_rate": 2.2493555316652698e-05, + "loss": 0.0352, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007985566230490803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003757298132404685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019728430546820164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7032519161701203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22010, + "weighted_flow_matching_loss": 0.021289705671370028, + "weighted_rkd_loss": 0.013873929623514414 + }, + { + "epoch": 6.406750072737853, + "flow_matching_loss": 0.022332469932734966, + "grad_norm": 0.1650175005197525, + "learning_rate": 2.248639239491977e-05, + "loss": 0.0358, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007812307449057698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034749782644212246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01909959316253662, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7030126929283143, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22020, + "weighted_flow_matching_loss": 0.022332469932734966, + "weighted_rkd_loss": 0.01342732524499297 + }, + { + "epoch": 6.4096595868489965, + "flow_matching_loss": 0.021785423159599304, + "grad_norm": 0.18619106709957123, + "learning_rate": 2.24792271990185e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00688191750086844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029398091370239853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016703644301742315, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7027734220027924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22030, + "weighted_flow_matching_loss": 0.021785423159599304, + "weighted_rkd_loss": 0.01173884654417634 + }, + { + "epoch": 6.412569100960139, + "flow_matching_loss": 0.020659992657601833, + "grad_norm": 0.18846213817596436, + "learning_rate": 2.2472059731125483e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007440419308841229, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029099744046106935, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0177908131852746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7025340735912323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22040, + "weighted_flow_matching_loss": 0.020659992657601833, + "weighted_rkd_loss": 0.01249871077015996 + }, + { + "epoch": 6.415478615071283, + "flow_matching_loss": 0.022317536734044553, + "grad_norm": 0.22404372692108154, + "learning_rate": 2.2464889993418003e-05, + "loss": 0.0351, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007424448849633336, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033230538247153163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018171951454132796, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7022946774959564, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22050, + "weighted_flow_matching_loss": 0.022317536734044553, + "weighted_rkd_loss": 0.012762168142944574 + }, + { + "epoch": 6.418388129182427, + "flow_matching_loss": 0.021817871928215028, + "grad_norm": 0.1803269237279892, + "learning_rate": 2.2457717988074047e-05, + "loss": 0.0339, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007091402355581522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029936045641079543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017176409345120193, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7020552396774292, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22060, + "weighted_flow_matching_loss": 0.021817871928215028, + "weighted_rkd_loss": 0.01205884087830782 + }, + { + "epoch": 6.4212976432935704, + "flow_matching_loss": 0.018065520841628314, + "grad_norm": 0.1761634200811386, + "learning_rate": 2.245054371727228e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072728254366666075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003200797806493938, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017746448703110217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7018157422542572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22070, + "weighted_flow_matching_loss": 0.018065520841628314, + "weighted_rkd_loss": 0.01245474824681878 + }, + { + "epoch": 6.424207157404713, + "flow_matching_loss": 0.020918654184788464, + "grad_norm": 0.16033661365509033, + "learning_rate": 2.2443367183192046e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007261448306962848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003241943498142064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01776484027504921, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7015761733055115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22080, + "weighted_flow_matching_loss": 0.020918654184788464, + "weighted_rkd_loss": 0.012463553715497256 + }, + { + "epoch": 6.427116671515857, + "flow_matching_loss": 0.024347754940390588, + "grad_norm": 0.2214665710926056, + "learning_rate": 2.2436188388013394e-05, + "loss": 0.0361, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006819738214835525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00311642843298614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016755904629826547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7013365626335144, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22090, + "weighted_flow_matching_loss": 0.024347754940390588, + "weighted_rkd_loss": 0.011751567665487527 + }, + { + "epoch": 6.430026185627001, + "flow_matching_loss": 0.02128191515803337, + "grad_norm": 0.23309652507305145, + "learning_rate": 2.2429007333917048e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007716509234160185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034759202040731906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01890893867239356, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.701096898317337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22100, + "weighted_flow_matching_loss": 0.02128191515803337, + "weighted_rkd_loss": 0.013256859965622425 + }, + { + "epoch": 6.4329356997381435, + "flow_matching_loss": 0.021457244269549848, + "grad_norm": 0.16701637208461761, + "learning_rate": 2.242182402308443e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006642663944512605, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027213783003389837, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016006706096231938, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7008571743965148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22110, + "weighted_flow_matching_loss": 0.021457244269549848, + "weighted_rkd_loss": 0.01121844733133912 + }, + { + "epoch": 6.435845213849287, + "flow_matching_loss": 0.02069199150428176, + "grad_norm": 0.1638089269399643, + "learning_rate": 2.2414638457697632e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007581889722496271, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032998825656250117, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018463661707937718, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7006174027919769, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22120, + "weighted_flow_matching_loss": 0.02069199150428176, + "weighted_rkd_loss": 0.01293608695268631 + }, + { + "epoch": 6.438754727960431, + "flow_matching_loss": 0.020418067835271357, + "grad_norm": 0.18787162005901337, + "learning_rate": 2.2407450639939448e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007703399891033769, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032445897813886404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018651389610022308, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7003775596618652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22130, + "weighted_flow_matching_loss": 0.020418067835271357, + "weighted_rkd_loss": 0.013063083402812481 + }, + { + "epoch": 6.441664242071574, + "flow_matching_loss": 0.019012821093201637, + "grad_norm": 0.2047697752714157, + "learning_rate": 2.2400260571993333e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007112670643255115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029254604829475283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017150801606476307, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.7001376748085022, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22140, + "weighted_flow_matching_loss": 0.019012821093201637, + "weighted_rkd_loss": 0.012007949315011502 + }, + { + "epoch": 6.444573756182717, + "flow_matching_loss": 0.02134476862847805, + "grad_norm": 0.18007220327854156, + "learning_rate": 2.2393068256043457e-05, + "loss": 0.0345, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007783335540443659, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032091882778331636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01877585919573903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6998977363109589, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22150, + "weighted_flow_matching_loss": 0.02134476862847805, + "weighted_rkd_loss": 0.013141240645200013 + }, + { + "epoch": 6.447483270293861, + "flow_matching_loss": 0.019366031792014838, + "grad_norm": 0.16956613957881927, + "learning_rate": 2.2385873694274646e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006916631245985627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002962321531958878, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01679558428004384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6996577322483063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22160, + "weighted_flow_matching_loss": 0.019366031792014838, + "weighted_rkd_loss": 0.011751228757202626 + }, + { + "epoch": 6.450392784405004, + "flow_matching_loss": 0.020064113289117814, + "grad_norm": 0.17224647104740143, + "learning_rate": 2.2378676888872418e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068907846696674825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029704596381634474, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016752028744667767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6994176805019379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22170, + "weighted_flow_matching_loss": 0.020064113289117814, + "weighted_rkd_loss": 0.011716631520539523 + }, + { + "epoch": 6.453302298516148, + "flow_matching_loss": 0.021389169059693812, + "grad_norm": 0.17821113765239716, + "learning_rate": 2.2371477842022976e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006604887871071696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002964980318211019, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016174756083637477, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6991775751113891, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22180, + "weighted_flow_matching_loss": 0.021389169059693812, + "weighted_rkd_loss": 0.011308902222663165 + }, + { + "epoch": 6.456211812627291, + "flow_matching_loss": 0.02116899024695158, + "grad_norm": 0.20399145781993866, + "learning_rate": 2.2364276555913195e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070424539037048815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031194361858069898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017204343900084497, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6989374041557312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22190, + "weighted_flow_matching_loss": 0.02116899024695158, + "weighted_rkd_loss": 0.012024763226509094 + }, + { + "epoch": 6.459121326738435, + "flow_matching_loss": 0.021121476776897907, + "grad_norm": 0.20775574445724487, + "learning_rate": 2.235707303273064e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007302099466323852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003062096796929836, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017666295543313025, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6986971795558929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22200, + "weighted_flow_matching_loss": 0.021121476776897907, + "weighted_rkd_loss": 0.012343409098684787 + }, + { + "epoch": 6.462030840849578, + "flow_matching_loss": 0.01942703127861023, + "grad_norm": 0.16945430636405945, + "learning_rate": 2.234986727466355e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071181896142661575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032625240506604313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017498903442174196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6984569251537323, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22210, + "weighted_flow_matching_loss": 0.01942703127861023, + "weighted_rkd_loss": 0.01222219504415989 + }, + { + "epoch": 6.464940354960722, + "flow_matching_loss": 0.020441916398704053, + "grad_norm": 0.20570506155490875, + "learning_rate": 2.2342659283900843e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007164524961262941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003305806522257626, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01763485660776496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6982165992259979, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22220, + "weighted_flow_matching_loss": 0.020441916398704053, + "weighted_rkd_loss": 0.012312971614301204 + }, + { + "epoch": 6.467849869071865, + "flow_matching_loss": 0.020349610969424246, + "grad_norm": 0.15493936836719513, + "learning_rate": 2.2335449062632117e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006918872380629182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031433323165401816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016981077007949354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6979762196540833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22230, + "weighted_flow_matching_loss": 0.020349610969424246, + "weighted_rkd_loss": 0.011852499470114708 + }, + { + "epoch": 6.470759383183008, + "flow_matching_loss": 0.020720424316823482, + "grad_norm": 0.1674731969833374, + "learning_rate": 2.232823661304765e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007072573062032461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002984225680120289, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017129371874034405, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6977357864379883, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22240, + "weighted_flow_matching_loss": 0.020720424316823482, + "weighted_rkd_loss": 0.011951796244829893 + }, + { + "epoch": 6.473668897294152, + "flow_matching_loss": 0.02200215179473162, + "grad_norm": 0.19721892476081848, + "learning_rate": 2.2321021937338392e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007144068880006671, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003096538409590721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017384676076471806, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6974952995777131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22250, + "weighted_flow_matching_loss": 0.02200215179473162, + "weighted_rkd_loss": 0.012125784903764725 + }, + { + "epoch": 6.4765784114052956, + "flow_matching_loss": 0.02137074349448085, + "grad_norm": 0.21080386638641357, + "learning_rate": 2.2313805037695965e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007486569881439209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003165484662167728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018138624634593724, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.697254753112793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22260, + "weighted_flow_matching_loss": 0.02137074349448085, + "weighted_rkd_loss": 0.012647124845534563 + }, + { + "epoch": 6.479487925516438, + "flow_matching_loss": 0.01777010653167963, + "grad_norm": 0.15920668840408325, + "learning_rate": 2.230658591631268e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006834412505850196, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029875258915126324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01665635099634528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6970141589641571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22270, + "weighted_flow_matching_loss": 0.01777010653167963, + "weighted_rkd_loss": 0.011609776224941016 + }, + { + "epoch": 6.482397439627582, + "flow_matching_loss": 0.020412917621433734, + "grad_norm": 0.18052563071250916, + "learning_rate": 2.2299364575381512e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006586211360991001, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002704743284266442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01587716620415449, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6967735111713409, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22280, + "weighted_flow_matching_loss": 0.020412917621433734, + "weighted_rkd_loss": 0.011062749195843935 + }, + { + "epoch": 6.485306953738726, + "flow_matching_loss": 0.019621992483735085, + "grad_norm": 0.17186619341373444, + "learning_rate": 2.2292141017096115e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006681843753904104, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029843842843547464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016348071955144407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6965328097343445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22290, + "weighted_flow_matching_loss": 0.019621992483735085, + "weighted_rkd_loss": 0.01138700470328331 + }, + { + "epoch": 6.488216467849869, + "flow_matching_loss": 0.020340745616704226, + "grad_norm": 0.1721014678478241, + "learning_rate": 2.228491524365081e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007263781176880002, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029891248792409896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01751668704673648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6962920427322388, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22300, + "weighted_flow_matching_loss": 0.020340745616704226, + "weighted_rkd_loss": 0.012196792010217904 + }, + { + "epoch": 6.491125981961012, + "flow_matching_loss": 0.023809326626360417, + "grad_norm": 0.18813888728618622, + "learning_rate": 2.227768725724061e-05, + "loss": 0.0354, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006770521914586425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003050666127819568, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01659170975908637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6960512399673462, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22310, + "weighted_flow_matching_loss": 0.023809326626360417, + "weighted_rkd_loss": 0.011548594571650028 + }, + { + "epoch": 6.494035496072156, + "flow_matching_loss": 0.02110514435917139, + "grad_norm": 0.19633372128009796, + "learning_rate": 2.227045706006117e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00734067615121603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003094896674156189, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01777624888345599, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6958103597164154, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22320, + "weighted_flow_matching_loss": 0.02110514435917139, + "weighted_rkd_loss": 0.012368873413652181 + }, + { + "epoch": 6.4969450101833, + "flow_matching_loss": 0.02240106500685215, + "grad_norm": 0.16959857940673828, + "learning_rate": 2.2263224654308845e-05, + "loss": 0.0349, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073685530107468365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003264875872991979, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018001982010900974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6955694496631623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22330, + "weighted_flow_matching_loss": 0.02240106500685215, + "weighted_rkd_loss": 0.012521690782159567 + }, + { + "epoch": 6.4998545242944425, + "flow_matching_loss": 0.020430783927440643, + "grad_norm": 0.19673611223697662, + "learning_rate": 2.2255990042180646e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007272741524502635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030496766557917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01759515982121229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6953284740447998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22340, + "weighted_flow_matching_loss": 0.020430783927440643, + "weighted_rkd_loss": 0.012234329152852297 + }, + { + "epoch": 6.502764038405586, + "flow_matching_loss": 0.019001948926597834, + "grad_norm": 0.16298145055770874, + "learning_rate": 2.2248753225874252e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006358599895611406, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002691655047237873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015408854838460684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6950874507427216, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22350, + "weighted_flow_matching_loss": 0.019001948926597834, + "weighted_rkd_loss": 0.010710597690194845 + }, + { + "epoch": 6.50567355251673, + "flow_matching_loss": 0.019452972710132597, + "grad_norm": 0.18880881369113922, + "learning_rate": 2.2241514207588018e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006303946999832988, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028419142588973047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015449808537960052, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6948463618755341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22360, + "weighted_flow_matching_loss": 0.019452972710132597, + "weighted_rkd_loss": 0.010735222976654768 + }, + { + "epoch": 6.508583066627873, + "flow_matching_loss": 0.019386610575020315, + "grad_norm": 0.22199052572250366, + "learning_rate": 2.223427298952097e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072590299416333435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033713505836203694, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017889410629868506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6946052372455597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22370, + "weighted_flow_matching_loss": 0.019386610575020315, + "weighted_rkd_loss": 0.01242592018097639 + }, + { + "epoch": 6.5114925807390165, + "flow_matching_loss": 0.020531590282917022, + "grad_norm": 0.2034223973751068, + "learning_rate": 2.2227029573872803e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006665352452546358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00290267551317811, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016233380511403085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.694364047050476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22380, + "weighted_flow_matching_loss": 0.020531590282917022, + "weighted_rkd_loss": 0.011271871533244849 + }, + { + "epoch": 6.51440209485016, + "flow_matching_loss": 0.020777556765824556, + "grad_norm": 0.1846657395362854, + "learning_rate": 2.2219783962843866e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005990069266408682, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002458438731264323, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01443857727572322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6941228091716767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22390, + "weighted_flow_matching_loss": 0.020777556765824556, + "weighted_rkd_loss": 0.010022131633013487 + }, + { + "epoch": 6.517311608961304, + "flow_matching_loss": 0.02006683312356472, + "grad_norm": 0.21245509386062622, + "learning_rate": 2.221253615863519e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006574109755456447, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002962880639825016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01611110018566251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6938815176486969, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22400, + "weighted_flow_matching_loss": 0.02006683312356472, + "weighted_rkd_loss": 0.011179053876549005 + }, + { + "epoch": 6.520221123072447, + "flow_matching_loss": 0.020183677971363067, + "grad_norm": 0.16883791983127594, + "learning_rate": 2.2205286163448466e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006460566492751241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026706710690632463, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015591803938150406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6936401665210724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22410, + "weighted_flow_matching_loss": 0.020183677971363067, + "weighted_rkd_loss": 0.010815128870308399 + }, + { + "epoch": 6.52313063718359, + "flow_matching_loss": 0.020596611127257348, + "grad_norm": 0.1669422686100006, + "learning_rate": 2.219803397948604e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006290427129715681, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027958233375102283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01537667764350772, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6933987677097321, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22420, + "weighted_flow_matching_loss": 0.020596611127257348, + "weighted_rkd_loss": 0.010662187449634075 + }, + { + "epoch": 6.526040151294734, + "flow_matching_loss": 0.02262784242630005, + "grad_norm": 0.15148845314979553, + "learning_rate": 2.2190779608950953e-05, + "loss": 0.0344, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006849741563200951, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032234824262559413, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01692296555265784, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6931573152542114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22430, + "weighted_flow_matching_loss": 0.02262784242630005, + "weighted_rkd_loss": 0.011730351531878114 + }, + { + "epoch": 6.528949665405877, + "flow_matching_loss": 0.020993330143392087, + "grad_norm": 0.1872478425502777, + "learning_rate": 2.218352305404688e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007774314424023032, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003637093910947442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019185722526162862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6929158091545105, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22440, + "weighted_flow_matching_loss": 0.020993330143392087, + "weighted_rkd_loss": 0.013294034358114003 + }, + { + "epoch": 6.531859179517021, + "flow_matching_loss": 0.018652412947267293, + "grad_norm": 0.1895076483488083, + "learning_rate": 2.217626431697818e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006870286911725998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028959894087165596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01663656309247017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6926742553710937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22450, + "weighted_flow_matching_loss": 0.018652412947267293, + "weighted_rkd_loss": 0.011523723974823952 + }, + { + "epoch": 6.534768693628164, + "flow_matching_loss": 0.018063431326299905, + "grad_norm": 0.26935240626335144, + "learning_rate": 2.216900339994985e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007081820396706462, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030374899390153585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017201130837202074, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6924326360225678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22460, + "weighted_flow_matching_loss": 0.018063431326299905, + "weighted_rkd_loss": 0.011910588713362813 + }, + { + "epoch": 6.537678207739307, + "flow_matching_loss": 0.02236155457794666, + "grad_norm": 0.18861764669418335, + "learning_rate": 2.2161740305167574e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069127557799220085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002964668767526746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016790180280804635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6921909749507904, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22470, + "weighted_flow_matching_loss": 0.02236155457794666, + "weighted_rkd_loss": 0.011622007563710212 + }, + { + "epoch": 6.540587721850451, + "flow_matching_loss": 0.022429322358220816, + "grad_norm": 0.18430104851722717, + "learning_rate": 2.215447503483769e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006397395767271519, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002709352783858776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015504144318401813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6919492602348327, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22480, + "weighted_flow_matching_loss": 0.022429322358220816, + "weighted_rkd_loss": 0.010728115355595947 + }, + { + "epoch": 6.543497235961595, + "flow_matching_loss": 0.021676387079060076, + "grad_norm": 0.20596691966056824, + "learning_rate": 2.21472075911672e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072489514015614985, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032834324985742567, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017781335022300483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6917074799537659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22490, + "weighted_flow_matching_loss": 0.021676387079060076, + "weighted_rkd_loss": 0.012299463525414467 + }, + { + "epoch": 6.546406750072737, + "flow_matching_loss": 0.021197991259396077, + "grad_norm": 0.19588157534599304, + "learning_rate": 2.2139937976363747e-05, + "loss": 0.0332, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007139527797698974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003126326762139797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01740538254380226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6914656698703766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22500, + "weighted_flow_matching_loss": 0.021197991259396077, + "weighted_rkd_loss": 0.012035253830254078 + }, + { + "epoch": 6.549316264183881, + "flow_matching_loss": 0.020078801456838845, + "grad_norm": 0.2007140964269638, + "learning_rate": 2.2132666192635658e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006144855124875903, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027352719800546764, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015024982206523419, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6912237942218781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22510, + "weighted_flow_matching_loss": 0.020078801456838845, + "weighted_rkd_loss": 0.01038551600649953 + }, + { + "epoch": 6.552225778295025, + "flow_matching_loss": 0.02035771096125245, + "grad_norm": 0.17147108912467957, + "learning_rate": 2.2125392242191905e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006424012128263712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00280483877286315, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015652862936258317, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6909818649291992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22520, + "weighted_flow_matching_loss": 0.02035771096125245, + "weighted_rkd_loss": 0.010815805476158857 + }, + { + "epoch": 6.5551352924061685, + "flow_matching_loss": 0.02067249622195959, + "grad_norm": 0.19530093669891357, + "learning_rate": 2.2118116127242128e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006400687992572785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028024771832861005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015603853203356267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6907398700714111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22530, + "weighted_flow_matching_loss": 0.02067249622195959, + "weighted_rkd_loss": 0.010778034571558237 + }, + { + "epoch": 6.558044806517311, + "flow_matching_loss": 0.020059451460838318, + "grad_norm": 0.16812610626220703, + "learning_rate": 2.2110837849996614e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007444362435489893, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003121387679129839, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018010112456977367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6904978573322296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22540, + "weighted_flow_matching_loss": 0.020059451460838318, + "weighted_rkd_loss": 0.012435914017260075 + }, + { + "epoch": 6.560954320628455, + "flow_matching_loss": 0.020146800950169564, + "grad_norm": 0.169820174574852, + "learning_rate": 2.210355741266631e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006854182342067361, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030211380450055002, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016729502752423287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6902557730674743, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22550, + "weighted_flow_matching_loss": 0.020146800950169564, + "weighted_rkd_loss": 0.01154767656698823 + }, + { + "epoch": 6.563863834739599, + "flow_matching_loss": 0.01727627422660589, + "grad_norm": 0.1955685168504715, + "learning_rate": 2.2096274817462823e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073930793441832066, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031648882664740085, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017951046768575906, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6900136411190033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22560, + "weighted_flow_matching_loss": 0.01727627422660589, + "weighted_rkd_loss": 0.01238647075369954 + }, + { + "epoch": 6.566773348850742, + "flow_matching_loss": 0.021652545034885406, + "grad_norm": 0.22423189878463745, + "learning_rate": 2.208899006659841e-05, + "loss": 0.0341, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007337030488997698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0034025658387690783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01807662695646286, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6897714495658874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22570, + "weighted_flow_matching_loss": 0.021652545034885406, + "weighted_rkd_loss": 0.012468592170625925 + }, + { + "epoch": 6.569682862961885, + "flow_matching_loss": 0.019653791561722755, + "grad_norm": 0.19317874312400818, + "learning_rate": 2.2081703162285994e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006995460530743003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030034355353564025, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016994356550276278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6895292043685913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22580, + "weighted_flow_matching_loss": 0.019653791561722755, + "weighted_rkd_loss": 0.011718119028955698 + }, + { + "epoch": 6.572592377073029, + "flow_matching_loss": 0.019694245420396328, + "grad_norm": 0.19088448584079742, + "learning_rate": 2.2074414106739133e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005823579337447882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002422392519656569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014069551136344671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6892869174480438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22590, + "weighted_flow_matching_loss": 0.019694245420396328, + "weighted_rkd_loss": 0.009697862202301621 + }, + { + "epoch": 6.575501891184173, + "flow_matching_loss": 0.02202911991626024, + "grad_norm": 0.18435361981391907, + "learning_rate": 2.2067122902172062e-05, + "loss": 0.0347, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007554571190848946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032603591214865448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01836950145661831, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6890445709228515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22600, + "weighted_flow_matching_loss": 0.02202911991626024, + "weighted_rkd_loss": 0.012657444272190333 + }, + { + "epoch": 6.5784114052953155, + "flow_matching_loss": 0.021451655961573123, + "grad_norm": 0.17181231081485748, + "learning_rate": 2.2059829550799645e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007387669431045652, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031874874723143874, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017962826509028675, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6888021767139435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22610, + "weighted_flow_matching_loss": 0.021451655961573123, + "weighted_rkd_loss": 0.012372771510854363 + }, + { + "epoch": 6.581320919406459, + "flow_matching_loss": 0.02154626101255417, + "grad_norm": 0.1689303070306778, + "learning_rate": 2.2052534054837417e-05, + "loss": 0.034, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072880987077951435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00344824253115803, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018024439737200738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6885597288608551, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22620, + "weighted_flow_matching_loss": 0.02154626101255417, + "weighted_rkd_loss": 0.01241101250052452 + }, + { + "epoch": 6.584230433517602, + "flow_matching_loss": 0.02100566662847996, + "grad_norm": 0.1763397604227066, + "learning_rate": 2.2045236416501552e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007146280445158482, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003190913051366806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01748347347602248, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6883172333240509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22630, + "weighted_flow_matching_loss": 0.02100566662847996, + "weighted_rkd_loss": 0.012034250982105732 + }, + { + "epoch": 6.587139947628746, + "flow_matching_loss": 0.020404805429279804, + "grad_norm": 0.16637501120567322, + "learning_rate": 2.2037936638008885e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007044135499745608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029717213241383434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01705999216064811, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6880746781826019, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22640, + "weighted_flow_matching_loss": 0.020404805429279804, + "weighted_rkd_loss": 0.011738532967865468 + }, + { + "epoch": 6.5900494617398895, + "flow_matching_loss": 0.018066099472343923, + "grad_norm": 0.15939489006996155, + "learning_rate": 2.203063472157689e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006938046310096979, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002834686543792486, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016710779257118703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6878320813179016, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22650, + "weighted_flow_matching_loss": 0.018066099472343923, + "weighted_rkd_loss": 0.011494300421327352 + }, + { + "epoch": 6.592958975851033, + "flow_matching_loss": 0.01991468798369169, + "grad_norm": 0.2188820093870163, + "learning_rate": 2.2023330669423703e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006968698743730783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033066523727029564, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017244049906730653, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.687589418888092, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22660, + "weighted_flow_matching_loss": 0.01991468798369169, + "weighted_rkd_loss": 0.011856671748682856 + }, + { + "epoch": 6.595868489962176, + "flow_matching_loss": 0.018818315491080283, + "grad_norm": 0.17255248129367828, + "learning_rate": 2.20160244837681e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006383941927924752, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029506044229492545, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01571848839521408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6873467206954956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22670, + "weighted_flow_matching_loss": 0.018818315491080283, + "weighted_rkd_loss": 0.010804110113531351 + }, + { + "epoch": 6.59877800407332, + "flow_matching_loss": 0.02033174429088831, + "grad_norm": 0.14875428378582, + "learning_rate": 2.20087161668295e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067975400947034356, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029441518476232887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016539232153445483, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6871039628982544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22680, + "weighted_flow_matching_loss": 0.02033174429088831, + "weighted_rkd_loss": 0.011364216450601816 + }, + { + "epoch": 6.601687518184463, + "flow_matching_loss": 0.019278619159013032, + "grad_norm": 0.17051362991333008, + "learning_rate": 2.200140572082799e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007128042681142688, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029815302463248373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01723761549219489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6868611514568329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22690, + "weighted_flow_matching_loss": 0.019278619159013032, + "weighted_rkd_loss": 0.011839941050857306 + }, + { + "epoch": 6.604597032295606, + "flow_matching_loss": 0.023085123859345913, + "grad_norm": 0.16542492806911469, + "learning_rate": 2.1994093147984283e-05, + "loss": 0.0346, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006946682929992676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002859916700981557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016753282584249973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.68661829829216, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22700, + "weighted_flow_matching_loss": 0.023085123859345913, + "weighted_rkd_loss": 0.011503191106021405 + }, + { + "epoch": 6.60750654640675, + "flow_matching_loss": 0.02049761712551117, + "grad_norm": 0.1591278463602066, + "learning_rate": 2.1986778450519742e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006835496053099632, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002965678251348436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01663667047396302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6863753855228424, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22710, + "weighted_flow_matching_loss": 0.02049761712551117, + "weighted_rkd_loss": 0.01141912629827857 + }, + { + "epoch": 6.610416060517894, + "flow_matching_loss": 0.019384615309536456, + "grad_norm": 0.18473108112812042, + "learning_rate": 2.1979461630656384e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068148689344525335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00284400403033942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016473741736263037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6861324191093445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22720, + "weighted_flow_matching_loss": 0.019384615309536456, + "weighted_rkd_loss": 0.011303199455142021 + }, + { + "epoch": 6.613325574629037, + "flow_matching_loss": 0.021642466075718403, + "grad_norm": 0.18198710680007935, + "learning_rate": 2.1972142690616866e-05, + "loss": 0.0331, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068456839304417375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00297420856077224, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01666557639837265, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6858894109725953, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22730, + "weighted_flow_matching_loss": 0.021642466075718403, + "weighted_rkd_loss": 0.011430754233151675 + }, + { + "epoch": 6.61623508874018, + "flow_matching_loss": 0.020411107409745454, + "grad_norm": 0.17063742876052856, + "learning_rate": 2.1964821632624484e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007098101265728474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003174475557170808, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017370678018778563, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6856463432312012, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22740, + "weighted_flow_matching_loss": 0.020411107409745454, + "weighted_rkd_loss": 0.01191010857000947 + }, + { + "epoch": 6.619144602851324, + "flow_matching_loss": 0.021046852692961693, + "grad_norm": 0.2124975323677063, + "learning_rate": 2.1957498458903188e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007586035737767815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033116083359345794, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018483679834753276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6854032278060913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22750, + "weighted_flow_matching_loss": 0.021046852692961693, + "weighted_rkd_loss": 0.012668797839432954 + }, + { + "epoch": 6.622054116962468, + "flow_matching_loss": 0.020813525188714266, + "grad_norm": 0.18098023533821106, + "learning_rate": 2.195017317167756e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065868858247995375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002899384894408286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01607315642759204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6851600408554077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22760, + "weighted_flow_matching_loss": 0.020813525188714266, + "weighted_rkd_loss": 0.011012679804116488 + }, + { + "epoch": 6.62496363107361, + "flow_matching_loss": 0.01993488324806094, + "grad_norm": 0.19068115949630737, + "learning_rate": 2.1942845773172836e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070648223627358675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031235184287652373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017253163270652294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6849168479442597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22770, + "weighted_flow_matching_loss": 0.01993488324806094, + "weighted_rkd_loss": 0.01181694194674492 + }, + { + "epoch": 6.627873145184754, + "flow_matching_loss": 0.01976005993783474, + "grad_norm": 0.15708398818969727, + "learning_rate": 2.193551626561488e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007402111450210214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003183718724176288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0179879417642951, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6846735835075378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22780, + "weighted_flow_matching_loss": 0.01976005993783474, + "weighted_rkd_loss": 0.012315992265939713 + }, + { + "epoch": 6.630782659295898, + "flow_matching_loss": 0.020642843656241895, + "grad_norm": 0.2054407298564911, + "learning_rate": 2.19281846512302e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006915809819474816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030042119324207306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016835831664502622, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6844302654266358, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22790, + "weighted_flow_matching_loss": 0.020642843656241895, + "weighted_rkd_loss": 0.011522933561354875 + }, + { + "epoch": 6.633692173407041, + "flow_matching_loss": 0.020270161610096694, + "grad_norm": 0.1744367480278015, + "learning_rate": 2.1920850932245955e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006492630718275905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027194939902983606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015704755671322346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6841869056224823, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22800, + "weighted_flow_matching_loss": 0.020270161610096694, + "weighted_rkd_loss": 0.010745020490139723 + }, + { + "epoch": 6.636601687518184, + "flow_matching_loss": 0.0199552021920681, + "grad_norm": 0.19297373294830322, + "learning_rate": 2.1913515110889932e-05, + "loss": 0.0325, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007531436858698726, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033313764492049814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01839424967765808, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6839434802532196, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22810, + "weighted_flow_matching_loss": 0.0199552021920681, + "weighted_rkd_loss": 0.012580592464655638 + }, + { + "epoch": 6.639511201629328, + "flow_matching_loss": 0.020379336923360823, + "grad_norm": 0.19105465710163116, + "learning_rate": 2.1906177189390557e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006857056636363268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002892902330495417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016607015579938888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6837000131607056, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22820, + "weighted_flow_matching_loss": 0.020379336923360823, + "weighted_rkd_loss": 0.011354181915521622 + }, + { + "epoch": 6.642420715740471, + "flow_matching_loss": 0.021280363202095032, + "grad_norm": 0.18147733807563782, + "learning_rate": 2.18988371699769e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007531027216464281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003215413959696889, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01827746843919158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6834564924240112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22830, + "weighted_flow_matching_loss": 0.021280363202095032, + "weighted_rkd_loss": 0.012491921335458756 + }, + { + "epoch": 6.645330229851615, + "flow_matching_loss": 0.018409256637096406, + "grad_norm": 0.1519308090209961, + "learning_rate": 2.189149505487866e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007194479927420616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003063133463729173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01745209312066436, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6832129180431366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22840, + "weighted_flow_matching_loss": 0.018409256637096406, + "weighted_rkd_loss": 0.011923705413937568 + }, + { + "epoch": 6.648239743962758, + "flow_matching_loss": 0.021242856793105603, + "grad_norm": 0.2035573124885559, + "learning_rate": 2.1884150846326186e-05, + "loss": 0.0337, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007464397605508566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032453578663989902, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0181741532869637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6829693019390106, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22850, + "weighted_flow_matching_loss": 0.021242856793105603, + "weighted_rkd_loss": 0.012412288505584002 + }, + { + "epoch": 6.651149258073902, + "flow_matching_loss": 0.02013679500669241, + "grad_norm": 0.17776024341583252, + "learning_rate": 2.187680454655045e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007614656537771225, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033722586231306195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018601571768522264, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6827256321907044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22860, + "weighted_flow_matching_loss": 0.02013679500669241, + "weighted_rkd_loss": 0.012699691951274872 + }, + { + "epoch": 6.654058772185045, + "flow_matching_loss": 0.01942859273403883, + "grad_norm": 0.17667783796787262, + "learning_rate": 2.186945615778306e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073615176603198055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031353789381682873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017858414072543384, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6824819147586823, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22870, + "weighted_flow_matching_loss": 0.01942859273403883, + "weighted_rkd_loss": 0.01218806579709053 + }, + { + "epoch": 6.6569682862961885, + "flow_matching_loss": 0.02153726890683174, + "grad_norm": 0.1721276193857193, + "learning_rate": 2.186210568225627e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006846582982689142, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003005362185649574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016698528360575438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6822381317615509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22880, + "weighted_flow_matching_loss": 0.02153726890683174, + "weighted_rkd_loss": 0.011392292566597462 + }, + { + "epoch": 6.659877800407332, + "flow_matching_loss": 0.01866226065903902, + "grad_norm": 0.17329926788806915, + "learning_rate": 2.1854753122202955e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006851608632132411, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029581632697954774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016661380417644978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6819943249225616, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22890, + "weighted_flow_matching_loss": 0.01866226065903902, + "weighted_rkd_loss": 0.01136311199516058 + }, + { + "epoch": 6.662787314518475, + "flow_matching_loss": 0.019075040891766547, + "grad_norm": 0.15475228428840637, + "learning_rate": 2.1847398479856638e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006390384072437883, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027683872729539873, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015549155417829752, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6817504525184631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22900, + "weighted_flow_matching_loss": 0.019075040891766547, + "weighted_rkd_loss": 0.010600647330284119 + }, + { + "epoch": 6.665696828629619, + "flow_matching_loss": 0.020466025173664092, + "grad_norm": 0.203701451420784, + "learning_rate": 2.184004175745145e-05, + "loss": 0.0321, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006933587975800038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031550890067592262, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017022265028208493, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6815065383911133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22910, + "weighted_flow_matching_loss": 0.020466025173664092, + "weighted_rkd_loss": 0.011600717250257731 + }, + { + "epoch": 6.6686063427407625, + "flow_matching_loss": 0.020096416492015122, + "grad_norm": 0.17808414995670319, + "learning_rate": 2.1832682957222182e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007017717882990837, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031160641927272083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01715150000527501, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6812625646591186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22920, + "weighted_flow_matching_loss": 0.020096416492015122, + "weighted_rkd_loss": 0.011684755235910416 + }, + { + "epoch": 6.671515856851906, + "flow_matching_loss": 0.021853104792535306, + "grad_norm": 0.20331601798534393, + "learning_rate": 2.1825322081404236e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070992860943079, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030376069946214555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01723617920652032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6810185492038727, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22930, + "weighted_flow_matching_loss": 0.021853104792535306, + "weighted_rkd_loss": 0.011738149635493755 + }, + { + "epoch": 6.674425370963049, + "flow_matching_loss": 0.020476136822253466, + "grad_norm": 0.17838092148303986, + "learning_rate": 2.1817959132233655e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00694957678206265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003060314676258713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016959467995911838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6807744801044464, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22940, + "weighted_flow_matching_loss": 0.020476136822253466, + "weighted_rkd_loss": 0.011545604374259711 + }, + { + "epoch": 6.677334885074193, + "flow_matching_loss": 0.019013004284352063, + "grad_norm": 0.22044789791107178, + "learning_rate": 2.1810594111947104e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070977875962853435, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003155130334198475, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017350705247372387, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6805303573608399, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22950, + "weighted_flow_matching_loss": 0.019013004284352063, + "weighted_rkd_loss": 0.011807622853666544 + }, + { + "epoch": 6.680244399185336, + "flow_matching_loss": 0.020806231535971165, + "grad_norm": 0.1414313018321991, + "learning_rate": 2.1803227022781892e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006103521771728992, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025951590156182646, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014802202489227057, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.680286192893982, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22960, + "weighted_flow_matching_loss": 0.020806231535971165, + "weighted_rkd_loss": 0.010069770272821188 + }, + { + "epoch": 6.683153913296479, + "flow_matching_loss": 0.0195733405649662, + "grad_norm": 0.167425736784935, + "learning_rate": 2.1795857866975933e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065901285968720915, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029143128311261536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01609457014128566, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6800419747829437, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22970, + "weighted_flow_matching_loss": 0.0195733405649662, + "weighted_rkd_loss": 0.010944954957813025 + }, + { + "epoch": 6.686063427407623, + "flow_matching_loss": 0.0204896729439497, + "grad_norm": 0.20435938239097595, + "learning_rate": 2.178848664676779e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007137733325362205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003113747714087367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017389214225113393, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6797977030277252, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22980, + "weighted_flow_matching_loss": 0.0204896729439497, + "weighted_rkd_loss": 0.011821131967008115 + }, + { + "epoch": 6.688972941518767, + "flow_matching_loss": 0.020840760599821807, + "grad_norm": 0.18469107151031494, + "learning_rate": 2.1781113364396643e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007143730018287897, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003268267074599862, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01755572697147727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6795533895492554, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 22990, + "weighted_flow_matching_loss": 0.020840760599821807, + "weighted_rkd_loss": 0.011929965857416392 + }, + { + "epoch": 6.6918824556299095, + "flow_matching_loss": 0.021130226925015448, + "grad_norm": 0.17272791266441345, + "learning_rate": 2.17737380221023e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007606753753498197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00336509607732296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018578603398054838, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6793090105056763, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23000, + "weighted_flow_matching_loss": 0.021130226925015448, + "weighted_rkd_loss": 0.012620763015002012 + }, + { + "epoch": 6.694791969741053, + "flow_matching_loss": 0.018842308316379786, + "grad_norm": 0.17438410222530365, + "learning_rate": 2.1766360622125198e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069462890736758705, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029505171114578843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016843095235526562, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6790646135807037, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23010, + "weighted_flow_matching_loss": 0.018842308316379786, + "weighted_rkd_loss": 0.011437662225216628 + }, + { + "epoch": 6.697701483852197, + "flow_matching_loss": 0.019061992503702642, + "grad_norm": 0.1611284613609314, + "learning_rate": 2.175898116670639e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006743737589567899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028628080617636442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016350283287465572, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.678820151090622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23020, + "weighted_flow_matching_loss": 0.019061992503702642, + "weighted_rkd_loss": 0.01109895296394825 + }, + { + "epoch": 6.70061099796334, + "flow_matching_loss": 0.019872472621500492, + "grad_norm": 0.1602354794740677, + "learning_rate": 2.175159965808757e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007079365011304617, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032707146368920804, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017429444938898087, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6785756349563599, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23030, + "weighted_flow_matching_loss": 0.019872472621500492, + "weighted_rkd_loss": 0.011827246751636267 + }, + { + "epoch": 6.703520512074483, + "flow_matching_loss": 0.019530495442450047, + "grad_norm": 0.1710231453180313, + "learning_rate": 2.1744216098511038e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071167512331157924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003248340776190162, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01748184310272336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6783310651779175, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23040, + "weighted_flow_matching_loss": 0.019530495442450047, + "weighted_rkd_loss": 0.011858467012643814 + }, + { + "epoch": 6.706430026185627, + "flow_matching_loss": 0.020085647609084843, + "grad_norm": 0.16265611350536346, + "learning_rate": 2.1736830490219722e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006449410552158951, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002676089957822114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015574910957366229, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6780864655971527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23050, + "weighted_flow_matching_loss": 0.020085647609084843, + "weighted_rkd_loss": 0.010561248194426298 + }, + { + "epoch": 6.709339540296771, + "flow_matching_loss": 0.021357670519500972, + "grad_norm": 0.15432360768318176, + "learning_rate": 2.1729442835457182e-05, + "loss": 0.0328, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006976095819845795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029810754116624595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016933266911655663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6778418123722076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23060, + "weighted_flow_matching_loss": 0.021357670519500972, + "weighted_rkd_loss": 0.011478130519390107 + }, + { + "epoch": 6.712249054407914, + "flow_matching_loss": 0.01934584239497781, + "grad_norm": 0.1945263147354126, + "learning_rate": 2.1722053136467594e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007232413347810507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003326447959989309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017791274841874836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6775970935821534, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23070, + "weighted_flow_matching_loss": 0.01934584239497781, + "weighted_rkd_loss": 0.012055221199989318 + }, + { + "epoch": 6.715158568519057, + "flow_matching_loss": 0.018901195377111435, + "grad_norm": 0.15882264077663422, + "learning_rate": 2.1714661395495743e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006867225654423237, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029837521724402905, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016718203388154506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6773523449897766, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23080, + "weighted_flow_matching_loss": 0.018901195377111435, + "weighted_rkd_loss": 0.011324105085805058 + }, + { + "epoch": 6.718068082630201, + "flow_matching_loss": 0.020350630581378936, + "grad_norm": 0.17528825998306274, + "learning_rate": 2.170726761478705e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006426106346771121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002642204612493515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015494417306035757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6771075367927551, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23090, + "weighted_flow_matching_loss": 0.020350630581378936, + "weighted_rkd_loss": 0.010491316206753254 + }, + { + "epoch": 6.720977596741344, + "flow_matching_loss": 0.019089158903807402, + "grad_norm": 0.203574076294899, + "learning_rate": 2.1699871796587554e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006972102774307131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003019001707434654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016963207442313433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6768626809120178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23100, + "weighted_flow_matching_loss": 0.019089158903807402, + "weighted_rkd_loss": 0.011481781024485826 + }, + { + "epoch": 6.723887110852488, + "flow_matching_loss": 0.020093489065766336, + "grad_norm": 0.18230539560317993, + "learning_rate": 2.1692473943143904e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006742691993713379, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031354714883491397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016620855405926706, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6766177713871002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23110, + "weighted_flow_matching_loss": 0.020093489065766336, + "weighted_rkd_loss": 0.011245987005531788 + }, + { + "epoch": 6.726796624963631, + "flow_matching_loss": 0.019501803815364836, + "grad_norm": 0.13613243401050568, + "learning_rate": 2.1685074056703373e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006231685355305672, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027970249415375293, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015260395593941212, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6763728439807892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23120, + "weighted_flow_matching_loss": 0.019501803815364836, + "weighted_rkd_loss": 0.01032176767475903 + }, + { + "epoch": 6.729706139074775, + "flow_matching_loss": 0.019735286943614483, + "grad_norm": 0.18778204917907715, + "learning_rate": 2.167767213951386e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065613111015409235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002899296104442328, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016021918319165707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6761278212070465, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23130, + "weighted_flow_matching_loss": 0.019735286943614483, + "weighted_rkd_loss": 0.010832898924127221 + }, + { + "epoch": 6.732615653185918, + "flow_matching_loss": 0.020808932092040777, + "grad_norm": 0.17345702648162842, + "learning_rate": 2.1670268193823853e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073791074100881815, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032375500770285724, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01799576496705413, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6758827805519104, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23140, + "weighted_flow_matching_loss": 0.020808932092040777, + "weighted_rkd_loss": 0.012162982299923897 + }, + { + "epoch": 6.7355251672970615, + "flow_matching_loss": 0.018169258534908295, + "grad_norm": 0.17359943687915802, + "learning_rate": 2.1662862221882495e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006241497024893761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026327292202040554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015115723200142383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6756376922130585, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23150, + "weighted_flow_matching_loss": 0.018169258534908295, + "weighted_rkd_loss": 0.01021276842802763 + }, + { + "epoch": 6.738434681408204, + "flow_matching_loss": 0.020993956923484804, + "grad_norm": 0.17755642533302307, + "learning_rate": 2.1655454225939514e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006929294019937515, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002929641946684569, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01678823009133339, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6753925442695617, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23160, + "weighted_flow_matching_loss": 0.020993956923484804, + "weighted_rkd_loss": 0.011338584776967763 + }, + { + "epoch": 6.741344195519348, + "flow_matching_loss": 0.020905961096286774, + "grad_norm": 0.19695225358009338, + "learning_rate": 2.1648044208245266e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00771062565036118, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003207463165745139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018628714606165886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6751473486423493, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23170, + "weighted_flow_matching_loss": 0.020905961096286774, + "weighted_rkd_loss": 0.012577132228761912 + }, + { + "epoch": 6.744253709630492, + "flow_matching_loss": 0.020452886819839478, + "grad_norm": 0.191159188747406, + "learning_rate": 2.164063217105072e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007007253030315042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00298406423535198, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01699857022613287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6749021112918854, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23180, + "weighted_flow_matching_loss": 0.020452886819839478, + "weighted_rkd_loss": 0.011472401581704617 + }, + { + "epoch": 6.7471632237416355, + "flow_matching_loss": 0.01952213766053319, + "grad_norm": 0.18641534447669983, + "learning_rate": 2.163321811660746e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067917924840003255, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031034384155645967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0166870235465467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6746568262577057, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23190, + "weighted_flow_matching_loss": 0.01952213766053319, + "weighted_rkd_loss": 0.011258145049214362 + }, + { + "epoch": 6.750072737852778, + "flow_matching_loss": 0.0204010215587914, + "grad_norm": 0.1578066349029541, + "learning_rate": 2.162580204716767e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00691004735417664, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029067585011944173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01672685332596302, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6744114875793457, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23200, + "weighted_flow_matching_loss": 0.0204010215587914, + "weighted_rkd_loss": 0.01128084324300289 + }, + { + "epoch": 6.752982251963922, + "flow_matching_loss": 0.02469745706766844, + "grad_norm": 0.17665384709835052, + "learning_rate": 2.1618383964984162e-05, + "loss": 0.0365, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007212860696017742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003073294460773468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01749901585280895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6741661071777344, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23210, + "weighted_flow_matching_loss": 0.02469745706766844, + "weighted_rkd_loss": 0.011797410901635885 + }, + { + "epoch": 6.755891766075066, + "flow_matching_loss": 0.017912314739078283, + "grad_norm": 0.1993674486875534, + "learning_rate": 2.1610963872310358e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006436016503721475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002646497776731849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015518530737608671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6739206731319427, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23220, + "weighted_flow_matching_loss": 0.017912314739078283, + "weighted_rkd_loss": 0.010458275023847819 + }, + { + "epoch": 6.7588012801862085, + "flow_matching_loss": 0.01888960925862193, + "grad_norm": 0.16315624117851257, + "learning_rate": 2.160354177140028e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006913519371300936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029016763321124016, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016728715039789677, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6736751973628998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23230, + "weighted_flow_matching_loss": 0.01888960925862193, + "weighted_rkd_loss": 0.011269703228026628 + }, + { + "epoch": 6.761710794297352, + "flow_matching_loss": 0.02147912448272109, + "grad_norm": 0.1743931621313095, + "learning_rate": 2.159611766450857e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006417558947578073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002770518884062767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01560563687235117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6734296679496765, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23240, + "weighted_flow_matching_loss": 0.02147912448272109, + "weighted_rkd_loss": 0.010509283281862736 + }, + { + "epoch": 6.764620308408496, + "flow_matching_loss": 0.01978527707979083, + "grad_norm": 0.18651193380355835, + "learning_rate": 2.1588691553890475e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006562012480571866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002869198564440012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01599322324618697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6731840968132019, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23250, + "weighted_flow_matching_loss": 0.01978527707979083, + "weighted_rkd_loss": 0.010766385775059461 + }, + { + "epoch": 6.76752982251964, + "flow_matching_loss": 0.0185541027225554, + "grad_norm": 0.22135841846466064, + "learning_rate": 2.1581263441801846e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006961323413997888, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003054155968129635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01697680288925767, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6729384779930114, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23260, + "weighted_flow_matching_loss": 0.0185541027225554, + "weighted_rkd_loss": 0.01142439730465412 + }, + { + "epoch": 6.7704393366307825, + "flow_matching_loss": 0.019658693112432957, + "grad_norm": 0.17814944684505463, + "learning_rate": 2.157383333049915e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064008547458797695, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028340989723801614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0156358084641397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6726928055286407, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23270, + "weighted_flow_matching_loss": 0.019658693112432957, + "weighted_rkd_loss": 0.010518169216811656 + }, + { + "epoch": 6.773348850741926, + "flow_matching_loss": 0.01923607597127557, + "grad_norm": 0.17359000444412231, + "learning_rate": 2.156640122223947e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007254942459985614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032734033651649953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017783288098871707, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6724470853805542, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23280, + "weighted_flow_matching_loss": 0.01923607597127557, + "weighted_rkd_loss": 0.011958308890461922 + }, + { + "epoch": 6.77625836485307, + "flow_matching_loss": 0.020182670466601847, + "grad_norm": 0.1552312672138214, + "learning_rate": 2.1558967119280468e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00720603265799582, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032102330354973675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017622297909110785, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6722013235092164, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23290, + "weighted_flow_matching_loss": 0.020182670466601847, + "weighted_rkd_loss": 0.011845707520842552 + }, + { + "epoch": 6.779167878964213, + "flow_matching_loss": 0.021058348752558232, + "grad_norm": 0.2990102171897888, + "learning_rate": 2.155153102388043e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007544987788423896, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003538010409101844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018627986032515764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6719555139541626, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23300, + "weighted_flow_matching_loss": 0.021058348752558232, + "weighted_rkd_loss": 0.012517148721963167 + }, + { + "epoch": 6.782077393075356, + "flow_matching_loss": 0.021246164478361605, + "grad_norm": 0.2010045349597931, + "learning_rate": 2.1544092938298258e-05, + "loss": 0.0324, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006862454628571868, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029160822508856655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01664099171757698, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6717096567153931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23310, + "weighted_flow_matching_loss": 0.021246164478361605, + "weighted_rkd_loss": 0.011177921527996659 + }, + { + "epoch": 6.7849869071865, + "flow_matching_loss": 0.01939122239127755, + "grad_norm": 0.18457849323749542, + "learning_rate": 2.153665286479343e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007878581574186683, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033477509394288065, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019104914274066687, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6714637517929077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23320, + "weighted_flow_matching_loss": 0.01939122239127755, + "weighted_rkd_loss": 0.0128281744197011 + }, + { + "epoch": 6.787896421297643, + "flow_matching_loss": 0.01912709977477789, + "grad_norm": 0.20301952958106995, + "learning_rate": 2.1529210805626058e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00636297520250082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026939881383441387, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015419938694685697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6712177991867065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23330, + "weighted_flow_matching_loss": 0.01912709977477789, + "weighted_rkd_loss": 0.010350107215344906 + }, + { + "epoch": 6.790805935408787, + "flow_matching_loss": 0.01970209963619709, + "grad_norm": 0.20347346365451813, + "learning_rate": 2.1521766763056835e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006206745002418757, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026356380200013517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015049128327518702, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.670971792936325, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23340, + "weighted_flow_matching_loss": 0.01970209963619709, + "weighted_rkd_loss": 0.010097548365592957 + }, + { + "epoch": 6.79371544951993, + "flow_matching_loss": 0.019292674586176874, + "grad_norm": 0.15440915524959564, + "learning_rate": 2.151432073934707e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006274046283215285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025075664976611733, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015055659133940935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6707257568836212, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23350, + "weighted_flow_matching_loss": 0.019292674586176874, + "weighted_rkd_loss": 0.0100981748662889 + }, + { + "epoch": 6.796624963631073, + "flow_matching_loss": 0.020822099409997463, + "grad_norm": 0.18683485686779022, + "learning_rate": 2.1506872736758665e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006247005891054869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002552171191200614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01504618264734745, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.670479667186737, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23360, + "weighted_flow_matching_loss": 0.020822099409997463, + "weighted_rkd_loss": 0.010088109038770198 + }, + { + "epoch": 6.799534477742217, + "flow_matching_loss": 0.019709521532058717, + "grad_norm": 0.18874713778495789, + "learning_rate": 2.149942275755412e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006864600256085396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029262777417898176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016655478067696093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6702335178852081, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23370, + "weighted_flow_matching_loss": 0.019709521532058717, + "weighted_rkd_loss": 0.011163217294961215 + }, + { + "epoch": 6.802443991853361, + "flow_matching_loss": 0.02075937632471323, + "grad_norm": 0.22241084277629852, + "learning_rate": 2.1491970803996557e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006536502577364445, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027415968477725984, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015814602188766003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6699873387813569, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23380, + "weighted_flow_matching_loss": 0.02075937632471323, + "weighted_rkd_loss": 0.010595417115837336 + }, + { + "epoch": 6.805353505964504, + "flow_matching_loss": 0.018768104910850524, + "grad_norm": 0.19072435796260834, + "learning_rate": 2.1484516878349667e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006516501074656844, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027270914986729624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015760093554854394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6697411000728607, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23390, + "weighted_flow_matching_loss": 0.018768104910850524, + "weighted_rkd_loss": 0.010555236041545868 + }, + { + "epoch": 6.808263020075647, + "flow_matching_loss": 0.018929014075547458, + "grad_norm": 0.16195452213287354, + "learning_rate": 2.147706098287777e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006309584435075522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025811977684497833, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015200366638600827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6694948256015778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23400, + "weighted_flow_matching_loss": 0.018929014075547458, + "weighted_rkd_loss": 0.010176614299416542 + }, + { + "epoch": 6.811172534186791, + "flow_matching_loss": 0.019586792215704918, + "grad_norm": 0.18195337057113647, + "learning_rate": 2.146960311984576e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007331177406013012, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033330629114061592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017995417583733797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6692484974861145, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23410, + "weighted_flow_matching_loss": 0.019586792215704918, + "weighted_rkd_loss": 0.012043390795588494 + }, + { + "epoch": 6.8140820482979345, + "flow_matching_loss": 0.019896240439265966, + "grad_norm": 0.1805996596813202, + "learning_rate": 2.146214329151914e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006998887797817588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032117300666868686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017209505569189786, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6690021276473999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23420, + "weighted_flow_matching_loss": 0.019896240439265966, + "weighted_rkd_loss": 0.011513048596680165 + }, + { + "epoch": 6.816991562409077, + "flow_matching_loss": 0.022608136013150214, + "grad_norm": 0.2047712802886963, + "learning_rate": 2.1454681500164012e-05, + "loss": 0.0336, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006661507626995444, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003063993831165135, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016387009155005218, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6687557101249695, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23430, + "weighted_flow_matching_loss": 0.022608136013150214, + "weighted_rkd_loss": 0.010958833899348975 + }, + { + "epoch": 6.819901076520221, + "flow_matching_loss": 0.01980469897389412, + "grad_norm": 0.18408431112766266, + "learning_rate": 2.144721774804707e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006612137984484434, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026760191423818467, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01590029541403055, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6685092449188232, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23440, + "weighted_flow_matching_loss": 0.01980469897389412, + "weighted_rkd_loss": 0.010629504919052124 + }, + { + "epoch": 6.822810590631365, + "flow_matching_loss": 0.02023162813857198, + "grad_norm": 0.17439839243888855, + "learning_rate": 2.1439752037435607e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006353472312912345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002680550469085574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015387495141476392, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6682627320289611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23450, + "weighted_flow_matching_loss": 0.02023162813857198, + "weighted_rkd_loss": 0.010282968077808619 + }, + { + "epoch": 6.8257201047425085, + "flow_matching_loss": 0.018813247419893743, + "grad_norm": 0.16919292509555817, + "learning_rate": 2.1432284370597505e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005861172219738364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002514578483533114, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014236923214048148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6680161774158477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23460, + "weighted_flow_matching_loss": 0.018813247419893743, + "weighted_rkd_loss": 0.00951042165979743 + }, + { + "epoch": 6.828629618853651, + "flow_matching_loss": 0.022227277420461176, + "grad_norm": 0.17968080937862396, + "learning_rate": 2.1424814749801246e-05, + "loss": 0.0334, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006870363373309374, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029481473611667752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01668887408450246, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6677695751190186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23470, + "weighted_flow_matching_loss": 0.022227277420461176, + "weighted_rkd_loss": 0.01114447358995676 + }, + { + "epoch": 6.831539132964795, + "flow_matching_loss": 0.020059741381555796, + "grad_norm": 0.1923702508211136, + "learning_rate": 2.1417343177315905e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00659738527610898, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029126397799700498, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016107410378754138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.667522931098938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23480, + "weighted_flow_matching_loss": 0.020059741381555796, + "weighted_rkd_loss": 0.01075195400044322 + }, + { + "epoch": 6.834448647075939, + "flow_matching_loss": 0.018509510159492492, + "grad_norm": 0.21513402462005615, + "learning_rate": 2.1409869655411145e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00684030344709754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031647320836782457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016845338977873325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6672762334346771, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23490, + "weighted_flow_matching_loss": 0.018509510159492492, + "weighted_rkd_loss": 0.011240388080477715 + }, + { + "epoch": 6.8373581611870815, + "flow_matching_loss": 0.019534897711127996, + "grad_norm": 0.17040184140205383, + "learning_rate": 2.140239418635723e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006568708317354321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029230503248982133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016060466784983872, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6670295000076294, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23500, + "weighted_flow_matching_loss": 0.019534897711127996, + "weighted_rkd_loss": 0.010712659033015371 + }, + { + "epoch": 6.840267675298225, + "flow_matching_loss": 0.02134823389351368, + "grad_norm": 0.16835889220237732, + "learning_rate": 2.1394916772425002e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006595052685588599, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027190987020730972, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01590920416638255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6667827010154724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23510, + "weighted_flow_matching_loss": 0.02134823389351368, + "weighted_rkd_loss": 0.010608033649623394 + }, + { + "epoch": 6.843177189409369, + "flow_matching_loss": 0.018366421479731797, + "grad_norm": 0.18254095315933228, + "learning_rate": 2.138743741588591e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007435356685891748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032020231941714884, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01807273644953966, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6665358781814575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23520, + "weighted_flow_matching_loss": 0.018366421479731797, + "weighted_rkd_loss": 0.012046016007661819 + }, + { + "epoch": 6.846086703520512, + "flow_matching_loss": 0.021093393489718438, + "grad_norm": 0.16278579831123352, + "learning_rate": 2.137995611901198e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007320722751319409, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032193001825362444, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017860745638608934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6662889897823334, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23530, + "weighted_flow_matching_loss": 0.021093393489718438, + "weighted_rkd_loss": 0.011900280322879552 + }, + { + "epoch": 6.8489962176316554, + "flow_matching_loss": 0.01909986212849617, + "grad_norm": 0.1898978054523468, + "learning_rate": 2.1372472884075835e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006882554898038506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031075459672138097, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016872655879706145, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6660420715808868, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23540, + "weighted_flow_matching_loss": 0.01909986212849617, + "weighted_rkd_loss": 0.0112378372810781 + }, + { + "epoch": 6.851905731742799, + "flow_matching_loss": 0.019176106713712216, + "grad_norm": 0.17293798923492432, + "learning_rate": 2.136498771335068e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006930849142372608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030906049534678457, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01695230333134532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.66579509973526, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23550, + "weighted_flow_matching_loss": 0.019176106713712216, + "weighted_rkd_loss": 0.011286842357367276 + }, + { + "epoch": 6.854815245853942, + "flow_matching_loss": 0.020883365720510482, + "grad_norm": 0.15818876028060913, + "learning_rate": 2.1357500609110324e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006847101030871272, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030044260201975703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016698628105223177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6655480861663818, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23560, + "weighted_flow_matching_loss": 0.020883365720510482, + "weighted_rkd_loss": 0.011113842483609915 + }, + { + "epoch": 6.857724759965086, + "flow_matching_loss": 0.019116059876978397, + "grad_norm": 0.18502439558506012, + "learning_rate": 2.1350011573629134e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00652860626578331, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002678025187924504, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01573523776605725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6653010249137878, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23570, + "weighted_flow_matching_loss": 0.019116059876978397, + "weighted_rkd_loss": 0.01046868897974491 + }, + { + "epoch": 6.860634274076229, + "flow_matching_loss": 0.018703957833349705, + "grad_norm": 0.15845631062984467, + "learning_rate": 2.1342520609182093e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006385138910263777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027613078709691763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015531585831195117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.665053915977478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23580, + "weighted_flow_matching_loss": 0.018703957833349705, + "weighted_rkd_loss": 0.010329327546060085 + }, + { + "epoch": 6.863543788187373, + "flow_matching_loss": 0.01970922378823161, + "grad_norm": 0.1879756599664688, + "learning_rate": 2.1335027718044753e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00738401017151773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031906398944556713, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017958660051226615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6648067653179168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23590, + "weighted_flow_matching_loss": 0.01970922378823161, + "weighted_rkd_loss": 0.011939094401896 + }, + { + "epoch": 6.866453302298516, + "flow_matching_loss": 0.01984830675646663, + "grad_norm": 0.22779153287410736, + "learning_rate": 2.1327532902493263e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006777146877720952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002826612256467342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016380906011909245, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6645595729351044, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23600, + "weighted_flow_matching_loss": 0.01984830675646663, + "weighted_rkd_loss": 0.010886008106172084 + }, + { + "epoch": 6.86936281640966, + "flow_matching_loss": 0.020927591249346733, + "grad_norm": 0.1712048202753067, + "learning_rate": 2.1320036164804345e-05, + "loss": 0.0327, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0073297138325870035, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031217446317896246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017781172413378954, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.664312332868576, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23610, + "weighted_flow_matching_loss": 0.020927591249346733, + "weighted_rkd_loss": 0.011812319234013557 + }, + { + "epoch": 6.872272330520803, + "flow_matching_loss": 0.01901209345087409, + "grad_norm": 0.15589891374111176, + "learning_rate": 2.131253750725531e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00667303828522563, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031361818313598634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016482258681207895, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6640650451183319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23620, + "weighted_flow_matching_loss": 0.01901209345087409, + "weighted_rkd_loss": 0.010945295915007591 + }, + { + "epoch": 6.875181844631946, + "flow_matching_loss": 0.02012834418565035, + "grad_norm": 0.2427712380886078, + "learning_rate": 2.1305036932124053e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006923940544947982, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029416464967653155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016789527516812085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6638177335262299, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23630, + "weighted_flow_matching_loss": 0.02012834418565035, + "weighted_rkd_loss": 0.011145040579140186 + }, + { + "epoch": 6.87809135874309, + "flow_matching_loss": 0.018609693646430968, + "grad_norm": 0.15720084309577942, + "learning_rate": 2.1297534441689054e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00607067821547389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025250890757888556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01466644536703825, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6635703384876251, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23640, + "weighted_flow_matching_loss": 0.018609693646430968, + "weighted_rkd_loss": 0.009732188191264868 + }, + { + "epoch": 6.881000872854234, + "flow_matching_loss": 0.02137002404779196, + "grad_norm": 0.18468832969665527, + "learning_rate": 2.1290030038229365e-05, + "loss": 0.0329, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007167564146220684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030444453470408915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017379573732614517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.663322925567627, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23650, + "weighted_flow_matching_loss": 0.02137002404779196, + "weighted_rkd_loss": 0.01152830310165882 + }, + { + "epoch": 6.883910386965377, + "flow_matching_loss": 0.01966956667602062, + "grad_norm": 0.1646897941827774, + "learning_rate": 2.1282523724024632e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006393974041566252, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002673195069655776, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015461143385618926, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6630754590034484, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23660, + "weighted_flow_matching_loss": 0.01966956667602062, + "weighted_rkd_loss": 0.0102519235573709 + }, + { + "epoch": 6.88681990107652, + "flow_matching_loss": 0.019131278339773416, + "grad_norm": 0.1822231411933899, + "learning_rate": 2.127501550135507e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007579302275553345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032649093540385366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018423513788729907, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6628279447555542, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23670, + "weighted_flow_matching_loss": 0.019131278339773416, + "weighted_rkd_loss": 0.012211558781564236 + }, + { + "epoch": 6.889729415187664, + "flow_matching_loss": 0.02020262535661459, + "grad_norm": 0.19355887174606323, + "learning_rate": 2.126750537250148e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007046111952513456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002979893796145916, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017072117421776056, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6625803887844086, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23680, + "weighted_flow_matching_loss": 0.02020262535661459, + "weighted_rkd_loss": 0.011311700288206338 + }, + { + "epoch": 6.892638929298807, + "flow_matching_loss": 0.019940916169434785, + "grad_norm": 0.17355047166347504, + "learning_rate": 2.1259993339745244e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00650642947293818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027913994155824183, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01580425864085555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6623327910900116, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23690, + "weighted_flow_matching_loss": 0.019940916169434785, + "weighted_rkd_loss": 0.01046762363985181 + }, + { + "epoch": 6.89554844340995, + "flow_matching_loss": 0.01987108550965786, + "grad_norm": 0.19490955770015717, + "learning_rate": 2.1252479405368314e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007599848136305809, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003364360029809177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018564056418836117, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6620851457118988, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23700, + "weighted_flow_matching_loss": 0.01987108550965786, + "weighted_rkd_loss": 0.012290850654244423 + }, + { + "epoch": 6.898457957521094, + "flow_matching_loss": 0.01877723094075918, + "grad_norm": 0.17595767974853516, + "learning_rate": 2.1244963571653232e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00689369305036962, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003129343152977526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01691672932356596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6618374586105347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23710, + "weighted_flow_matching_loss": 0.01877723094075918, + "weighted_rkd_loss": 0.011196123715490103 + }, + { + "epoch": 6.901367471632238, + "flow_matching_loss": 0.020878207683563233, + "grad_norm": 0.14374017715454102, + "learning_rate": 2.1237445840883103e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068248953204602, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003176527703180909, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016826318018138407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6615897297859192, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23720, + "weighted_flow_matching_loss": 0.020878207683563233, + "weighted_rkd_loss": 0.011132165882736444 + }, + { + "epoch": 6.904276985743381, + "flow_matching_loss": 0.019500110298395157, + "grad_norm": 0.15030714869499207, + "learning_rate": 2.1229926215341616e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007072944473475218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003087543463334441, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017233432456851005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6613419532775879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23730, + "weighted_flow_matching_loss": 0.019500110298395157, + "weighted_rkd_loss": 0.011397275701165199 + }, + { + "epoch": 6.907186499854524, + "flow_matching_loss": 0.021202181372791527, + "grad_norm": 0.22704441845417023, + "learning_rate": 2.1222404697313036e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006933363294228912, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029449967434629796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016811723355203867, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6610941290855408, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23740, + "weighted_flow_matching_loss": 0.021202181372791527, + "weighted_rkd_loss": 0.011114062555134297 + }, + { + "epoch": 6.910096013965668, + "flow_matching_loss": 0.020668233931064605, + "grad_norm": 0.17625467479228973, + "learning_rate": 2.1214881289082205e-05, + "loss": 0.0308, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006225724657997489, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028909680899232627, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015342417359352111, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6608462631702423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23750, + "weighted_flow_matching_loss": 0.020668233931064605, + "weighted_rkd_loss": 0.01013898621313274 + }, + { + "epoch": 6.913005528076811, + "flow_matching_loss": 0.021356114652007818, + "grad_norm": 0.17262409627437592, + "learning_rate": 2.1207355992934535e-05, + "loss": 0.0323, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006774238683283329, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002969847177155316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016518324613571167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6605983376502991, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23760, + "weighted_flow_matching_loss": 0.021356114652007818, + "weighted_rkd_loss": 0.010911989025771618 + }, + { + "epoch": 6.9159150421879545, + "flow_matching_loss": 0.01999296434223652, + "grad_norm": 0.17833566665649414, + "learning_rate": 2.119982881115601e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007275443058460951, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003159738378599286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017710624355822803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6603503823280334, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23770, + "weighted_flow_matching_loss": 0.01999296434223652, + "weighted_rkd_loss": 0.011695114057511091 + }, + { + "epoch": 6.918824556299098, + "flow_matching_loss": 0.020850694738328457, + "grad_norm": 0.1733839362859726, + "learning_rate": 2.1192299746033187e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006397132482379675, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027635985519737006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01555786356329918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6601024031639099, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23780, + "weighted_flow_matching_loss": 0.020850694738328457, + "weighted_rkd_loss": 0.0102697123773396 + }, + { + "epoch": 6.921734070410242, + "flow_matching_loss": 0.021790686622262002, + "grad_norm": 0.15728065371513367, + "learning_rate": 2.1184768799853195e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006922257179394364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031076469109393655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016952161211520433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6598543584346771, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23790, + "weighted_flow_matching_loss": 0.021790686622262002, + "weighted_rkd_loss": 0.011185851646587253 + }, + { + "epoch": 6.924643584521385, + "flow_matching_loss": 0.019570725224912166, + "grad_norm": 0.17450988292694092, + "learning_rate": 2.1177235974903748e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006927005341276526, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030209605349227784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016874970961362125, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6596062779426575, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23800, + "weighted_flow_matching_loss": 0.019570725224912166, + "weighted_rkd_loss": 0.01113081192597747 + }, + { + "epoch": 6.927553098632528, + "flow_matching_loss": 0.019447764195501805, + "grad_norm": 0.17310914397239685, + "learning_rate": 2.116970127347311e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006327124172821641, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029572034254670143, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015611451771110296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.659358149766922, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23810, + "weighted_flow_matching_loss": 0.019447764195501805, + "weighted_rkd_loss": 0.010293541755527259 + }, + { + "epoch": 6.930462612743672, + "flow_matching_loss": 0.019576706737279893, + "grad_norm": 0.15260383486747742, + "learning_rate": 2.1162164697850127e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006195916654542088, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002590016950853169, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01498185032978654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6591099739074707, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23820, + "weighted_flow_matching_loss": 0.019576706737279893, + "weighted_rkd_loss": 0.009874726086854935 + }, + { + "epoch": 6.933372126854815, + "flow_matching_loss": 0.019194467552006246, + "grad_norm": 0.21317258477210999, + "learning_rate": 2.1154626250324206e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006764716934412718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002872750349342823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016402184125036003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6588617563247681, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23830, + "weighted_flow_matching_loss": 0.019194467552006246, + "weighted_rkd_loss": 0.010806651320308447 + }, + { + "epoch": 6.936281640965959, + "flow_matching_loss": 0.02153355609625578, + "grad_norm": 0.2201886922121048, + "learning_rate": 2.1147085933185338e-05, + "loss": 0.0333, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007325005158782006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003288918756879866, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01793892905116081, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6586134970188141, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23840, + "weighted_flow_matching_loss": 0.02153355609625578, + "weighted_rkd_loss": 0.01181465438567102 + }, + { + "epoch": 6.939191155077102, + "flow_matching_loss": 0.018281488586217164, + "grad_norm": 0.16510450839996338, + "learning_rate": 2.113954374872407e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006373049132525921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027789097977802156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015525007899850608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6583651959896087, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23850, + "weighted_flow_matching_loss": 0.018281488586217164, + "weighted_rkd_loss": 0.0102211176417768 + }, + { + "epoch": 6.942100669188245, + "flow_matching_loss": 0.02132126558572054, + "grad_norm": 0.19689366221427917, + "learning_rate": 2.1131999699231513e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007269961154088378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003249346138909459, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0177892685867846, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6581168532371521, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23860, + "weighted_flow_matching_loss": 0.02132126558572054, + "weighted_rkd_loss": 0.011707337480038405 + }, + { + "epoch": 6.945010183299389, + "flow_matching_loss": 0.020252611301839352, + "grad_norm": 0.17994607985019684, + "learning_rate": 2.1124453786999354e-05, + "loss": 0.032, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007262798072770238, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003267565811984241, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017793162167072295, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6578684628009797, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23870, + "weighted_flow_matching_loss": 0.020252611301839352, + "weighted_rkd_loss": 0.011705605871975422 + }, + { + "epoch": 6.947919697410533, + "flow_matching_loss": 0.02036970164626837, + "grad_norm": 0.15683592855930328, + "learning_rate": 2.111690601431984e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006909663230180741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029701212421059608, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016789447609335185, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6576200187206268, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23880, + "weighted_flow_matching_loss": 0.02036970164626837, + "weighted_rkd_loss": 0.011041022557765245 + }, + { + "epoch": 6.950829211521675, + "flow_matching_loss": 0.019199433270841837, + "grad_norm": 0.188261941075325, + "learning_rate": 2.110935638348579e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006747224554419518, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030128915328532456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01650734068825841, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6573715507984161, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23890, + "weighted_flow_matching_loss": 0.019199433270841837, + "weighted_rkd_loss": 0.010851389355957508 + }, + { + "epoch": 6.953738725632819, + "flow_matching_loss": 0.020535687170922756, + "grad_norm": 0.15894007682800293, + "learning_rate": 2.1101804896790576e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006602808414027095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028290055226534603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016034622583538292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6571230351924896, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23900, + "weighted_flow_matching_loss": 0.020535687170922756, + "weighted_rkd_loss": 0.01053681718185544 + }, + { + "epoch": 6.956648239743963, + "flow_matching_loss": 0.020003476366400718, + "grad_norm": 0.1700890213251114, + "learning_rate": 2.1094251556528143e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063404649496078495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028600708115845918, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015541000571101905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6568744540214538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23910, + "weighted_flow_matching_loss": 0.020003476366400718, + "weighted_rkd_loss": 0.010208542644977569 + }, + { + "epoch": 6.9595577538551066, + "flow_matching_loss": 0.02128913290798664, + "grad_norm": 0.17292669415473938, + "learning_rate": 2.1086696364993e-05, + "loss": 0.0322, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006795988418161869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002967305644415319, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01655928250402212, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6566258609294892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23920, + "weighted_flow_matching_loss": 0.02128913290798664, + "weighted_rkd_loss": 0.010873416811227799 + }, + { + "epoch": 6.962467267966249, + "flow_matching_loss": 0.020172863267362118, + "grad_norm": 0.18385440111160278, + "learning_rate": 2.107913932448021e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0069720680359750984, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002940464019775391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016884599812328815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6563772082328796, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23930, + "weighted_flow_matching_loss": 0.020172863267362118, + "weighted_rkd_loss": 0.01108265370130539 + }, + { + "epoch": 6.965376782077393, + "flow_matching_loss": 0.020474801398813726, + "grad_norm": 0.1865844875574112, + "learning_rate": 2.10715804372854e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006079624779522419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028356082271784543, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014994857553392648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6561285197734833, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23940, + "weighted_flow_matching_loss": 0.020474801398813726, + "weighted_rkd_loss": 0.009838659781962633 + }, + { + "epoch": 6.968286296188537, + "flow_matching_loss": 0.019707621075212955, + "grad_norm": 0.14871975779533386, + "learning_rate": 2.1064019705704767e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006097713112831115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026225100620649753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01481793625280261, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6558797895908356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23950, + "weighted_flow_matching_loss": 0.019707621075212955, + "weighted_rkd_loss": 0.00971878352575004 + }, + { + "epoch": 6.97119581029968, + "flow_matching_loss": 0.01879999879747629, + "grad_norm": 0.17674998939037323, + "learning_rate": 2.1056457132035062e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006830076640471816, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002784404926933348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016444558370858432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.655631011724472, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23960, + "weighted_flow_matching_loss": 0.01879999879747629, + "weighted_rkd_loss": 0.010781476134434342 + }, + { + "epoch": 6.974105324410823, + "flow_matching_loss": 0.01928390460088849, + "grad_norm": 0.1488366723060608, + "learning_rate": 2.1048892718573586e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006283520627766848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026596594601869584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015226700529456138, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6553821861743927, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23970, + "weighted_flow_matching_loss": 0.01928390460088849, + "weighted_rkd_loss": 0.009979348629713058 + }, + { + "epoch": 6.977014838521967, + "flow_matching_loss": 0.01846208730712533, + "grad_norm": 0.16301587224006653, + "learning_rate": 2.1041326467618215e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006455266196280718, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028606079518795013, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015771140251308678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6551333248615265, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23980, + "weighted_flow_matching_loss": 0.01846208730712533, + "weighted_rkd_loss": 0.010332275554537773 + }, + { + "epoch": 6.979924352633111, + "flow_matching_loss": 0.018816633615642786, + "grad_norm": 0.16522304713726044, + "learning_rate": 2.103375838146738e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006221818597987294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026648615021258593, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01510849865153432, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.654884421825409, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 23990, + "weighted_flow_matching_loss": 0.018816633615642786, + "weighted_rkd_loss": 0.009894385328516364 + }, + { + "epoch": 6.9828338667442535, + "flow_matching_loss": 0.020428409986197948, + "grad_norm": 0.1673552393913269, + "learning_rate": 2.1026188462420052e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006736415950581431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030398856848478316, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016512717679142953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6546354770660401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24000, + "weighted_flow_matching_loss": 0.020428409986197948, + "weighted_rkd_loss": 0.010809763846918941 + }, + { + "epoch": 6.985743380855397, + "flow_matching_loss": 0.019076308328658344, + "grad_norm": 0.1554856300354004, + "learning_rate": 2.1018616712775792e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064764510374516245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002788638207130134, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015741540305316448, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6543864846229553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24010, + "weighted_flow_matching_loss": 0.019076308328658344, + "weighted_rkd_loss": 0.010301019623875618 + }, + { + "epoch": 6.988652894966541, + "flow_matching_loss": 0.01873426968231797, + "grad_norm": 0.1660907119512558, + "learning_rate": 2.1011043134834676e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006787238409742713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00288686043350026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01646133717149496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6541374564170838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24020, + "weighted_flow_matching_loss": 0.01873426968231797, + "weighted_rkd_loss": 0.010767928883433342 + }, + { + "epoch": 6.991562409077684, + "flow_matching_loss": 0.019617212004959583, + "grad_norm": 0.2059113085269928, + "learning_rate": 2.1003467730897374e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006681879982352257, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003125609550625086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0164893695153296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6538883805274963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24030, + "weighted_flow_matching_loss": 0.019617212004959583, + "weighted_rkd_loss": 0.010782165173441171 + }, + { + "epoch": 6.9944719231888275, + "flow_matching_loss": 0.01896830480545759, + "grad_norm": 0.16433271765708923, + "learning_rate": 2.0995890503265086e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00672814822755754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003056453261524439, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01651274962350726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6536392569541931, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24040, + "weighted_flow_matching_loss": 0.01896830480545759, + "weighted_rkd_loss": 0.010793435014784336 + }, + { + "epoch": 6.997381437299971, + "flow_matching_loss": 0.019471969921141863, + "grad_norm": 0.17409445345401764, + "learning_rate": 2.0988311454239575e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068925054743886, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003046300634741783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016831311490386726, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.653390109539032, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24050, + "weighted_flow_matching_loss": 0.019471969921141863, + "weighted_rkd_loss": 0.010997381061315536 + }, + { + "epoch": 7.000290951411114, + "flow_matching_loss": 0.020617708563804626, + "grad_norm": 0.19422757625579834, + "learning_rate": 2.098073058612316e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006354573788121343, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025914085446856914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01530055608600378, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6531409025192261, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24060, + "weighted_flow_matching_loss": 0.020617708563804626, + "weighted_rkd_loss": 0.00999347069300711 + }, + { + "epoch": 7.003200465522258, + "flow_matching_loss": 0.018520653247833252, + "grad_norm": 0.15130218863487244, + "learning_rate": 2.0973147901218705e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068687489721924065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030812902841717005, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016818788554519414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6528916597366333, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24070, + "weighted_flow_matching_loss": 0.018520653247833252, + "weighted_rkd_loss": 0.010980826243758202 + }, + { + "epoch": 7.006109979633401, + "flow_matching_loss": 0.019612252619117497, + "grad_norm": 0.1738128513097763, + "learning_rate": 2.0965563401829625e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005924047343432904, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025830274098552763, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014431122224777937, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6526423752307892, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24080, + "weighted_flow_matching_loss": 0.019612252619117497, + "weighted_rkd_loss": 0.009418326988816261 + }, + { + "epoch": 7.009019493744544, + "flow_matching_loss": 0.020248602516949175, + "grad_norm": 0.19021201133728027, + "learning_rate": 2.0957977090259897e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007119173975661397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003244648431427777, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017482996080070735, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6523930549621582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24090, + "weighted_flow_matching_loss": 0.020248602516949175, + "weighted_rkd_loss": 0.011405908735468984 + }, + { + "epoch": 7.011929007855688, + "flow_matching_loss": 0.020273049548268317, + "grad_norm": 0.16769880056381226, + "learning_rate": 2.095038896881404e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006057382328435779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026644258410669862, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014779190439730883, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.652143681049347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24100, + "weighted_flow_matching_loss": 0.020273049548268317, + "weighted_rkd_loss": 0.009638147056102752 + }, + { + "epoch": 7.014838521966832, + "flow_matching_loss": 0.018975222110748292, + "grad_norm": 0.15120501816272736, + "learning_rate": 2.0942799039797128e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006879924656823277, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031036837724968793, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01686353301629424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6518942713737488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24110, + "weighted_flow_matching_loss": 0.018975222110748292, + "weighted_rkd_loss": 0.010993311833590268 + }, + { + "epoch": 7.017748036077975, + "flow_matching_loss": 0.02112442869693041, + "grad_norm": 0.19482354819774628, + "learning_rate": 2.0935207305514777e-05, + "loss": 0.033, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007459389045834541, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0033017259323969484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01822050418704748, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6516448199748993, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24120, + "weighted_flow_matching_loss": 0.02112442869693041, + "weighted_rkd_loss": 0.011873188428580761 + }, + { + "epoch": 7.020657550189118, + "flow_matching_loss": 0.018868250679224728, + "grad_norm": 0.1594027280807495, + "learning_rate": 2.0927613768273156e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062425690703094006, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025645756628364324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015049713663756847, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6513953268527984, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24130, + "weighted_flow_matching_loss": 0.018868250679224728, + "weighted_rkd_loss": 0.009803276974707842 + }, + { + "epoch": 7.023567064300262, + "flow_matching_loss": 0.017933933716267347, + "grad_norm": 0.17452283203601837, + "learning_rate": 2.0920018430378984e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006345262890681625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027235973626375197, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015414123330265284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6511457920074463, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24140, + "weighted_flow_matching_loss": 0.017933933716267347, + "weighted_rkd_loss": 0.01003688359633088 + }, + { + "epoch": 7.026476578411406, + "flow_matching_loss": 0.019261434860527515, + "grad_norm": 0.18930362164974213, + "learning_rate": 2.0912421294139517e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006663290457800031, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002832661778666079, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016159242764115334, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6508962213993073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24150, + "weighted_flow_matching_loss": 0.019261434860527515, + "weighted_rkd_loss": 0.01051794639788568 + }, + { + "epoch": 7.029386092522548, + "flow_matching_loss": 0.019569861702620983, + "grad_norm": 0.17062340676784515, + "learning_rate": 2.0904822361862572e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006609896523877978, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002789158350788057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01600895160809159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6506465971469879, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24160, + "weighted_flow_matching_loss": 0.019569861702620983, + "weighted_rkd_loss": 0.010416293516755104 + }, + { + "epoch": 7.032295606633692, + "flow_matching_loss": 0.018498343974351884, + "grad_norm": 0.17093198001384735, + "learning_rate": 2.0897221635856498e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006678959028795362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028976785484701394, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01625559674575925, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6503969430923462, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24170, + "weighted_flow_matching_loss": 0.018498343974351884, + "weighted_rkd_loss": 0.010572695359587669 + }, + { + "epoch": 7.035205120744836, + "flow_matching_loss": 0.01944584110751748, + "grad_norm": 0.24986478686332703, + "learning_rate": 2.08896191184302e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006441871915012598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027587288292124867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015642472729086875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6501472294330597, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24180, + "weighted_flow_matching_loss": 0.01944584110751748, + "weighted_rkd_loss": 0.010169815830886364 + }, + { + "epoch": 7.038114634855979, + "flow_matching_loss": 0.019593711011111736, + "grad_norm": 0.1745249629020691, + "learning_rate": 2.0882014811893124e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006728818314149976, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029177706339396535, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016375407204031946, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6498974859714508, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24190, + "weighted_flow_matching_loss": 0.019593711011111736, + "weighted_rkd_loss": 0.010642274701967835 + }, + { + "epoch": 7.041024148967122, + "flow_matching_loss": 0.01980693796649575, + "grad_norm": 0.15308959782123566, + "learning_rate": 2.0874408718555254e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006272217445075512, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027208812651224436, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015265316143631935, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.649647718667984, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24200, + "weighted_flow_matching_loss": 0.01980693796649575, + "weighted_rkd_loss": 0.009917089901864528 + }, + { + "epoch": 7.043933663078266, + "flow_matching_loss": 0.020327314641326664, + "grad_norm": 0.13703669607639313, + "learning_rate": 2.0866800840727114e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006435195216909051, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026752337347716094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01554562421515584, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6493978917598724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24210, + "weighted_flow_matching_loss": 0.020327314641326664, + "weighted_rkd_loss": 0.010095346393063665 + }, + { + "epoch": 7.04684317718941, + "flow_matching_loss": 0.019941933266818524, + "grad_norm": 0.17642785608768463, + "learning_rate": 2.0859191180719786e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006368244253098965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028919093078002335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015628397837281226, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.649148029088974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24220, + "weighted_flow_matching_loss": 0.019941933266818524, + "weighted_rkd_loss": 0.010145093128085137 + }, + { + "epoch": 7.049752691300553, + "flow_matching_loss": 0.02039275076240301, + "grad_norm": 0.16060103476047516, + "learning_rate": 2.085157974084488e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006414107047021389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028441987466067075, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01567241260781884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6488981246948242, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24230, + "weighted_flow_matching_loss": 0.02039275076240301, + "weighted_rkd_loss": 0.010169818811118603 + }, + { + "epoch": 7.052662205411696, + "flow_matching_loss": 0.018580000940710308, + "grad_norm": 0.18050207197666168, + "learning_rate": 2.0843966523414553e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006389084411785007, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027184737497009336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015496642608195544, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6486481785774231, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24240, + "weighted_flow_matching_loss": 0.018580000940710308, + "weighted_rkd_loss": 0.010051767947152257 + }, + { + "epoch": 7.05557171952284, + "flow_matching_loss": 0.021922169346362354, + "grad_norm": 0.16731742024421692, + "learning_rate": 2.0836351530741492e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006301462417468429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002756305132061243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0153592299669981, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6483981907367706, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24250, + "weighted_flow_matching_loss": 0.021922169346362354, + "weighted_rkd_loss": 0.009958979906514287 + }, + { + "epoch": 7.058481233633983, + "flow_matching_loss": 0.019866112712770702, + "grad_norm": 0.15811093151569366, + "learning_rate": 2.082873476513894e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006343485601246357, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028173798695206643, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01550435097888112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6481481611728668, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24260, + "weighted_flow_matching_loss": 0.019866112712770702, + "weighted_rkd_loss": 0.01004912327043712 + }, + { + "epoch": 7.0613907477451265, + "flow_matching_loss": 0.020888568088412284, + "grad_norm": 0.13993678987026215, + "learning_rate": 2.0821116228920663e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064971345942467455, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026425563264638186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01563682556152344, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6478980958461762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24270, + "weighted_flow_matching_loss": 0.020888568088412284, + "weighted_rkd_loss": 0.010131119284778833 + }, + { + "epoch": 7.06430026185627, + "flow_matching_loss": 0.02180771566927433, + "grad_norm": 0.23325952887535095, + "learning_rate": 2.0813495924400977e-05, + "loss": 0.0335, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007446371391415596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032357853138819336, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018128527980297803, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6476479887962341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24280, + "weighted_flow_matching_loss": 0.02180771566927433, + "weighted_rkd_loss": 0.01174077307805419 + }, + { + "epoch": 7.067209775967413, + "flow_matching_loss": 0.019346400629729033, + "grad_norm": 0.16570498049259186, + "learning_rate": 2.0805873853894723e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006439625192433596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026803767774254083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015559627115726471, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6473978400230408, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24290, + "weighted_flow_matching_loss": 0.019346400629729033, + "weighted_rkd_loss": 0.010073269158601761 + }, + { + "epoch": 7.070119290078557, + "flow_matching_loss": 0.020173139218240977, + "grad_norm": 0.19172360002994537, + "learning_rate": 2.0798250019717284e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006636105244979262, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002760121808387339, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016032332461327313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6471476495265961, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24300, + "weighted_flow_matching_loss": 0.020173139218240977, + "weighted_rkd_loss": 0.01037521967664361 + }, + { + "epoch": 7.0730288041897005, + "flow_matching_loss": 0.018867839500308035, + "grad_norm": 0.1812107414007187, + "learning_rate": 2.079062442418459e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006533570308238268, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028591692447662354, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01592630986124277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6468974173069, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24310, + "weighted_flow_matching_loss": 0.018867839500308035, + "weighted_rkd_loss": 0.010302740056067706 + }, + { + "epoch": 7.075938318300844, + "flow_matching_loss": 0.0200631333515048, + "grad_norm": 0.21118849515914917, + "learning_rate": 2.078299706961308e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007238899497315288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032824515597894787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01776025043800473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6466471493244171, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24320, + "weighted_flow_matching_loss": 0.0200631333515048, + "weighted_rkd_loss": 0.01148453070782125 + }, + { + "epoch": 7.078847832411987, + "flow_matching_loss": 0.020404080394655466, + "grad_norm": 0.1916942000389099, + "learning_rate": 2.077536795831976e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006432650797069073, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027016432024538515, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01556694470345974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.646396815776825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24330, + "weighted_flow_matching_loss": 0.020404080394655466, + "weighted_rkd_loss": 0.010062333662062883 + }, + { + "epoch": 7.081757346523131, + "flow_matching_loss": 0.018873784597963093, + "grad_norm": 0.1664956659078598, + "learning_rate": 2.0767737092622137e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006076783873140812, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024890877422876654, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014642655663192273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6461464762687683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24340, + "weighted_flow_matching_loss": 0.018873784597963093, + "weighted_rkd_loss": 0.009461297793313861 + }, + { + "epoch": 7.084666860634274, + "flow_matching_loss": 0.020014575868844985, + "grad_norm": 0.14970996975898743, + "learning_rate": 2.076010447483828e-05, + "loss": 0.0318, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007508435193449259, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031881607603281735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018205031286925077, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6458961009979248, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24350, + "weighted_flow_matching_loss": 0.020014575868844985, + "weighted_rkd_loss": 0.011758788116276264 + }, + { + "epoch": 7.087576374745417, + "flow_matching_loss": 0.020303031615912914, + "grad_norm": 0.20331892371177673, + "learning_rate": 2.0752470107286765e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063676406163722275, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002631733822636306, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015367015078663827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6456456661224366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24360, + "weighted_flow_matching_loss": 0.020303031615912914, + "weighted_rkd_loss": 0.00992163149639964 + }, + { + "epoch": 7.090485888856561, + "flow_matching_loss": 0.022658869996666907, + "grad_norm": 0.1701364666223526, + "learning_rate": 2.074483399228672e-05, + "loss": 0.0343, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007386636454612017, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032892473274841903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018062520213425158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6453951954841614, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24370, + "weighted_flow_matching_loss": 0.022658869996666907, + "weighted_rkd_loss": 0.011657452676445246 + }, + { + "epoch": 7.093395402967705, + "flow_matching_loss": 0.018737333361059427, + "grad_norm": 0.16285623610019684, + "learning_rate": 2.0737196132157788e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006089150905609131, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027007216587662695, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014879023469984532, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6451446890830994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24380, + "weighted_flow_matching_loss": 0.018737333361059427, + "weighted_rkd_loss": 0.009599212557077408 + }, + { + "epoch": 7.0963049170788475, + "flow_matching_loss": 0.02007515411823988, + "grad_norm": 0.18755599856376648, + "learning_rate": 2.0729556529220156e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00666031138971448, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002868150197900832, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016188772954046728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6448941349983215, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24390, + "weighted_flow_matching_loss": 0.02007515411823988, + "weighted_rkd_loss": 0.010440020123496652 + }, + { + "epoch": 7.099214431189991, + "flow_matching_loss": 0.01912135733291507, + "grad_norm": 0.15060554444789886, + "learning_rate": 2.0721915185794532e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060322983656078575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025242774747312067, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01458887429907918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6446435451507568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24400, + "weighted_flow_matching_loss": 0.01912135733291507, + "weighted_rkd_loss": 0.009404582716524602 + }, + { + "epoch": 7.102123945301135, + "flow_matching_loss": 0.01914413757622242, + "grad_norm": 0.15686112642288208, + "learning_rate": 2.071427210420215e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006083925114944577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002669951424468309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014837801363319159, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6443929076194763, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24410, + "weighted_flow_matching_loss": 0.01914413757622242, + "weighted_rkd_loss": 0.009561353735625744 + }, + { + "epoch": 7.105033459412278, + "flow_matching_loss": 0.019199175387620927, + "grad_norm": 0.18236057460308075, + "learning_rate": 2.070662728676478e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006253767758607864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002799285657238215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015306821186095476, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6441422522068023, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24420, + "weighted_flow_matching_loss": 0.019199175387620927, + "weighted_rkd_loss": 0.009859785344451666 + }, + { + "epoch": 7.107942973523421, + "flow_matching_loss": 0.018138977140188216, + "grad_norm": 0.23114921152591705, + "learning_rate": 2.0698980735804712e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006222689384594559, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027831395622342826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01522851837798953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6438915312290192, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24430, + "weighted_flow_matching_loss": 0.018138977140188216, + "weighted_rkd_loss": 0.009805408027023076 + }, + { + "epoch": 7.110852487634565, + "flow_matching_loss": 0.019750750437378884, + "grad_norm": 0.17389599978923798, + "learning_rate": 2.0691332453644775e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068741978611797094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003011934459209442, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016760330088436602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6436407923698425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24440, + "weighted_flow_matching_loss": 0.019750750437378884, + "weighted_rkd_loss": 0.01078770887106657 + }, + { + "epoch": 7.113762001745709, + "flow_matching_loss": 0.019969477504491805, + "grad_norm": 0.15177319943904877, + "learning_rate": 2.0683682442608306e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0071058147586882114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031047313008457423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017316360771656037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6433899998664856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24450, + "weighted_flow_matching_loss": 0.019969477504491805, + "weighted_rkd_loss": 0.011141321994364262 + }, + { + "epoch": 7.116671515856852, + "flow_matching_loss": 0.018502840399742128, + "grad_norm": 0.1522241234779358, + "learning_rate": 2.0676030705019182e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006362431915476918, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002630936913192272, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015355800744146109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6431391775608063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24460, + "weighted_flow_matching_loss": 0.018502840399742128, + "weighted_rkd_loss": 0.009876010660082101 + }, + { + "epoch": 7.119581029967995, + "flow_matching_loss": 0.021573592722415925, + "grad_norm": 0.172258198261261, + "learning_rate": 2.066837724320179e-05, + "loss": 0.0342, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.008007658226415516, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003690333245322108, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.019705649837851523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6428883135318756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24470, + "weighted_flow_matching_loss": 0.021573592722415925, + "weighted_rkd_loss": 0.012668540887534618 + }, + { + "epoch": 7.122490544079139, + "flow_matching_loss": 0.022430446185171603, + "grad_norm": 0.17204038798809052, + "learning_rate": 2.0660722059481066e-05, + "loss": 0.0338, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007348801800981164, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003033119230531156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017730722948908807, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6426374018192291, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24480, + "weighted_flow_matching_loss": 0.022430446185171603, + "weighted_rkd_loss": 0.011394331976771355 + }, + { + "epoch": 7.125400058190282, + "flow_matching_loss": 0.01877261195331812, + "grad_norm": 0.17966711521148682, + "learning_rate": 2.065306515618244e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006815978186205029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030067365383729338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016638693027198316, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6423864662647247, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24490, + "weighted_flow_matching_loss": 0.01877261195331812, + "weighted_rkd_loss": 0.010688557848334312 + }, + { + "epoch": 7.128309572301426, + "flow_matching_loss": 0.020127716846764086, + "grad_norm": 0.18267710506916046, + "learning_rate": 2.064540653563188e-05, + "loss": 0.0319, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00749064153060317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003423316846601665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018404599837958814, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.64213547706604, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24500, + "weighted_flow_matching_loss": 0.020127716846764086, + "weighted_rkd_loss": 0.01181824030354619 + }, + { + "epoch": 7.131219086412569, + "flow_matching_loss": 0.020269881561398508, + "grad_norm": 0.20989416539669037, + "learning_rate": 2.063774620015587e-05, + "loss": 0.031, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006898994836956262, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029744474915787576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016772437281906603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6418844521045685, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24510, + "weighted_flow_matching_loss": 0.020269881561398508, + "weighted_rkd_loss": 0.010765888821333646 + }, + { + "epoch": 7.134128600523712, + "flow_matching_loss": 0.018998218420892953, + "grad_norm": 0.15397833287715912, + "learning_rate": 2.0630084152081422e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006168978707864881, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026223955675959585, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014960353169590235, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6416333913803101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24520, + "weighted_flow_matching_loss": 0.018998218420892953, + "weighted_rkd_loss": 0.009599131206050516 + }, + { + "epoch": 7.137038114634856, + "flow_matching_loss": 0.020116801653057338, + "grad_norm": 0.21042904257774353, + "learning_rate": 2.0622420393736065e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006398522621020674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026659373892471195, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015462982561439275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6413822948932648, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24530, + "weighted_flow_matching_loss": 0.020116801653057338, + "weighted_rkd_loss": 0.009917660057544709 + }, + { + "epoch": 7.1399476287459995, + "flow_matching_loss": 0.019473679829388857, + "grad_norm": 0.17147380113601685, + "learning_rate": 2.061475492744784e-05, + "loss": 0.0312, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007509627984836698, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0032579631078988314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018277219124138357, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6411311566829682, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24540, + "weighted_flow_matching_loss": 0.019473679829388857, + "weighted_rkd_loss": 0.011718225292861461 + }, + { + "epoch": 7.142857142857143, + "flow_matching_loss": 0.0204137422144413, + "grad_norm": 0.14900480210781097, + "learning_rate": 2.060708775554532e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005921330908313394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002808431745506823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014651093445718289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6408799767494202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24550, + "weighted_flow_matching_loss": 0.0204137422144413, + "weighted_rkd_loss": 0.009389673918485641 + }, + { + "epoch": 7.145766656968286, + "flow_matching_loss": 0.019811037369072436, + "grad_norm": 0.15850961208343506, + "learning_rate": 2.0599418880357586e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062665292527526615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003030367963947356, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0155634262599051, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6406287550926208, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24560, + "weighted_flow_matching_loss": 0.019811037369072436, + "weighted_rkd_loss": 0.009970476664602757 + }, + { + "epoch": 7.14867617107943, + "flow_matching_loss": 0.020108886156231166, + "grad_norm": 0.15105994045734406, + "learning_rate": 2.059174830421424e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006040268810465932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025374955264851453, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014618033077567816, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6403775095939637, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24570, + "weighted_flow_matching_loss": 0.020108886156231166, + "weighted_rkd_loss": 0.009361144527792931 + }, + { + "epoch": 7.1515856851905735, + "flow_matching_loss": 0.019129762426018715, + "grad_norm": 0.1624314934015274, + "learning_rate": 2.05840760294454e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006717092031612993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029163484694436194, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01635053250938654, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.640126210451126, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24580, + "weighted_flow_matching_loss": 0.019129762426018715, + "weighted_rkd_loss": 0.010466522024944424 + }, + { + "epoch": 7.154495199301716, + "flow_matching_loss": 0.02064306177198887, + "grad_norm": 0.18826156854629517, + "learning_rate": 2.05764020583817e-05, + "loss": 0.0317, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007064604107290507, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031594177009537814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017288625705987215, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6398748815059662, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24590, + "weighted_flow_matching_loss": 0.02064306177198887, + "weighted_rkd_loss": 0.011062483163550496 + }, + { + "epoch": 7.15740471341286, + "flow_matching_loss": 0.019164581038057803, + "grad_norm": 0.19216687977313995, + "learning_rate": 2.056872639335429e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006704406486824155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002838723221793771, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016247536428272723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6396235167980194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24600, + "weighted_flow_matching_loss": 0.019164581038057803, + "weighted_rkd_loss": 0.010392375197261572 + }, + { + "epoch": 7.160314227524004, + "flow_matching_loss": 0.019746967125684023, + "grad_norm": 0.17375408113002777, + "learning_rate": 2.056104903669484e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006389378383755684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002811113803181797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015589870512485504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6393721044063568, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24610, + "weighted_flow_matching_loss": 0.019746967125684023, + "weighted_rkd_loss": 0.009967678412795066 + }, + { + "epoch": 7.1632237416351465, + "flow_matching_loss": 0.019779067486524582, + "grad_norm": 0.20521996915340424, + "learning_rate": 2.055336999073552e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006068853149190545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025556946988217534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014693401195108891, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6391206622123718, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24620, + "weighted_flow_matching_loss": 0.019779067486524582, + "weighted_rkd_loss": 0.00939082526601851 + }, + { + "epoch": 7.16613325574629, + "flow_matching_loss": 0.01773989275097847, + "grad_norm": 0.2156130075454712, + "learning_rate": 2.0545689257809028e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006396783329546451, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028539714869111775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015647538378834723, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6388691663742065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24630, + "weighted_flow_matching_loss": 0.01773989275097847, + "weighted_rkd_loss": 0.009996728086844087 + }, + { + "epoch": 7.169042769857434, + "flow_matching_loss": 0.020951592735946177, + "grad_norm": 0.17615830898284912, + "learning_rate": 2.053800684024856e-05, + "loss": 0.0316, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00683504668995738, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030379546107724307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01670804787427187, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6386176347732544, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24640, + "weighted_flow_matching_loss": 0.020951592735946177, + "weighted_rkd_loss": 0.010670150583609938 + }, + { + "epoch": 7.171952283968578, + "flow_matching_loss": 0.019124712236225604, + "grad_norm": 0.19401417672634125, + "learning_rate": 2.0530322740387842e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007109479978680611, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003201720688957721, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017420680727809668, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6383660972118378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24650, + "weighted_flow_matching_loss": 0.019124712236225604, + "weighted_rkd_loss": 0.011120601138100029 + }, + { + "epoch": 7.1748617980797205, + "flow_matching_loss": 0.018765253014862537, + "grad_norm": 0.15787945687770844, + "learning_rate": 2.0522636960561096e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006906744744628668, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030201702495105563, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016833659913390874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6381144881248474, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24660, + "weighted_flow_matching_loss": 0.018765253014862537, + "weighted_rkd_loss": 0.010741747170686721 + }, + { + "epoch": 7.177771312190864, + "flow_matching_loss": 0.019046188052743673, + "grad_norm": 0.17762520909309387, + "learning_rate": 2.0514949503103056e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064014483243227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002730978699401021, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015533875580877066, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6378628551959992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24670, + "weighted_flow_matching_loss": 0.019046188052743673, + "weighted_rkd_loss": 0.009908405318856239 + }, + { + "epoch": 7.180680826302008, + "flow_matching_loss": 0.02012899164110422, + "grad_norm": 0.1775650680065155, + "learning_rate": 2.0507260370348968e-05, + "loss": 0.0303, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065275007393211125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029162569204345345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015971258375793697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6376111805438995, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24680, + "weighted_flow_matching_loss": 0.02012899164110422, + "weighted_rkd_loss": 0.010183309577405453 + }, + { + "epoch": 7.183590340413151, + "flow_matching_loss": 0.018957509193569422, + "grad_norm": 0.14794494211673737, + "learning_rate": 2.049956956463459e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006081702653318644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002596306789200753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014759712014347314, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6373594701290131, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24690, + "weighted_flow_matching_loss": 0.018957509193569422, + "weighted_rkd_loss": 0.009407252492383122 + }, + { + "epoch": 7.186499854524294, + "flow_matching_loss": 0.018263879232108593, + "grad_norm": 0.1743929237127304, + "learning_rate": 2.049187708829619e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005824005859903991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002551718207541853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014199730101972818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6371077179908753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24700, + "weighted_flow_matching_loss": 0.018263879232108593, + "weighted_rkd_loss": 0.009046730818226933 + }, + { + "epoch": 7.189409368635438, + "flow_matching_loss": 0.019627750478684904, + "grad_norm": 0.17063793540000916, + "learning_rate": 2.0484182943670537e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006795011414214969, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029917534440755842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016581776458770035, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6368559300899506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24710, + "weighted_flow_matching_loss": 0.019627750478684904, + "weighted_rkd_loss": 0.010560206323862075 + }, + { + "epoch": 7.192318882746581, + "flow_matching_loss": 0.019205853529274463, + "grad_norm": 0.17929191887378693, + "learning_rate": 2.04764871330949e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006208152836188674, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027630914468318225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015179397165775299, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6366041004657745, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24720, + "weighted_flow_matching_loss": 0.019205853529274463, + "weighted_rkd_loss": 0.009663253230974079 + }, + { + "epoch": 7.195228396857725, + "flow_matching_loss": 0.01831228584051132, + "grad_norm": 0.14827798306941986, + "learning_rate": 2.0468789658907075e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006751670967787504, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029538385570049286, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016457180306315423, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6363522469997406, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24730, + "weighted_flow_matching_loss": 0.01831228584051132, + "weighted_rkd_loss": 0.01047254605218768 + }, + { + "epoch": 7.198137910968868, + "flow_matching_loss": 0.019592416658997535, + "grad_norm": 0.22180019319057465, + "learning_rate": 2.0461090523445344e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006358128087595105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028057557763531806, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015522011928260327, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6361003518104553, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24740, + "weighted_flow_matching_loss": 0.019592416658997535, + "weighted_rkd_loss": 0.009873522538691759 + }, + { + "epoch": 7.201047425080012, + "flow_matching_loss": 0.021009220462292432, + "grad_norm": 0.1764668971300125, + "learning_rate": 2.04533897290485e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006664525298401713, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028639092808589338, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016192959900945426, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6358484148979187, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24750, + "weighted_flow_matching_loss": 0.021009220462292432, + "weighted_rkd_loss": 0.010296347457915545 + }, + { + "epoch": 7.203956939191155, + "flow_matching_loss": 0.018920392077416182, + "grad_norm": 0.19430875778198242, + "learning_rate": 2.0445687278055852e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061560965608805415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026318499352782966, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014944043010473251, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6355964362621307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24760, + "weighted_flow_matching_loss": 0.018920392077416182, + "weighted_rkd_loss": 0.00949835181236267 + }, + { + "epoch": 7.206866453302299, + "flow_matching_loss": 0.02000249084085226, + "grad_norm": 0.16023385524749756, + "learning_rate": 2.0437983172807196e-05, + "loss": 0.0314, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007417239248752594, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003180967504158616, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.018015445675700903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6353444278240203, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24770, + "weighted_flow_matching_loss": 0.02000249084085226, + "weighted_rkd_loss": 0.01144616287201643 + }, + { + "epoch": 7.209775967413442, + "flow_matching_loss": 0.018868003971874714, + "grad_norm": 0.16912466287612915, + "learning_rate": 2.043027741564283e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006643063528463245, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031195034505799414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016405630484223367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6350923776626587, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24780, + "weighted_flow_matching_loss": 0.018868003971874714, + "weighted_rkd_loss": 0.010419164132326842 + }, + { + "epoch": 7.212685481524585, + "flow_matching_loss": 0.020127305760979652, + "grad_norm": 0.20542554557323456, + "learning_rate": 2.042257000890356e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061602231580764055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002559583308175206, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014880029857158661, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6348402798175812, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24790, + "weighted_flow_matching_loss": 0.020127305760979652, + "weighted_rkd_loss": 0.009446457400918007 + }, + { + "epoch": 7.215594995635729, + "flow_matching_loss": 0.018695538491010667, + "grad_norm": 0.1744772046804428, + "learning_rate": 2.0414860954930698e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006937019526958466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029374010162428023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016811439860612155, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6345881521701813, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24800, + "weighted_flow_matching_loss": 0.018695538491010667, + "weighted_rkd_loss": 0.010668294318020344 + }, + { + "epoch": 7.2185045097468725, + "flow_matching_loss": 0.01859460221603513, + "grad_norm": 0.1935155689716339, + "learning_rate": 2.0407150256066052e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006396743003278971, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002672163804527372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015465649776160717, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6343360126018525, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24810, + "weighted_flow_matching_loss": 0.01859460221603513, + "weighted_rkd_loss": 0.009810423105955124 + }, + { + "epoch": 7.221414023858015, + "flow_matching_loss": 0.01869628606364131, + "grad_norm": 0.2040613740682602, + "learning_rate": 2.0399437914651927e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006389711936935783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002658193768002093, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015437617711722851, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6340838134288788, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24820, + "weighted_flow_matching_loss": 0.01869628606364131, + "weighted_rkd_loss": 0.009788627130910754 + }, + { + "epoch": 7.224323537969159, + "flow_matching_loss": 0.018785938434302808, + "grad_norm": 0.20254433155059814, + "learning_rate": 2.039172393303112e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00665849894285202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003020150726661086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016337148565798997, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6338315844535828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24830, + "weighted_flow_matching_loss": 0.018785938434302808, + "weighted_rkd_loss": 0.010354920662939549 + }, + { + "epoch": 7.227233052080303, + "flow_matching_loss": 0.018980601523071527, + "grad_norm": 0.16886451840400696, + "learning_rate": 2.0384008313546947e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006058432720601558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026489306590519847, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014765795972198247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6335793137550354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24840, + "weighted_flow_matching_loss": 0.018980601523071527, + "weighted_rkd_loss": 0.009355216473340987 + }, + { + "epoch": 7.2301425661914465, + "flow_matching_loss": 0.020158670842647552, + "grad_norm": 0.15971936285495758, + "learning_rate": 2.0376291058543206e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00631789998151362, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028067894978448748, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01544258939102292, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6333270072937012, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24850, + "weighted_flow_matching_loss": 0.020158670842647552, + "weighted_rkd_loss": 0.009780210861936211 + }, + { + "epoch": 7.233052080302589, + "flow_matching_loss": 0.01898397644981742, + "grad_norm": 0.14108797907829285, + "learning_rate": 2.036857217036419e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00645353221334517, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027870456455275417, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01569411000236869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6330746591091156, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24860, + "weighted_flow_matching_loss": 0.01898397644981742, + "weighted_rkd_loss": 0.009935671556740999 + }, + { + "epoch": 7.235961594413733, + "flow_matching_loss": 0.01946585513651371, + "grad_norm": 0.17092207074165344, + "learning_rate": 2.0360851651354702e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006164686009287834, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002708262298256159, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01503763422369957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6328222870826721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24870, + "weighted_flow_matching_loss": 0.01946585513651371, + "weighted_rkd_loss": 0.009516129549592734 + }, + { + "epoch": 7.238871108524877, + "flow_matching_loss": 0.021141277626156807, + "grad_norm": 0.17657652497291565, + "learning_rate": 2.0353129503860025e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005976015608757735, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00271806251257658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01467009363695979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6325698733329773, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24880, + "weighted_flow_matching_loss": 0.021141277626156807, + "weighted_rkd_loss": 0.00927987489849329 + }, + { + "epoch": 7.2417806226360195, + "flow_matching_loss": 0.019562962371855973, + "grad_norm": 0.15404947102069855, + "learning_rate": 2.034540573022595e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006358333397656679, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027600569068454207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015476723853498697, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6323174178600312, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24890, + "weighted_flow_matching_loss": 0.019562962371855973, + "weighted_rkd_loss": 0.009786200989037753 + }, + { + "epoch": 7.244690136747163, + "flow_matching_loss": 0.020036750100553037, + "grad_norm": 0.1598680317401886, + "learning_rate": 2.0337680332798742e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006490254588425159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002845985069870949, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01582649452611804, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6320649266242981, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24900, + "weighted_flow_matching_loss": 0.020036750100553037, + "weighted_rkd_loss": 0.010003390070050956 + }, + { + "epoch": 7.247599650858307, + "flow_matching_loss": 0.019271176867187025, + "grad_norm": 0.16686566174030304, + "learning_rate": 2.0329953313925188e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063757164869457485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026557889534160493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01540722157806158, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6318123996257782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24910, + "weighted_flow_matching_loss": 0.019271176867187025, + "weighted_rkd_loss": 0.009734569024294615 + }, + { + "epoch": 7.25050916496945, + "flow_matching_loss": 0.018131049629300833, + "grad_norm": 0.14967337250709534, + "learning_rate": 2.032222467595255e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006566854752600193, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028505032416433094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01598421288654208, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6315598487854004, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24920, + "weighted_flow_matching_loss": 0.018131049629300833, + "weighted_rkd_loss": 0.010095053073018789 + }, + { + "epoch": 7.2534186790805935, + "flow_matching_loss": 0.020771260000765325, + "grad_norm": 0.1757412701845169, + "learning_rate": 2.0314494421228577e-05, + "loss": 0.0305, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006387245282530785, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026085575809702277, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015383047983050346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6313072443008423, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24930, + "weighted_flow_matching_loss": 0.020771260000765325, + "weighted_rkd_loss": 0.00971145867370069 + }, + { + "epoch": 7.256328193191737, + "flow_matching_loss": 0.019730310700833798, + "grad_norm": 0.17370158433914185, + "learning_rate": 2.0306762552101516e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005734839430078864, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023785923840478063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01384827122092247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6310546159744262, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24940, + "weighted_flow_matching_loss": 0.019730310700833798, + "weighted_rkd_loss": 0.00873909592628479 + }, + { + "epoch": 7.25923770730288, + "flow_matching_loss": 0.01868783850222826, + "grad_norm": 0.17742346227169037, + "learning_rate": 2.0299029070920113e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006359230121597647, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029238136950880287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01564227370545268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6308019459247589, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24950, + "weighted_flow_matching_loss": 0.01868783850222826, + "weighted_rkd_loss": 0.009867230569943786 + }, + { + "epoch": 7.262147221414024, + "flow_matching_loss": 0.017317189183086158, + "grad_norm": 0.18417952954769135, + "learning_rate": 2.0291293980033594e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065031684003770355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002729543496388942, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01573588028550148, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6305492520332336, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24960, + "weighted_flow_matching_loss": 0.017317189183086158, + "weighted_rkd_loss": 0.009922173153609038 + }, + { + "epoch": 7.265056735525167, + "flow_matching_loss": 0.018630532547831535, + "grad_norm": 0.18443433940410614, + "learning_rate": 2.028355728179167e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006166994757950306, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002562027936801314, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014896017592400312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.630296504497528, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24970, + "weighted_flow_matching_loss": 0.018630532547831535, + "weighted_rkd_loss": 0.009388929512351752 + }, + { + "epoch": 7.267966249636311, + "flow_matching_loss": 0.019894724525511265, + "grad_norm": 0.15896400809288025, + "learning_rate": 2.0275818978544546e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005790172796696425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025627626688219606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014143108297139407, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6300437211990356, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24980, + "weighted_flow_matching_loss": 0.019894724525511265, + "weighted_rkd_loss": 0.008910825615748763 + }, + { + "epoch": 7.270875763747454, + "flow_matching_loss": 0.01891034943982959, + "grad_norm": 0.17244011163711548, + "learning_rate": 2.026807907264292e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006795011227950454, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029035426210612058, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016493565030395986, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6297909140586853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 24990, + "weighted_flow_matching_loss": 0.01891034943982959, + "weighted_rkd_loss": 0.010387460375204683 + }, + { + "epoch": 7.273785277858598, + "flow_matching_loss": 0.02010029386729002, + "grad_norm": 0.1804649829864502, + "learning_rate": 2.026033756643796e-05, + "loss": 0.0304, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006681558070704341, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002953835646621883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01631695181131363, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6295380651950836, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25000, + "weighted_flow_matching_loss": 0.02010029386729002, + "weighted_rkd_loss": 0.010272148251533508 + }, + { + "epoch": 7.276694791969741, + "flow_matching_loss": 0.01960023557767272, + "grad_norm": 0.17085114121437073, + "learning_rate": 2.0252594462281347e-05, + "loss": 0.0301, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006811773683875799, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003135395632125437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01675894297659397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6292851805686951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25010, + "weighted_flow_matching_loss": 0.01960023557767272, + "weighted_rkd_loss": 0.010546125005930662 + }, + { + "epoch": 7.279604306080884, + "flow_matching_loss": 0.019792513363063334, + "grad_norm": 0.18001128733158112, + "learning_rate": 2.024484976252522e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006189152738079429, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026036232244223355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014981928654015065, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6290322661399841, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25020, + "weighted_flow_matching_loss": 0.019792513363063334, + "weighted_rkd_loss": 0.009424093877896667 + }, + { + "epoch": 7.282513820192028, + "flow_matching_loss": 0.018992950581014155, + "grad_norm": 0.15625955164432526, + "learning_rate": 2.0237103469522212e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007063339976593852, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030792790930718184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017205958999693394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6287793099880219, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25030, + "weighted_flow_matching_loss": 0.018992950581014155, + "weighted_rkd_loss": 0.010818724706768989 + }, + { + "epoch": 7.285423334303172, + "flow_matching_loss": 0.020691151823848487, + "grad_norm": 0.17507719993591309, + "learning_rate": 2.022935558562545e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005816996889188885, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002565934881567955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014199928473681211, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6285263299942017, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25040, + "weighted_flow_matching_loss": 0.020691151823848487, + "weighted_rkd_loss": 0.008924992941319942 + }, + { + "epoch": 7.288332848414315, + "flow_matching_loss": 0.018797586858272552, + "grad_norm": 0.15350645780563354, + "learning_rate": 2.0221606113188538e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00563355372287333, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024409713805653155, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013708078674972058, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6282733023166657, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25050, + "weighted_flow_matching_loss": 0.018797586858272552, + "weighted_rkd_loss": 0.00861251107417047 + }, + { + "epoch": 7.291242362525458, + "flow_matching_loss": 0.019160189758986236, + "grad_norm": 0.13059915602207184, + "learning_rate": 2.0213855054565554e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065121558960527185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002766475826501846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015790787618607284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6280202388763427, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25060, + "weighted_flow_matching_loss": 0.019160189758986236, + "weighted_rkd_loss": 0.009917040402069689 + }, + { + "epoch": 7.294151876636602, + "flow_matching_loss": 0.02062683692201972, + "grad_norm": 0.16133204102516174, + "learning_rate": 2.020610241211107e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006524525070562959, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002966795628890395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01601584581658244, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6277671456336975, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25070, + "weighted_flow_matching_loss": 0.02062683692201972, + "weighted_rkd_loss": 0.010054253228008746 + }, + { + "epoch": 7.2970613907477455, + "flow_matching_loss": 0.018989817053079606, + "grad_norm": 0.16964131593704224, + "learning_rate": 2.0198348188180137e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007346750330179929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031748808454722167, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017868381273001433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6275140166282653, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25080, + "weighted_flow_matching_loss": 0.018989817053079606, + "weighted_rkd_loss": 0.01121281748637557 + }, + { + "epoch": 7.299970904858888, + "flow_matching_loss": 0.018650331161916257, + "grad_norm": 0.1478884220123291, + "learning_rate": 2.0190592385128274e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006333474162966013, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027161910897120834, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015383139438927173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6272608458995819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25090, + "weighted_flow_matching_loss": 0.018650331161916257, + "weighted_rkd_loss": 0.009649296896532178 + }, + { + "epoch": 7.302880418970032, + "flow_matching_loss": 0.01708229389041662, + "grad_norm": 0.13618671894073486, + "learning_rate": 2.018283500531149e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0072116201743483545, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003396193590015173, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017819433752447367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6270076513290406, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25100, + "weighted_flow_matching_loss": 0.01708229389041662, + "weighted_rkd_loss": 0.011173038883134723 + }, + { + "epoch": 7.305789933081176, + "flow_matching_loss": 0.019659112393856048, + "grad_norm": 0.18136478960514069, + "learning_rate": 2.0175076051086282e-05, + "loss": 0.03, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006854658992961049, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028418223257176576, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01655114032328129, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6267544150352478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25110, + "weighted_flow_matching_loss": 0.019659112393856048, + "weighted_rkd_loss": 0.010373415984213352 + }, + { + "epoch": 7.308699447192319, + "flow_matching_loss": 0.019412625674158336, + "grad_norm": 0.1646254062652588, + "learning_rate": 2.016731552480961e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006023307237774134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002430209331214428, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014476823806762695, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6265011548995971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25120, + "weighted_flow_matching_loss": 0.019412625674158336, + "weighted_rkd_loss": 0.009069784078747034 + }, + { + "epoch": 7.311608961303462, + "flow_matching_loss": 0.017887889593839645, + "grad_norm": 0.18404661118984222, + "learning_rate": 2.0159553428838914e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064280132297426466, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002709749806672335, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015565776359289884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6262478530406952, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25130, + "weighted_flow_matching_loss": 0.017887889593839645, + "weighted_rkd_loss": 0.009748022304847836 + }, + { + "epoch": 7.314518475414606, + "flow_matching_loss": 0.01894193170592189, + "grad_norm": 0.17514853179454803, + "learning_rate": 2.0151789765532117e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006639926647767425, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028785979375243187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01615845113992691, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6259945094585418, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25140, + "weighted_flow_matching_loss": 0.01894193170592189, + "weighted_rkd_loss": 0.01011502486653626 + }, + { + "epoch": 7.317427989525749, + "flow_matching_loss": 0.018264447432011367, + "grad_norm": 0.18513086438179016, + "learning_rate": 2.0144024537247612e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006329141464084387, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002748787240125239, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015407070238143205, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6257411360740661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25150, + "weighted_flow_matching_loss": 0.018264447432011367, + "weighted_rkd_loss": 0.00964084775187075 + }, + { + "epoch": 7.3203375036368925, + "flow_matching_loss": 0.019082209561020136, + "grad_norm": 0.16883864998817444, + "learning_rate": 2.013625774634428e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062673629727214575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028207373339682817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015355463325977325, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6254877269268035, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25160, + "weighted_flow_matching_loss": 0.019082209561020136, + "weighted_rkd_loss": 0.0096045286860317 + }, + { + "epoch": 7.323247017748036, + "flow_matching_loss": 0.018639357294887304, + "grad_norm": 0.15738141536712646, + "learning_rate": 2.0128489395181455e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005854713870212436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002364168781787157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01407359642907977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6252342879772186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25170, + "weighted_flow_matching_loss": 0.018639357294887304, + "weighted_rkd_loss": 0.008799284789711237 + }, + { + "epoch": 7.32615653185918, + "flow_matching_loss": 0.020101531781256197, + "grad_norm": 0.15034978091716766, + "learning_rate": 2.0120719486118958e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006690730294212699, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028283220599405468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016209782846271992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6249808132648468, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25180, + "weighted_flow_matching_loss": 0.020101531781256197, + "weighted_rkd_loss": 0.010130725242197514 + }, + { + "epoch": 7.329066045970323, + "flow_matching_loss": 0.018002257402986287, + "grad_norm": 0.12739554047584534, + "learning_rate": 2.0112948021517093e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005745949409902096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028161954251118004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014308094140142202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6247273027896881, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25190, + "weighted_flow_matching_loss": 0.018002257402986287, + "weighted_rkd_loss": 0.008938806690275669 + }, + { + "epoch": 7.3319755600814664, + "flow_matching_loss": 0.021385280787944792, + "grad_norm": 0.15757563710212708, + "learning_rate": 2.0105175003736615e-05, + "loss": 0.0313, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065138097386807205, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002814309054519981, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01584192821756005, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6244737565517425, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25200, + "weighted_flow_matching_loss": 0.021385280787944792, + "weighted_rkd_loss": 0.00989298396743834 + }, + { + "epoch": 7.33488507419261, + "flow_matching_loss": 0.02055684309452772, + "grad_norm": 0.17117632925510406, + "learning_rate": 2.0097400435138764e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059000262524932625, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025255305226892233, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014325582794845105, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6242201805114747, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25210, + "weighted_flow_matching_loss": 0.02055684309452772, + "weighted_rkd_loss": 0.008942395076155662 + }, + { + "epoch": 7.337794588303753, + "flow_matching_loss": 0.017148298677057028, + "grad_norm": 0.13929623365402222, + "learning_rate": 2.0089624318085253e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056750202551484105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023210238432511687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013671064283698797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6239665746688843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25220, + "weighted_flow_matching_loss": 0.017148298677057028, + "weighted_rkd_loss": 0.008530368469655514 + }, + { + "epoch": 7.340704102414897, + "flow_matching_loss": 0.01821196610108018, + "grad_norm": 0.15091471374034882, + "learning_rate": 2.0081846654938256e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006398300221189856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027976322919130324, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01559423254802823, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6237129151821137, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25230, + "weighted_flow_matching_loss": 0.01821196610108018, + "weighted_rkd_loss": 0.009726380929350853 + }, + { + "epoch": 7.34361361652604, + "flow_matching_loss": 0.01807507062330842, + "grad_norm": 0.14844052493572235, + "learning_rate": 2.0074067448060423e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006070362217724323, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026204169495031237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014761141408234835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6234592437744141, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25240, + "weighted_flow_matching_loss": 0.01807507062330842, + "weighted_rkd_loss": 0.009203080972656607 + }, + { + "epoch": 7.346523130637183, + "flow_matching_loss": 0.018505018670111895, + "grad_norm": 0.1429368108510971, + "learning_rate": 2.0066286699814873e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059452835470438, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002430736436508596, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014321303460747004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6232055246829986, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25250, + "weighted_flow_matching_loss": 0.018505018670111895, + "weighted_rkd_loss": 0.00892515517771244 + }, + { + "epoch": 7.349432644748327, + "flow_matching_loss": 0.01737221172079444, + "grad_norm": 0.161385178565979, + "learning_rate": 2.005850441256519e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005616822838783264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002394971204921603, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01362861655652523, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6229517757892609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25260, + "weighted_flow_matching_loss": 0.01737221172079444, + "weighted_rkd_loss": 0.008489967603236438 + }, + { + "epoch": 7.352342158859471, + "flow_matching_loss": 0.01937236124649644, + "grad_norm": 0.15170913934707642, + "learning_rate": 2.0050720588675432e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060851153451949355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024628126760944725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01463304329663515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6226979911327362, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25270, + "weighted_flow_matching_loss": 0.01937236124649644, + "weighted_rkd_loss": 0.009111955109983683 + }, + { + "epoch": 7.355251672970613, + "flow_matching_loss": 0.020343449898064137, + "grad_norm": 0.14897111058235168, + "learning_rate": 2.0042935230510113e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058347830083221195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026026745676063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014272240549325943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6224441826343536, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25280, + "weighted_flow_matching_loss": 0.020343449898064137, + "weighted_rkd_loss": 0.008883758401498198 + }, + { + "epoch": 7.358161187081757, + "flow_matching_loss": 0.01981536587700248, + "grad_norm": 0.16113540530204773, + "learning_rate": 2.0035148340434222e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00555301527492702, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002425606397446245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013531637098640203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6221903324127197, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25290, + "weighted_flow_matching_loss": 0.01981536587700248, + "weighted_rkd_loss": 0.008419251209124923 + }, + { + "epoch": 7.361070701192901, + "flow_matching_loss": 0.018727642577141525, + "grad_norm": 0.1517467051744461, + "learning_rate": 2.0027359920813212e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006092217192053795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002667878451757133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01485231276601553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6219364404678345, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25300, + "weighted_flow_matching_loss": 0.018727642577141525, + "weighted_rkd_loss": 0.009237138368189336 + }, + { + "epoch": 7.363980215304045, + "flow_matching_loss": 0.019170534797012806, + "grad_norm": 0.1663498878479004, + "learning_rate": 2.0019569974012994e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006041538994759321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024847027962096035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014567780774086714, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6216825246810913, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25310, + "weighted_flow_matching_loss": 0.019170534797012806, + "weighted_rkd_loss": 0.009056406933814288 + }, + { + "epoch": 7.366889729415187, + "flow_matching_loss": 0.017987347114831208, + "grad_norm": 0.1820872724056244, + "learning_rate": 2.0011778502399954e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006183764291927218, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025129078654572367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014880436658859252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6214285850524902, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25320, + "weighted_flow_matching_loss": 0.017987347114831208, + "weighted_rkd_loss": 0.009247179608792067 + }, + { + "epoch": 7.369799243526331, + "flow_matching_loss": 0.019716139510273935, + "grad_norm": 0.17335563898086548, + "learning_rate": 2.0003985508340937e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005830153916031122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002377677196636796, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014037985168397427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6211746037006378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25330, + "weighted_flow_matching_loss": 0.019716139510273935, + "weighted_rkd_loss": 0.00871994737535715 + }, + { + "epoch": 7.372708757637475, + "flow_matching_loss": 0.018814904522150754, + "grad_norm": 0.15494121611118317, + "learning_rate": 1.9996190994203244e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006432091537863016, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002705928171053529, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015570111293345689, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.620920580625534, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25340, + "weighted_flow_matching_loss": 0.018814904522150754, + "weighted_rkd_loss": 0.00966790532693267 + }, + { + "epoch": 7.375618271748618, + "flow_matching_loss": 0.01974297408014536, + "grad_norm": 0.1849089115858078, + "learning_rate": 1.998839496235464e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006611572904512286, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002801901800557971, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01602504765614867, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6206665337085724, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25350, + "weighted_flow_matching_loss": 0.01974297408014536, + "weighted_rkd_loss": 0.00994613696821034 + }, + { + "epoch": 7.378527785859761, + "flow_matching_loss": 0.021409381832927465, + "grad_norm": 0.16611050069332123, + "learning_rate": 1.9980597415163358e-05, + "loss": 0.0311, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063680749852210285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00280710831284523, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015543258283287287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6204124569892884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25360, + "weighted_flow_matching_loss": 0.021409381832927465, + "weighted_rkd_loss": 0.009643191564828157 + }, + { + "epoch": 7.381437299970905, + "flow_matching_loss": 0.021172606293112038, + "grad_norm": 0.1412234604358673, + "learning_rate": 1.997279835499809e-05, + "loss": 0.0309, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006502057006582618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027254460612311958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015729560051113367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.620158338546753, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25370, + "weighted_flow_matching_loss": 0.021172606293112038, + "weighted_rkd_loss": 0.009754959447309374 + }, + { + "epoch": 7.384346814082049, + "flow_matching_loss": 0.018270062748342752, + "grad_norm": 0.16457772254943848, + "learning_rate": 1.9964997784227976e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005912353098392486, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002476732130162418, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014301438350230455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6199041903018951, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25380, + "weighted_flow_matching_loss": 0.018270062748342752, + "weighted_rkd_loss": 0.00886548226699233 + }, + { + "epoch": 7.387256328193192, + "flow_matching_loss": 0.018779121804982424, + "grad_norm": 0.19459567964076996, + "learning_rate": 1.9957195705222633e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006202266039326787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026282742619514465, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015032806433737278, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6196500062942505, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25390, + "weighted_flow_matching_loss": 0.018779121804982424, + "weighted_rkd_loss": 0.009315016912296414 + }, + { + "epoch": 7.390165842304335, + "flow_matching_loss": 0.01944208089262247, + "grad_norm": 0.1676364243030548, + "learning_rate": 1.994939212035212e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006174209481105209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002724615577608347, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01507303435355425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6193957924842834, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25400, + "weighted_flow_matching_loss": 0.01944208089262247, + "weighted_rkd_loss": 0.009336108108982444 + }, + { + "epoch": 7.393075356415479, + "flow_matching_loss": 0.018394377268850803, + "grad_norm": 0.1391160935163498, + "learning_rate": 1.9941587031986963e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005852261371910572, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002514155418612063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014218678046017886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.619141548871994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25410, + "weighted_flow_matching_loss": 0.018394377268850803, + "weighted_rkd_loss": 0.008803465031087398 + }, + { + "epoch": 7.395984870526622, + "flow_matching_loss": 0.01824201913550496, + "grad_norm": 0.1736932098865509, + "learning_rate": 1.9933780442498137e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065448879264295105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002862606733106077, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015952382516115905, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6188872694969177, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25420, + "weighted_flow_matching_loss": 0.01824201913550496, + "weighted_rkd_loss": 0.00987274176441133 + }, + { + "epoch": 7.3988943846377655, + "flow_matching_loss": 0.021435052435845137, + "grad_norm": 0.1698068529367447, + "learning_rate": 1.9925972354257085e-05, + "loss": 0.0315, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006641842983663082, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030027990695089104, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016286485083401202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6186329603195191, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25430, + "weighted_flow_matching_loss": 0.021435052435845137, + "weighted_rkd_loss": 0.010075269686058163 + }, + { + "epoch": 7.401803898748909, + "flow_matching_loss": 0.019343104120343922, + "grad_norm": 0.17452165484428406, + "learning_rate": 1.9918162769635696e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006809367192909121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002883804007433355, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01650253850966692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.618378609418869, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25440, + "weighted_flow_matching_loss": 0.019343104120343922, + "weighted_rkd_loss": 0.01020495593547821 + }, + { + "epoch": 7.404713412860052, + "flow_matching_loss": 0.018212045542895794, + "grad_norm": 0.15978218615055084, + "learning_rate": 1.9910351691006308e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006654586410149932, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028963617514818907, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016205534525215625, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6181242346763611, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25450, + "weighted_flow_matching_loss": 0.018212045542895794, + "weighted_rkd_loss": 0.01001688945107162 + }, + { + "epoch": 7.407622926971196, + "flow_matching_loss": 0.019222764950245618, + "grad_norm": 0.1446511149406433, + "learning_rate": 1.990253912074173e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006371470866724849, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027783986181020735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015521340258419513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6178698301315307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25460, + "weighted_flow_matching_loss": 0.019222764950245618, + "weighted_rkd_loss": 0.00959015330299735 + }, + { + "epoch": 7.410532441082339, + "flow_matching_loss": 0.019771741796284915, + "grad_norm": 0.14462462067604065, + "learning_rate": 1.989472506121521e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056242936756461855, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023229250917211177, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013571512419730425, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6176153898239136, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25470, + "weighted_flow_matching_loss": 0.019771741796284915, + "weighted_rkd_loss": 0.00838192249648273 + }, + { + "epoch": 7.413441955193482, + "flow_matching_loss": 0.016932308953255416, + "grad_norm": 0.15195755660533905, + "learning_rate": 1.9886909514800454e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006046896474435926, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002630817494355142, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0147246103733778, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.617360919713974, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25480, + "weighted_flow_matching_loss": 0.016932308953255416, + "weighted_rkd_loss": 0.009090432012453676 + }, + { + "epoch": 7.416351469304626, + "flow_matching_loss": 0.021041833236813544, + "grad_norm": 0.1693124622106552, + "learning_rate": 1.9879092483871622e-05, + "loss": 0.0307, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006391727598384023, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002921407762914896, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015704863145947455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6171064078807831, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25490, + "weighted_flow_matching_loss": 0.021041833236813544, + "weighted_rkd_loss": 0.009691677102819086 + }, + { + "epoch": 7.41926098341577, + "flow_matching_loss": 0.020067875925451516, + "grad_norm": 0.18142829835414886, + "learning_rate": 1.987127397080332e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006539442343637347, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002798846561927348, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015877731423825026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6168518722057342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25500, + "weighted_flow_matching_loss": 0.020067875925451516, + "weighted_rkd_loss": 0.009794233134016395 + }, + { + "epoch": 7.422170497526913, + "flow_matching_loss": 0.017392253316938878, + "grad_norm": 0.17539741098880768, + "learning_rate": 1.98634539779706e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005702262558043003, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026374227134510876, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0140419477596879, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6165973007678985, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25510, + "weighted_flow_matching_loss": 0.017392253316938878, + "weighted_rkd_loss": 0.008658181736245751 + }, + { + "epoch": 7.425080011638056, + "flow_matching_loss": 0.020197810232639314, + "grad_norm": 0.17096824944019318, + "learning_rate": 1.985563250774897e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0068817419931292536, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003086746553890407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016850230656564234, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6163426995277405, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25520, + "weighted_flow_matching_loss": 0.020197810232639314, + "weighted_rkd_loss": 0.010385640524327755 + }, + { + "epoch": 7.4279895257492, + "flow_matching_loss": 0.01903786836192012, + "grad_norm": 0.1887311488389969, + "learning_rate": 1.98478095625144e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064210712909698485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028855090262368322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01572765177115798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6160880625247955, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25530, + "weighted_flow_matching_loss": 0.01903786836192012, + "weighted_rkd_loss": 0.009689571196213364 + }, + { + "epoch": 7.430899039860344, + "flow_matching_loss": 0.018213766720145942, + "grad_norm": 0.1862039417028427, + "learning_rate": 1.9839985144643275e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00603619716130197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002499581058509648, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014571975357830525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6158333957195282, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25540, + "weighted_flow_matching_loss": 0.018213766720145942, + "weighted_rkd_loss": 0.008973779808729886 + }, + { + "epoch": 7.433808553971486, + "flow_matching_loss": 0.020345381554216148, + "grad_norm": 0.16469867527484894, + "learning_rate": 1.9832159256512463e-05, + "loss": 0.0293, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005966267175972461, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002542236726731062, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014474770985543728, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6155787110328674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25550, + "weighted_flow_matching_loss": 0.020345381554216148, + "weighted_rkd_loss": 0.008910260302945972 + }, + { + "epoch": 7.43671806808263, + "flow_matching_loss": 0.02012297883629799, + "grad_norm": 0.15592297911643982, + "learning_rate": 1.982433190049925e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005537895113229751, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022303837235085667, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013306173961609602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6153239727020263, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25560, + "weighted_flow_matching_loss": 0.02012297883629799, + "weighted_rkd_loss": 0.008187636313959957 + }, + { + "epoch": 7.439627582193774, + "flow_matching_loss": 0.018418794590979816, + "grad_norm": 0.16256040334701538, + "learning_rate": 1.981650307898138e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005837384751066566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024236002354882658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014098369609564542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6150692164897918, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25570, + "weighted_flow_matching_loss": 0.018418794590979816, + "weighted_rkd_loss": 0.008671551616862417 + }, + { + "epoch": 7.442537096304917, + "flow_matching_loss": 0.020185638032853604, + "grad_norm": 0.17616765201091766, + "learning_rate": 1.980867279433705e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006065196171402931, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002544402878265828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01467479532584548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.614814430475235, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25580, + "weighted_flow_matching_loss": 0.020185638032853604, + "weighted_rkd_loss": 0.009022287977859377 + }, + { + "epoch": 7.44544661041606, + "flow_matching_loss": 0.01917604310438037, + "grad_norm": 0.18002194166183472, + "learning_rate": 1.98008410489449e-05, + "loss": 0.0287, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006289359228685498, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028958902577869592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015474608726799488, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6145595967769623, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25590, + "weighted_flow_matching_loss": 0.01917604310438037, + "weighted_rkd_loss": 0.009509976068511606 + }, + { + "epoch": 7.448356124527204, + "flow_matching_loss": 0.018577981553971768, + "grad_norm": 0.16360723972320557, + "learning_rate": 1.9793007845183987e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006534382095560431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027549428399652244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015823707077652217, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6143047451972962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25600, + "weighted_flow_matching_loss": 0.018577981553971768, + "weighted_rkd_loss": 0.009720646776258945 + }, + { + "epoch": 7.451265638638348, + "flow_matching_loss": 0.01839731205254793, + "grad_norm": 0.19915850460529327, + "learning_rate": 1.9785173185433846e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00587208210490644, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027215218404307963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014465686026960611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6140498578548431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25610, + "weighted_flow_matching_loss": 0.01839731205254793, + "weighted_rkd_loss": 0.008882646262645722 + }, + { + "epoch": 7.454175152749491, + "flow_matching_loss": 0.019133557751774787, + "grad_norm": 0.1972091943025589, + "learning_rate": 1.977733707207444e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005983687983825803, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026656302623450758, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014633006416261196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6137949407100678, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25620, + "weighted_flow_matching_loss": 0.019133557751774787, + "weighted_rkd_loss": 0.008981598168611526 + }, + { + "epoch": 7.457084666860634, + "flow_matching_loss": 0.01805643131956458, + "grad_norm": 0.16123642027378082, + "learning_rate": 1.976949950748616e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061295194085687395, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025926445610821246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014851683471351861, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.61353999376297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25630, + "weighted_flow_matching_loss": 0.01805643131956458, + "weighted_rkd_loss": 0.009112062770873309 + }, + { + "epoch": 7.459994180971778, + "flow_matching_loss": 0.018457640428096056, + "grad_norm": 0.16558139026165009, + "learning_rate": 1.976166049404987e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006636540265753865, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029097928665578364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016182873211801053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6132850110530853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25640, + "weighted_flow_matching_loss": 0.018457640428096056, + "weighted_rkd_loss": 0.009924642136320471 + }, + { + "epoch": 7.462903695082921, + "flow_matching_loss": 0.01856429800391197, + "grad_norm": 0.1595182567834854, + "learning_rate": 1.9753820034146848e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006403571786358952, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027140827616676686, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015521226357668639, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6130299985408783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25650, + "weighted_flow_matching_loss": 0.01856429800391197, + "weighted_rkd_loss": 0.00951507557183504 + }, + { + "epoch": 7.4658132091940645, + "flow_matching_loss": 0.02015570057556033, + "grad_norm": 0.1755800098180771, + "learning_rate": 1.9745978130158813e-05, + "loss": 0.0297, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006412415113300085, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002795975445769727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01562080578878522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6127749562263489, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25660, + "weighted_flow_matching_loss": 0.02015570057556033, + "weighted_rkd_loss": 0.009572062874212861 + }, + { + "epoch": 7.468722723305208, + "flow_matching_loss": 0.017675520107150077, + "grad_norm": 0.12497859448194504, + "learning_rate": 1.9738134784467932e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006315891304984689, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002672799548599869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0153045822866261, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6125198662281036, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25670, + "weighted_flow_matching_loss": 0.017675520107150077, + "weighted_rkd_loss": 0.009374397993087768 + }, + { + "epoch": 7.471632237416351, + "flow_matching_loss": 0.018658605590462684, + "grad_norm": 0.1491982787847519, + "learning_rate": 1.9730289999456804e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0070275743026286365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029789617750793696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017034110240638257, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6122647762298584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25680, + "weighted_flow_matching_loss": 0.018658605590462684, + "weighted_rkd_loss": 0.010429603233933448 + }, + { + "epoch": 7.474541751527495, + "flow_matching_loss": 0.019293418247252704, + "grad_norm": 0.14237208664417267, + "learning_rate": 1.9722443777508473e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005617692647501826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002313413773663342, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013548799138516188, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6120096445083618, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25690, + "weighted_flow_matching_loss": 0.019293418247252704, + "weighted_rkd_loss": 0.008292106539011001 + }, + { + "epoch": 7.4774512656386385, + "flow_matching_loss": 0.018948030844330787, + "grad_norm": 0.1405995488166809, + "learning_rate": 1.9714596121006414e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00638593859039247, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024862880120053886, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01525816498324275, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6117544770240784, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25700, + "weighted_flow_matching_loss": 0.018948030844330787, + "weighted_rkd_loss": 0.009334307396784425 + }, + { + "epoch": 7.480360779749782, + "flow_matching_loss": 0.019219513423740862, + "grad_norm": 0.1603570133447647, + "learning_rate": 1.9706747032334535e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006156811956316233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027426005457527936, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01505622435361147, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.611499285697937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25710, + "weighted_flow_matching_loss": 0.019219513423740862, + "weighted_rkd_loss": 0.009206880861893297 + }, + { + "epoch": 7.483270293860925, + "flow_matching_loss": 0.01820267690345645, + "grad_norm": 0.19729262590408325, + "learning_rate": 1.9698896513877177e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005872502271085977, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023814972955733538, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014126501884311438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6112440586090088, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25720, + "weighted_flow_matching_loss": 0.01820267690345645, + "weighted_rkd_loss": 0.008634723210707306 + }, + { + "epoch": 7.486179807972069, + "flow_matching_loss": 0.01882317420095205, + "grad_norm": 0.1507948637008667, + "learning_rate": 1.9691044568019128e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006588261714205146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027685826877132057, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015945106092840434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6109888017177582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25730, + "weighted_flow_matching_loss": 0.01882317420095205, + "weighted_rkd_loss": 0.009742230968549847 + }, + { + "epoch": 7.489089322083212, + "flow_matching_loss": 0.018744513392448425, + "grad_norm": 0.18759089708328247, + "learning_rate": 1.96831911971456e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005588087206706405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024965506861917676, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013672724924981594, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6107335209846496, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25740, + "weighted_flow_matching_loss": 0.018744513392448425, + "weighted_rkd_loss": 0.008350250776857138 + }, + { + "epoch": 7.491998836194355, + "flow_matching_loss": 0.01840162416920066, + "grad_norm": 0.17844931781291962, + "learning_rate": 1.9675336403642234e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.007073025824502111, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0030495191225782038, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.017195570934563875, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6104781925678253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25750, + "weighted_flow_matching_loss": 0.01840162416920066, + "weighted_rkd_loss": 0.01049741506576538 + }, + { + "epoch": 7.494908350305499, + "flow_matching_loss": 0.019680945202708244, + "grad_norm": 0.15946097671985626, + "learning_rate": 1.966748018989512e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005746534047648311, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025002541719004513, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013993322383612394, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6102228462696075, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25760, + "weighted_flow_matching_loss": 0.019680945202708244, + "weighted_rkd_loss": 0.008539065532386304 + }, + { + "epoch": 7.497817864416643, + "flow_matching_loss": 0.01861228058114648, + "grad_norm": 0.18719421327114105, + "learning_rate": 1.965962255829076e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067184594925493, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002930063859093934, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01636698320508003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6099674701690674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25770, + "weighted_flow_matching_loss": 0.01861228058114648, + "weighted_rkd_loss": 0.009983322257176042 + }, + { + "epoch": 7.5007273785277855, + "flow_matching_loss": 0.020907349418848753, + "grad_norm": 0.21203413605690002, + "learning_rate": 1.9651763511216096e-05, + "loss": 0.0294, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005787129839882254, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023793199798092246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01395357958972454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6097120642662048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25780, + "weighted_flow_matching_loss": 0.020907349418848753, + "weighted_rkd_loss": 0.008507660310715436 + }, + { + "epoch": 7.503636892638929, + "flow_matching_loss": 0.02066908832639456, + "grad_norm": 0.1845691055059433, + "learning_rate": 1.96439030510585e-05, + "loss": 0.0299, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006328221503645181, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024178055464290082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015074248611927032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6094566226005554, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25790, + "weighted_flow_matching_loss": 0.02066908832639456, + "weighted_rkd_loss": 0.009187122946605087 + }, + { + "epoch": 7.506546406750073, + "flow_matching_loss": 0.018124494515359402, + "grad_norm": 0.20110099017620087, + "learning_rate": 1.9636041180205774e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057278202380985025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025264084921218454, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013982048816978931, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6092011511325837, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25800, + "weighted_flow_matching_loss": 0.018124494515359402, + "weighted_rkd_loss": 0.008517868956550956 + }, + { + "epoch": 7.509455920861216, + "flow_matching_loss": 0.020207790285348894, + "grad_norm": 0.1604485809803009, + "learning_rate": 1.9628177901046144e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006323313969187438, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002761109103448689, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015407737158238888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6089456498622894, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25810, + "weighted_flow_matching_loss": 0.020207790285348894, + "weighted_rkd_loss": 0.009382520383223892 + }, + { + "epoch": 7.512365434972359, + "flow_matching_loss": 0.017733958084136248, + "grad_norm": 0.15266011655330658, + "learning_rate": 1.962031321596827e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006128425803035497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025270701153203845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014783921744674444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6086901247501373, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25820, + "weighted_flow_matching_loss": 0.017733958084136248, + "weighted_rkd_loss": 0.008998833363875747 + }, + { + "epoch": 7.515274949083503, + "flow_matching_loss": 0.020286211185157298, + "grad_norm": 0.18985489010810852, + "learning_rate": 1.961244712736123e-05, + "loss": 0.0298, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065192586276680235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026448442018590867, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015683361049741506, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6084345698356628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25830, + "weighted_flow_matching_loss": 0.020286211185157298, + "weighted_rkd_loss": 0.009542348561808467 + }, + { + "epoch": 7.518184463194647, + "flow_matching_loss": 0.019840582739561796, + "grad_norm": 0.19718176126480103, + "learning_rate": 1.9604579637614535e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006044694455340505, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026219942141324283, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01471138307824731, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6081789791584015, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25840, + "weighted_flow_matching_loss": 0.019840582739561796, + "weighted_rkd_loss": 0.008947068778797983 + }, + { + "epoch": 7.52109397730579, + "flow_matching_loss": 0.018166476394981146, + "grad_norm": 0.17760616540908813, + "learning_rate": 1.9596710749118126e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006143278535455465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00269417364615947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014980730693787336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6079233646392822, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25850, + "weighted_flow_matching_loss": 0.018166476394981146, + "weighted_rkd_loss": 0.009107160149142146 + }, + { + "epoch": 7.524003491416933, + "flow_matching_loss": 0.018150623142719268, + "grad_norm": 0.20290498435497284, + "learning_rate": 1.9588840464262364e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006893552746623755, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003151148138567805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0169382537715137, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6076677143573761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25860, + "weighted_flow_matching_loss": 0.018150623142719268, + "weighted_rkd_loss": 0.010292772576212883 + }, + { + "epoch": 7.526913005528077, + "flow_matching_loss": 0.01807677550241351, + "grad_norm": 0.15051738917827606, + "learning_rate": 1.9580968785438026e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006048728479072452, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002765264711342752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01486272132024169, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6074120402336121, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25870, + "weighted_flow_matching_loss": 0.01807677550241351, + "weighted_rkd_loss": 0.009027802618220448 + }, + { + "epoch": 7.52982251963922, + "flow_matching_loss": 0.018092304840683936, + "grad_norm": 0.17732764780521393, + "learning_rate": 1.9573095715036324e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006158828129991889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002705158106982708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015022814180701971, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6071563303470612, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25880, + "weighted_flow_matching_loss": 0.018092304840683936, + "weighted_rkd_loss": 0.00912107778713107 + }, + { + "epoch": 7.532732033750364, + "flow_matching_loss": 0.0189851937815547, + "grad_norm": 0.16037267446517944, + "learning_rate": 1.9565221255448892e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061522744130343195, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026022182777523995, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014906767010688781, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6069006025791168, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25890, + "weighted_flow_matching_loss": 0.0189851937815547, + "weighted_rkd_loss": 0.009046884253621101 + }, + { + "epoch": 7.535641547861507, + "flow_matching_loss": 0.017606256995350122, + "grad_norm": 0.16753314435482025, + "learning_rate": 1.9557345409067775e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062583211809396746, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002689418429508805, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015206060744822026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6066448271274567, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25900, + "weighted_flow_matching_loss": 0.017606256995350122, + "weighted_rkd_loss": 0.009224715689197183 + }, + { + "epoch": 7.538551061972651, + "flow_matching_loss": 0.01811903193593025, + "grad_norm": 0.15869878232479095, + "learning_rate": 1.954946817828546e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006479910016059876, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029474995099008083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015907319635152815, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6063890397548676, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25910, + "weighted_flow_matching_loss": 0.01811903193593025, + "weighted_rkd_loss": 0.009645929466933012 + }, + { + "epoch": 7.541460576083794, + "flow_matching_loss": 0.01878797085955739, + "grad_norm": 0.17703478038311005, + "learning_rate": 1.954158956549483e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060529538430273535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027209475869312883, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014826855156570673, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6061332166194916, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25920, + "weighted_flow_matching_loss": 0.01878797085955739, + "weighted_rkd_loss": 0.008986950479447841 + }, + { + "epoch": 7.5443700901949375, + "flow_matching_loss": 0.016825034376233818, + "grad_norm": 0.14507819712162018, + "learning_rate": 1.9533709573089204e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067022575065493585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029125350527465345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016317049972712994, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6058773696422577, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25930, + "weighted_flow_matching_loss": 0.016825034376233818, + "weighted_rkd_loss": 0.00988619215786457 + }, + { + "epoch": 7.547279604306081, + "flow_matching_loss": 0.018099791184067727, + "grad_norm": 0.16128501296043396, + "learning_rate": 1.9525828203462313e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005559716979041696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002338156150653958, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013457590155303479, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.605621474981308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25940, + "weighted_flow_matching_loss": 0.018099791184067727, + "weighted_rkd_loss": 0.008150128787383438 + }, + { + "epoch": 7.550189118417224, + "flow_matching_loss": 0.01943221837282181, + "grad_norm": 0.14344607293605804, + "learning_rate": 1.9517945459008318e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005655809864401817, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024501346051692963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013761754520237446, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6053655743598938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25950, + "weighted_flow_matching_loss": 0.01943221837282181, + "weighted_rkd_loss": 0.008330966299399733 + }, + { + "epoch": 7.553098632528368, + "flow_matching_loss": 0.018426321540027858, + "grad_norm": 0.13676534593105316, + "learning_rate": 1.9510061342121776e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0062589906621724365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027789194486103953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015296900644898415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6051096320152283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25960, + "weighted_flow_matching_loss": 0.018426321540027858, + "weighted_rkd_loss": 0.009256369387730955 + }, + { + "epoch": 7.5560081466395115, + "flow_matching_loss": 0.019424332492053508, + "grad_norm": 0.15095949172973633, + "learning_rate": 1.950217585519768e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006484969891607762, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026099953101947904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015579935256391763, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6048536598682404, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25970, + "weighted_flow_matching_loss": 0.019424332492053508, + "weighted_rkd_loss": 0.009423581976443529 + }, + { + "epoch": 7.558917660750654, + "flow_matching_loss": 0.01681311447173357, + "grad_norm": 0.1508292555809021, + "learning_rate": 1.9494289000631437e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0066366433165967464, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028900396078824996, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016163326147943736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.60459765791893, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25980, + "weighted_flow_matching_loss": 0.01681311447173357, + "weighted_rkd_loss": 0.009772320091724396 + }, + { + "epoch": 7.561827174861798, + "flow_matching_loss": 0.019945411756634713, + "grad_norm": 0.12998515367507935, + "learning_rate": 1.9486400780818857e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061889241449534895, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002687618718482554, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015065467078238725, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6043416440486908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 25990, + "weighted_flow_matching_loss": 0.019945411756634713, + "weighted_rkd_loss": 0.009104749513790011 + }, + { + "epoch": 7.564736688972942, + "flow_matching_loss": 0.019059528410434724, + "grad_norm": 0.18097734451293945, + "learning_rate": 1.9478511198156175e-05, + "loss": 0.0291, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006781788822263479, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029796739108860493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01654325155541301, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6040855884552002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26000, + "weighted_flow_matching_loss": 0.019059528410434724, + "weighted_rkd_loss": 0.009993558330461384 + }, + { + "epoch": 7.5676462030840845, + "flow_matching_loss": 0.0175659766420722, + "grad_norm": 0.15477880835533142, + "learning_rate": 1.9470620255040043e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006891329120844602, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0029337863670662045, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01671644449234009, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6038295090198517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26010, + "weighted_flow_matching_loss": 0.0175659766420722, + "weighted_rkd_loss": 0.010093893017619848 + }, + { + "epoch": 7.570555717195228, + "flow_matching_loss": 0.01912190029397607, + "grad_norm": 0.15804028511047363, + "learning_rate": 1.9462727953867515e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005484541039913892, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023237226181663574, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01329280473291874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6035733997821808, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26020, + "weighted_flow_matching_loss": 0.01912190029397607, + "weighted_rkd_loss": 0.008023216435685755 + }, + { + "epoch": 7.573465231306372, + "flow_matching_loss": 0.018345714081078768, + "grad_norm": 0.15977828204631805, + "learning_rate": 1.9454834297036066e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00566400783136487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023497703135944904, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013677786011248827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6033172607421875, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26030, + "weighted_flow_matching_loss": 0.018345714081078768, + "weighted_rkd_loss": 0.008252135617658496 + }, + { + "epoch": 7.576374745417516, + "flow_matching_loss": 0.018146778084337713, + "grad_norm": 0.16320866346359253, + "learning_rate": 1.944693928694358e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006416786462068558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002920318208634853, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015753890946507455, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6030610859394073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26040, + "weighted_flow_matching_loss": 0.018146778084337713, + "weighted_rkd_loss": 0.009500538185238838 + }, + { + "epoch": 7.5792842595286585, + "flow_matching_loss": 0.016333207301795483, + "grad_norm": 0.176224023103714, + "learning_rate": 1.943904292598835e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006370560359209776, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002715684799477458, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015456805471330881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6028048932552338, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26050, + "weighted_flow_matching_loss": 0.016333207301795483, + "weighted_rkd_loss": 0.0093173089902848 + }, + { + "epoch": 7.582193773639802, + "flow_matching_loss": 0.018799323122948408, + "grad_norm": 0.16461822390556335, + "learning_rate": 1.9431145216569093e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006233695475384593, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002695640514139086, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015163031406700611, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6025486707687377, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26060, + "weighted_flow_matching_loss": 0.018799323122948408, + "weighted_rkd_loss": 0.009136512130498885 + }, + { + "epoch": 7.585103287750946, + "flow_matching_loss": 0.018894466664642097, + "grad_norm": 0.17866221070289612, + "learning_rate": 1.942324616108491e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006532631954178214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026876442367210983, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015752908308058977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6022924184799194, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26070, + "weighted_flow_matching_loss": 0.018894466664642097, + "weighted_rkd_loss": 0.009487852593883872 + }, + { + "epoch": 7.588012801862089, + "flow_matching_loss": 0.01949778515845537, + "grad_norm": 0.17747139930725098, + "learning_rate": 1.9415345761935342e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005459730001166463, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022501465515233575, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013169606495648623, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6020361185073853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26080, + "weighted_flow_matching_loss": 0.01949778515845537, + "weighted_rkd_loss": 0.007928636018186807 + }, + { + "epoch": 7.590922315973232, + "flow_matching_loss": 0.01877340106293559, + "grad_norm": 0.13862407207489014, + "learning_rate": 1.9407444021520302e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00611268999055028, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025681252009235322, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014793505147099494, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6017798185348511, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26090, + "weighted_flow_matching_loss": 0.01877340106293559, + "weighted_rkd_loss": 0.00890233856625855 + }, + { + "epoch": 7.593831830084376, + "flow_matching_loss": 0.01736952383071184, + "grad_norm": 0.14985324442386627, + "learning_rate": 1.9399540942240144e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061991569120436905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002580378367565572, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014978692308068276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.601523494720459, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26100, + "weighted_flow_matching_loss": 0.01736952383071184, + "weighted_rkd_loss": 0.0090100581292063 + }, + { + "epoch": 7.59674134419552, + "flow_matching_loss": 0.017716426961123945, + "grad_norm": 0.15186476707458496, + "learning_rate": 1.9391636526495613e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006342053879052401, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002763839822728187, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015447947662323713, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6012671291828156, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26110, + "weighted_flow_matching_loss": 0.017716426961123945, + "weighted_rkd_loss": 0.00928834592923522 + }, + { + "epoch": 7.599650858306663, + "flow_matching_loss": 0.018869967479258776, + "grad_norm": 0.17669247090816498, + "learning_rate": 1.9383730776687858e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005612073792144656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024222190957516434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013646366726607084, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6010107338428498, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26120, + "weighted_flow_matching_loss": 0.018869967479258776, + "weighted_rkd_loss": 0.008201572159305216 + }, + { + "epoch": 7.602560372417806, + "flow_matching_loss": 0.017788988538086414, + "grad_norm": 0.20076440274715424, + "learning_rate": 1.9375823695218437e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006247450364753604, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002841830789111555, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015336731355637312, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.600754314661026, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26130, + "weighted_flow_matching_loss": 0.017788988538086414, + "weighted_rkd_loss": 0.009213501727208495 + }, + { + "epoch": 7.605469886528949, + "flow_matching_loss": 0.01811247868463397, + "grad_norm": 0.14798933267593384, + "learning_rate": 1.9367915284489316e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005960528738796711, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002477688994258642, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014398746564984322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6004978716373444, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26140, + "weighted_flow_matching_loss": 0.01811247868463397, + "weighted_rkd_loss": 0.008646342996507883 + }, + { + "epoch": 7.608379400640093, + "flow_matching_loss": 0.01844437588006258, + "grad_norm": 0.15403634309768677, + "learning_rate": 1.9360005546902856e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006060458812862634, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027107425848953426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014831660129129887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.6002413928508759, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26150, + "weighted_flow_matching_loss": 0.01844437588006258, + "weighted_rkd_loss": 0.00890259356237948 + }, + { + "epoch": 7.611288914751237, + "flow_matching_loss": 0.020374423172324897, + "grad_norm": 0.166426882147789, + "learning_rate": 1.9352094484861836e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005335917975753546, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002257397084031254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012929232884198427, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5999848902225494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26160, + "weighted_flow_matching_loss": 0.020374423172324897, + "weighted_rkd_loss": 0.007757270988076925 + }, + { + "epoch": 7.61419842886238, + "flow_matching_loss": 0.019495410658419133, + "grad_norm": 0.15223635733127594, + "learning_rate": 1.934418210076942e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005355844413861633, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002247951948083937, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012959640845656395, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5997283637523652, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26170, + "weighted_flow_matching_loss": 0.019495410658419133, + "weighted_rkd_loss": 0.007772335084155202 + }, + { + "epoch": 7.617107942973523, + "flow_matching_loss": 0.019779003225266933, + "grad_norm": 0.14762641489505768, + "learning_rate": 1.9336268397029176e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006344897393137217, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026523027569055556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015342097729444504, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5994718074798584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26180, + "weighted_flow_matching_loss": 0.019779003225266933, + "weighted_rkd_loss": 0.00919732116162777 + }, + { + "epoch": 7.620017457084667, + "flow_matching_loss": 0.018045196402817966, + "grad_norm": 0.1596168577671051, + "learning_rate": 1.9328353376045096e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006123089278116822, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026260192273184655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014872197806835175, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5992152094841003, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26190, + "weighted_flow_matching_loss": 0.018045196402817966, + "weighted_rkd_loss": 0.008911729464307428 + }, + { + "epoch": 7.6229269711958105, + "flow_matching_loss": 0.01965146539732814, + "grad_norm": 0.16463381052017212, + "learning_rate": 1.932043704022154e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005385087803006172, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023578973836265505, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013128073047846555, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5989586055278778, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26200, + "weighted_flow_matching_loss": 0.01965146539732814, + "weighted_rkd_loss": 0.007863173307850958 + }, + { + "epoch": 7.625836485306953, + "flow_matching_loss": 0.017190129868686198, + "grad_norm": 0.15574470162391663, + "learning_rate": 1.9312519391963284e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00570313292555511, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025022972491569816, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01390856308862567, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5987019717693329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26210, + "weighted_flow_matching_loss": 0.017190129868686198, + "weighted_rkd_loss": 0.008327090926468373 + }, + { + "epoch": 7.628745999418097, + "flow_matching_loss": 0.019209830183535816, + "grad_norm": 0.17957060039043427, + "learning_rate": 1.9304600433675504e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00605374698061496, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024930524406954644, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014600546564906836, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5984453082084655, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26220, + "weighted_flow_matching_loss": 0.019209830183535816, + "weighted_rkd_loss": 0.00873757041990757 + }, + { + "epoch": 7.631655513529241, + "flow_matching_loss": 0.018374269548803567, + "grad_norm": 0.194804847240448, + "learning_rate": 1.929668016776377e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006689327163621783, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002737521647941321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016116175893694164, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5981886148452759, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26230, + "weighted_flow_matching_loss": 0.018374269548803567, + "weighted_rkd_loss": 0.009640508005395532 + }, + { + "epoch": 7.6345650276403845, + "flow_matching_loss": 0.01778389047831297, + "grad_norm": 0.14938387274742126, + "learning_rate": 1.9288758596634046e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005861528776586055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024476039106957614, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01417066128924489, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5979318916797638, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26240, + "weighted_flow_matching_loss": 0.01778389047831297, + "weighted_rkd_loss": 0.00847322870977223 + }, + { + "epoch": 7.637474541751527, + "flow_matching_loss": 0.01828001169487834, + "grad_norm": 0.13897787034511566, + "learning_rate": 1.9280835722692698e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005986814014613628, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025848780758678915, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014558506198227406, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5976751446723938, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26250, + "weighted_flow_matching_loss": 0.01828001169487834, + "weighted_rkd_loss": 0.008701342064887286 + }, + { + "epoch": 7.640384055862671, + "flow_matching_loss": 0.01961019691079855, + "grad_norm": 0.15184758603572845, + "learning_rate": 1.9272911548346487e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006623997492715717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028047890635207297, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01605278393253684, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5974183738231659, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26260, + "weighted_flow_matching_loss": 0.01961019691079855, + "weighted_rkd_loss": 0.009590245597064496 + }, + { + "epoch": 7.643293569973815, + "flow_matching_loss": 0.019549050275236367, + "grad_norm": 0.15652413666248322, + "learning_rate": 1.926498607600257e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0065481121186167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003037413500715047, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016133637446910144, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5971615791320801, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26270, + "weighted_flow_matching_loss": 0.019549050275236367, + "weighted_rkd_loss": 0.009634367655962706 + }, + { + "epoch": 7.6462030840849575, + "flow_matching_loss": 0.02014795709401369, + "grad_norm": 0.14797939360141754, + "learning_rate": 1.9257059308068492e-05, + "loss": 0.0292, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006281135231256485, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026266245055012404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015188894979655743, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5969047486782074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26280, + "weighted_flow_matching_loss": 0.02014795709401369, + "weighted_rkd_loss": 0.009066407941281796 + }, + { + "epoch": 7.649112598196101, + "flow_matching_loss": 0.018680213205516338, + "grad_norm": 0.1842467337846756, + "learning_rate": 1.9249131246952195e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0064277640543878075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028064949670806526, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01566202286630869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5966478884220123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26290, + "weighted_flow_matching_loss": 0.018680213205516338, + "weighted_rkd_loss": 0.009344662306830287 + }, + { + "epoch": 7.652022112307245, + "flow_matching_loss": 0.02017693119123578, + "grad_norm": 0.16097192466259003, + "learning_rate": 1.924120189506202e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055374626070261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024146226001903414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013489547837525605, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5963909924030304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26300, + "weighted_flow_matching_loss": 0.02017693119123578, + "weighted_rkd_loss": 0.008045119093731045 + }, + { + "epoch": 7.654931626418388, + "flow_matching_loss": 0.02040873859077692, + "grad_norm": 0.1721460074186325, + "learning_rate": 1.9233271254806686e-05, + "loss": 0.0296, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006401535868644714, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026616953080520036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015464767068624496, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5961340963840485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26310, + "weighted_flow_matching_loss": 0.02040873859077692, + "weighted_rkd_loss": 0.00921908370219171 + }, + { + "epoch": 7.6578411405295315, + "flow_matching_loss": 0.017158705089241267, + "grad_norm": 0.15736688673496246, + "learning_rate": 1.9225339328595323e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005790608515962958, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024668656173162163, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01404808284714818, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.595877182483673, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26320, + "weighted_flow_matching_loss": 0.017158705089241267, + "weighted_rkd_loss": 0.00837079486809671 + }, + { + "epoch": 7.660750654640675, + "flow_matching_loss": 0.019346489757299423, + "grad_norm": 0.15802697837352753, + "learning_rate": 1.9217406118837435e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005912157497368753, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023554972372949123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014179812278598547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5956202208995819, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26330, + "weighted_flow_matching_loss": 0.019346489757299423, + "weighted_rkd_loss": 0.00844587329775095 + }, + { + "epoch": 7.663660168751818, + "flow_matching_loss": 0.018635240849107504, + "grad_norm": 0.14786426723003387, + "learning_rate": 1.9209471627942922e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006194496201351285, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002772038895636797, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015161031484603881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5953632354736328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26340, + "weighted_flow_matching_loss": 0.018635240849107504, + "weighted_rkd_loss": 0.009026304772123694 + }, + { + "epoch": 7.666569682862962, + "flow_matching_loss": 0.018124727811664344, + "grad_norm": 0.13959334790706635, + "learning_rate": 1.9201535858322072e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0067590669263154265, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0031772566959261893, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016695390548557044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5951062202453613, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26350, + "weighted_flow_matching_loss": 0.018124727811664344, + "weighted_rkd_loss": 0.009935573767870664 + }, + { + "epoch": 7.669479196974105, + "flow_matching_loss": 0.017818915471434594, + "grad_norm": 0.1637018918991089, + "learning_rate": 1.9193598812385562e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006108907982707024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026335015427321196, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014851317647844553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5948491871356965, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26360, + "weighted_flow_matching_loss": 0.017818915471434594, + "weighted_rkd_loss": 0.008834306662902237 + }, + { + "epoch": 7.672388711085249, + "flow_matching_loss": 0.018289452791213988, + "grad_norm": 0.17642328143119812, + "learning_rate": 1.9185660492544464e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005821089353412389, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024828479159623384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014125026669353247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5945921242237091, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26370, + "weighted_flow_matching_loss": 0.018289452791213988, + "weighted_rkd_loss": 0.008398607140406966 + }, + { + "epoch": 7.675298225196392, + "flow_matching_loss": 0.018309073615819217, + "grad_norm": 0.17648883163928986, + "learning_rate": 1.917772090121022e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005533455451950431, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023520701564848424, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013418981060385703, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5943350374698639, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26380, + "weighted_flow_matching_loss": 0.018309073615819217, + "weighted_rkd_loss": 0.007975318795070052 + }, + { + "epoch": 7.678207739307536, + "flow_matching_loss": 0.018130281940102577, + "grad_norm": 0.13972772657871246, + "learning_rate": 1.9169780040794678e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005037212651222945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022330240113660692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012307449243962765, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5940779209136963, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26390, + "weighted_flow_matching_loss": 0.018130281940102577, + "weighted_rkd_loss": 0.007311569806188345 + }, + { + "epoch": 7.681117253418679, + "flow_matching_loss": 0.018993467465043067, + "grad_norm": 0.18908607959747314, + "learning_rate": 1.916183791371005e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006497527938336134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027968170470558107, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015791873168200254, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5938207745552063, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26400, + "weighted_flow_matching_loss": 0.018993467465043067, + "weighted_rkd_loss": 0.009377551963552833 + }, + { + "epoch": 7.684026767529822, + "flow_matching_loss": 0.018785600364208222, + "grad_norm": 0.1582397073507309, + "learning_rate": 1.9153894522368952e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00558945806697011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023222323739901184, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013501148391515016, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5935636162757874, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26410, + "weighted_flow_matching_loss": 0.018785600364208222, + "weighted_rkd_loss": 0.008013695990666748 + }, + { + "epoch": 7.686936281640966, + "flow_matching_loss": 0.019704232458025217, + "grad_norm": 0.1542808711528778, + "learning_rate": 1.9145949869184377e-05, + "loss": 0.0295, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006751482980325818, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002940904558636248, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016443870775401593, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5933064222335815, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26420, + "weighted_flow_matching_loss": 0.019704232458025217, + "weighted_rkd_loss": 0.009756259573623537 + }, + { + "epoch": 7.68984579575211, + "flow_matching_loss": 0.016604889277368785, + "grad_norm": 0.1532808244228363, + "learning_rate": 1.91380039565697e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006355676893144846, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027429496636614203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015454303473234177, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5930492043495178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26430, + "weighted_flow_matching_loss": 0.016604889277368785, + "weighted_rkd_loss": 0.00916532906703651 + }, + { + "epoch": 7.692755309863253, + "flow_matching_loss": 0.01949808420613408, + "grad_norm": 0.18204747140407562, + "learning_rate": 1.9130056786938678e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005499123269692064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024417323991656305, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013439979031682014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5927919626235962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26440, + "weighted_flow_matching_loss": 0.01949808420613408, + "weighted_rkd_loss": 0.00796713875606656 + }, + { + "epoch": 7.695664823974396, + "flow_matching_loss": 0.01879888279363513, + "grad_norm": 0.17246031761169434, + "learning_rate": 1.9122108362705447e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005875787418335676, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002562690048944205, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014314264990389347, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5925346910953522, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26450, + "weighted_flow_matching_loss": 0.01879888279363513, + "weighted_rkd_loss": 0.008481627143919469 + }, + { + "epoch": 7.69857433808554, + "flow_matching_loss": 0.019225182384252547, + "grad_norm": 0.1783042997121811, + "learning_rate": 1.9114158686284538e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005880924500524997, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027232327032834292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014485081657767296, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5922773957252503, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26460, + "weighted_flow_matching_loss": 0.019225182384252547, + "weighted_rkd_loss": 0.008579137129709124 + }, + { + "epoch": 7.7014838521966835, + "flow_matching_loss": 0.018282170686870814, + "grad_norm": 0.17707140743732452, + "learning_rate": 1.910620776009085e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006585964048281312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028214263264089825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015993354376405478, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5920200824737549, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26470, + "weighted_flow_matching_loss": 0.018282170686870814, + "weighted_rkd_loss": 0.009468312747776509 + }, + { + "epoch": 7.704393366307826, + "flow_matching_loss": 0.018578093405812977, + "grad_norm": 0.1695687621831894, + "learning_rate": 1.9098255586539653e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060255571734160185, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026423473726026716, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014693461824208498, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5917627334594726, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26480, + "weighted_flow_matching_loss": 0.018578093405812977, + "weighted_rkd_loss": 0.008694929629564285 + }, + { + "epoch": 7.70730288041897, + "flow_matching_loss": 0.019320129603147506, + "grad_norm": 0.14758923649787903, + "learning_rate": 1.909030216804663e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006026501068845391, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025312208570539953, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014584222994744777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.591505366563797, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26490, + "weighted_flow_matching_loss": 0.019320129603147506, + "weighted_rkd_loss": 0.008626635977998376 + }, + { + "epoch": 7.710212394530114, + "flow_matching_loss": 0.019410362746566535, + "grad_norm": 0.1436455100774765, + "learning_rate": 1.9082347507027797e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005516575975343585, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022863500402309, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013319501839578151, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.591247969865799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26500, + "weighted_flow_matching_loss": 0.019410362746566535, + "weighted_rkd_loss": 0.007875180011615157 + }, + { + "epoch": 7.713121908641257, + "flow_matching_loss": 0.01783551266416907, + "grad_norm": 0.1579979509115219, + "learning_rate": 1.9074391605899584e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005623024562373758, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002422130841296166, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013668179884552956, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.590990549325943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26510, + "weighted_flow_matching_loss": 0.01783551266416907, + "weighted_rkd_loss": 0.008077795431017876 + }, + { + "epoch": 7.7160314227524, + "flow_matching_loss": 0.018320364970713855, + "grad_norm": 0.1489427089691162, + "learning_rate": 1.906643446707878e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006153046851977706, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025761729455552997, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01488226680085063, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5907331109046936, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26520, + "weighted_flow_matching_loss": 0.018320364970713855, + "weighted_rkd_loss": 0.008791304379701614 + }, + { + "epoch": 7.718940936863544, + "flow_matching_loss": 0.019466729275882243, + "grad_norm": 0.16868340969085693, + "learning_rate": 1.9058476092982552e-05, + "loss": 0.0279, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005877867108210921, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002565353875979781, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014321088138967753, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5904756188392639, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26530, + "weighted_flow_matching_loss": 0.019466729275882243, + "weighted_rkd_loss": 0.00845614611171186 + }, + { + "epoch": 7.721850450974687, + "flow_matching_loss": 0.018338189460337162, + "grad_norm": 0.2005978524684906, + "learning_rate": 1.9050516486028444e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061266646720468994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024752163561061026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01472854558378458, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5902181267738342, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26540, + "weighted_flow_matching_loss": 0.018338189460337162, + "weighted_rkd_loss": 0.008693034015595912 + }, + { + "epoch": 7.7247599650858305, + "flow_matching_loss": 0.0168434358201921, + "grad_norm": 0.18364964425563812, + "learning_rate": 1.9042555648634375e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057005220092833046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002380408474709839, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013781452551484108, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5899606287479401, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26550, + "weighted_flow_matching_loss": 0.0168434358201921, + "weighted_rkd_loss": 0.008130485936999322 + }, + { + "epoch": 7.727669479196974, + "flow_matching_loss": 0.018106393329799176, + "grad_norm": 0.16430196166038513, + "learning_rate": 1.903459358321864e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005788078019395471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002369412628468126, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013945568539202213, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5897030770778656, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26560, + "weighted_flow_matching_loss": 0.018106393329799176, + "weighted_rkd_loss": 0.00822383356280625 + }, + { + "epoch": 7.730578993308118, + "flow_matching_loss": 0.01792710991576314, + "grad_norm": 0.1604715883731842, + "learning_rate": 1.9026630292199896e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005555375316180289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024135987157933416, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013524349313229322, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5894455075263977, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26570, + "weighted_flow_matching_loss": 0.01792710991576314, + "weighted_rkd_loss": 0.007971808128058911 + }, + { + "epoch": 7.733488507419261, + "flow_matching_loss": 0.017763447668403386, + "grad_norm": 0.14880895614624023, + "learning_rate": 1.901866577799719e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006271505216136575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025298429653048517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015072853490710258, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5891879141330719, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26580, + "weighted_flow_matching_loss": 0.017763447668403386, + "weighted_rkd_loss": 0.008880794141441583 + }, + { + "epoch": 7.7363980215304045, + "flow_matching_loss": 0.01921672113239765, + "grad_norm": 0.1612374186515808, + "learning_rate": 1.901070004302993e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006188222067430615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00263938867719844, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015015832893550397, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5889303028583527, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26590, + "weighted_flow_matching_loss": 0.01921672113239765, + "weighted_rkd_loss": 0.008843361400067806 + }, + { + "epoch": 7.739307535641548, + "flow_matching_loss": 0.017109622061252595, + "grad_norm": 0.16992206871509552, + "learning_rate": 1.900273308971789e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005728102056309581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024925631005316973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01394876716658473, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5886726617813111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26600, + "weighted_flow_matching_loss": 0.017109622061252595, + "weighted_rkd_loss": 0.00821119910106063 + }, + { + "epoch": 7.742217049752691, + "flow_matching_loss": 0.019366617687046526, + "grad_norm": 0.2214490920305252, + "learning_rate": 1.899476492048122e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005875187832862139, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025645118323154746, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01431488748639822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5884149968624115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26610, + "weighted_flow_matching_loss": 0.019366617687046526, + "weighted_rkd_loss": 0.008423107955604792 + }, + { + "epoch": 7.745126563863835, + "flow_matching_loss": 0.018538920674473047, + "grad_norm": 0.18515540659427643, + "learning_rate": 1.8986795537740448e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053371374495327474, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002687417273409665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01336169233545661, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.588157308101654, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26620, + "weighted_flow_matching_loss": 0.018538920674473047, + "weighted_rkd_loss": 0.007858832366764545 + }, + { + "epoch": 7.748036077974978, + "flow_matching_loss": 0.021513899601995946, + "grad_norm": 0.14824733138084412, + "learning_rate": 1.897882494391645e-05, + "loss": 0.0306, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063375512603670355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027652312768623234, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015440333727747202, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5878995895385742, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26630, + "weighted_flow_matching_loss": 0.021513899601995946, + "weighted_rkd_loss": 0.00907729659229517 + }, + { + "epoch": 7.750945592086122, + "flow_matching_loss": 0.01780034266412258, + "grad_norm": 0.20670932531356812, + "learning_rate": 1.8970853141430497e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005911988113075495, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024704826064407825, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014294459018856287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5876418471336364, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26640, + "weighted_flow_matching_loss": 0.01780034266412258, + "weighted_rkd_loss": 0.008399938978254796 + }, + { + "epoch": 7.753855106197265, + "flow_matching_loss": 0.019692080840468407, + "grad_norm": 0.14842215180397034, + "learning_rate": 1.89628801327042e-05, + "loss": 0.0289, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006470295926555991, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027964656706899404, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01573705757036805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5873840868473053, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26650, + "weighted_flow_matching_loss": 0.019692080840468407, + "weighted_rkd_loss": 0.009243761841207742 + }, + { + "epoch": 7.756764620308409, + "flow_matching_loss": 0.018516074866056442, + "grad_norm": 0.15407082438468933, + "learning_rate": 1.8954905920159555e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005973224574699998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026374880922958254, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014583937264978886, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5871263027191163, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26660, + "weighted_flow_matching_loss": 0.018516074866056442, + "weighted_rkd_loss": 0.008562586316838861 + }, + { + "epoch": 7.7596741344195515, + "flow_matching_loss": 0.018930798582732677, + "grad_norm": 0.16688679158687592, + "learning_rate": 1.8946930506218916e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006544109666720033, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002998869679868221, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.016087089106440543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5868684887886048, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26670, + "weighted_flow_matching_loss": 0.018930798582732677, + "weighted_rkd_loss": 0.009440980525687337 + }, + { + "epoch": 7.762583648530695, + "flow_matching_loss": 0.018736127763986588, + "grad_norm": 0.1478426158428192, + "learning_rate": 1.8938953893305e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005413509649224579, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00223487337352708, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013061892706900835, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5866106569766998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26680, + "weighted_flow_matching_loss": 0.018736127763986588, + "weighted_rkd_loss": 0.0076623589266091585 + }, + { + "epoch": 7.765493162641839, + "flow_matching_loss": 0.017796213366091253, + "grad_norm": 0.15229561924934387, + "learning_rate": 1.893097608384089e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005704414984211326, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023602049332112074, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013769034948199987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.586352801322937, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26690, + "weighted_flow_matching_loss": 0.017796213366091253, + "weighted_rkd_loss": 0.008073479495942592 + }, + { + "epoch": 7.768402676752983, + "flow_matching_loss": 0.017067156825214626, + "grad_norm": 0.16606292128562927, + "learning_rate": 1.892299708025004e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005560253094881773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002381850080564618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013502356410026551, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5860949158668518, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26700, + "weighted_flow_matching_loss": 0.017067156825214626, + "weighted_rkd_loss": 0.00791367320343852 + }, + { + "epoch": 7.771312190864125, + "flow_matching_loss": 0.017083107121288778, + "grad_norm": 0.14384089410305023, + "learning_rate": 1.891501688495626e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006039953837171197, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025288144242949784, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014608722273260355, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5858370125293731, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26710, + "weighted_flow_matching_loss": 0.017083107121288778, + "weighted_rkd_loss": 0.008558371430262923 + }, + { + "epoch": 7.774221704975269, + "flow_matching_loss": 0.017724875174462796, + "grad_norm": 0.15275651216506958, + "learning_rate": 1.8907035500383717e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00589658971875906, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024990206584334373, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014292200282216071, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5855790793895721, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26720, + "weighted_flow_matching_loss": 0.017724875174462796, + "weighted_rkd_loss": 0.008369268896058202 + }, + { + "epoch": 7.777131219086413, + "flow_matching_loss": 0.02000788748264313, + "grad_norm": 0.15846632421016693, + "learning_rate": 1.889905292895695e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005210767220705748, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022879887837916613, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012709523364901543, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5853211283683777, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26730, + "weighted_flow_matching_loss": 0.02000788748264313, + "weighted_rkd_loss": 0.00743916081264615 + }, + { + "epoch": 7.780040733197556, + "flow_matching_loss": 0.01586626973003149, + "grad_norm": 0.16387784481048584, + "learning_rate": 1.889106917310085e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005334302503615618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022586048813536762, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012927209865301848, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5850631594657898, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26740, + "weighted_flow_matching_loss": 0.01586626973003149, + "weighted_rkd_loss": 0.007563214935362339 + }, + { + "epoch": 7.782950247308699, + "flow_matching_loss": 0.01989711420610547, + "grad_norm": 0.1498013287782669, + "learning_rate": 1.8883084235240672e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005758269969373941, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023958203033544122, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013912360277026891, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5848051488399506, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26750, + "weighted_flow_matching_loss": 0.01989711420610547, + "weighted_rkd_loss": 0.008136000763624907 + }, + { + "epoch": 7.785859761419843, + "flow_matching_loss": 0.01869755983352661, + "grad_norm": 0.16386154294013977, + "learning_rate": 1.887509811780203e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006232068222016096, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026878634467720985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015151999983936548, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5845471322536469, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26760, + "weighted_flow_matching_loss": 0.01869755983352661, + "weighted_rkd_loss": 0.008857122343033551 + }, + { + "epoch": 7.788769275530987, + "flow_matching_loss": 0.0188650437630713, + "grad_norm": 0.15425066649913788, + "learning_rate": 1.88671108232109e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005837410362437368, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002600055350922048, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014274876192212104, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5842890977859497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26770, + "weighted_flow_matching_loss": 0.0188650437630713, + "weighted_rkd_loss": 0.008340713614597917 + }, + { + "epoch": 7.79167878964213, + "flow_matching_loss": 0.01905276011675596, + "grad_norm": 0.1682288497686386, + "learning_rate": 1.88591223538936e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005484634451568127, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024262752151116727, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013395544234663248, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5840310156345367, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26780, + "weighted_flow_matching_loss": 0.01905276011675596, + "weighted_rkd_loss": 0.007823389954864979 + }, + { + "epoch": 7.794588303753273, + "flow_matching_loss": 0.019883243553340434, + "grad_norm": 0.1577252596616745, + "learning_rate": 1.885113271227683e-05, + "loss": 0.028, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00572744666133076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002408043190371245, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013862936478108167, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5837729275226593, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26790, + "weighted_flow_matching_loss": 0.019883243553340434, + "weighted_rkd_loss": 0.008092853659763932 + }, + { + "epoch": 7.797497817864417, + "flow_matching_loss": 0.018281731661409138, + "grad_norm": 0.14510385692119598, + "learning_rate": 1.8843141900787625e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005740510439500213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002514400437939912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013995421305298805, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5835148215293884, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26800, + "weighted_flow_matching_loss": 0.018281731661409138, + "weighted_rkd_loss": 0.008166634384542704 + }, + { + "epoch": 7.80040733197556, + "flow_matching_loss": 0.019555175118148325, + "grad_norm": 0.21388928592205048, + "learning_rate": 1.8835149921853383e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006192208034917712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027276397682726382, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01511205593124032, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5832566797733307, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26810, + "weighted_flow_matching_loss": 0.019555175118148325, + "weighted_rkd_loss": 0.008814165648072959 + }, + { + "epoch": 7.8033168460867035, + "flow_matching_loss": 0.018659673817455768, + "grad_norm": 0.16851310431957245, + "learning_rate": 1.8827156777901853e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006033269921317697, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027703687082976103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014836908504366874, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5829985201358795, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26820, + "weighted_flow_matching_loss": 0.018659673817455768, + "weighted_rkd_loss": 0.00865007913671434 + }, + { + "epoch": 7.806226360197847, + "flow_matching_loss": 0.018986627366393803, + "grad_norm": 0.1636441946029663, + "learning_rate": 1.8819162471361144e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005947037041187287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026275802752934395, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014521654322743415, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5827403366565704, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26830, + "weighted_flow_matching_loss": 0.018986627366393803, + "weighted_rkd_loss": 0.008462352678179741 + }, + { + "epoch": 7.80913587430899, + "flow_matching_loss": 0.017836721800267696, + "grad_norm": 0.1727214753627777, + "learning_rate": 1.8811167004659715e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006179759185761214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026872079819440843, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015046726353466511, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5824821293354034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26840, + "weighted_flow_matching_loss": 0.017836721800267696, + "weighted_rkd_loss": 0.008764406805858016 + }, + { + "epoch": 7.812045388420134, + "flow_matching_loss": 0.017293765861541032, + "grad_norm": 0.1446101814508438, + "learning_rate": 1.8803170380226383e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006171859428286552, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026632246794179083, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015006943419575692, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.582223904132843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26850, + "weighted_flow_matching_loss": 0.017293765861541032, + "weighted_rkd_loss": 0.008737423503771425 + }, + { + "epoch": 7.8149549025312774, + "flow_matching_loss": 0.01763680884614587, + "grad_norm": 0.14545702934265137, + "learning_rate": 1.8795172600490302e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005813940986990928, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023669076734222473, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013994789589196444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5819656491279602, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26860, + "weighted_flow_matching_loss": 0.01763680884614587, + "weighted_rkd_loss": 0.008144431700929999 + }, + { + "epoch": 7.81786441664242, + "flow_matching_loss": 0.01947342175990343, + "grad_norm": 0.17004550993442535, + "learning_rate": 1.878717366788099e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005436475411988795, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002420056250412017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013293007109314203, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.581707376241684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26870, + "weighted_flow_matching_loss": 0.01947342175990343, + "weighted_rkd_loss": 0.007732596900314093 + }, + { + "epoch": 7.820773930753564, + "flow_matching_loss": 0.018710755556821824, + "grad_norm": 0.23200146853923798, + "learning_rate": 1.8779173584828316e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006130044301971793, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026052128756418823, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014865301642566919, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5814490854740143, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26880, + "weighted_flow_matching_loss": 0.018710755556821824, + "weighted_rkd_loss": 0.008643298596143722 + }, + { + "epoch": 7.823683444864708, + "flow_matching_loss": 0.018017966113984586, + "grad_norm": 0.1603008508682251, + "learning_rate": 1.8771172353762488e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006197632430121303, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002544401597697288, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014939666632562876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5811907589435578, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26890, + "weighted_flow_matching_loss": 0.018017966113984586, + "weighted_rkd_loss": 0.008682764507830143 + }, + { + "epoch": 7.826592958975851, + "flow_matching_loss": 0.018391946982592344, + "grad_norm": 0.16118896007537842, + "learning_rate": 1.876316997711407e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006387441093102098, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025900029577314854, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015364885143935681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5809324204921722, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26900, + "weighted_flow_matching_loss": 0.018391946982592344, + "weighted_rkd_loss": 0.00892594801262021 + }, + { + "epoch": 7.829502473086994, + "flow_matching_loss": 0.018570001516491175, + "grad_norm": 0.19044871628284454, + "learning_rate": 1.8755166457313972e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006175392353907227, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027767505729570986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015127535350620747, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5806740522384644, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26910, + "weighted_flow_matching_loss": 0.018570001516491175, + "weighted_rkd_loss": 0.008784037176519632 + }, + { + "epoch": 7.832411987198138, + "flow_matching_loss": 0.018272156454622747, + "grad_norm": 0.16494154930114746, + "learning_rate": 1.8747161796793457e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005923157557845116, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024714436265639963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014317758474498987, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.580415666103363, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26920, + "weighted_flow_matching_loss": 0.018272156454622747, + "weighted_rkd_loss": 0.008310193754732608 + }, + { + "epoch": 7.835321501309282, + "flow_matching_loss": 0.020515230670571327, + "grad_norm": 0.15594643354415894, + "learning_rate": 1.8739155997984124e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00577484299428761, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002458076039329171, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014007762167602777, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5801572740077973, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26930, + "weighted_flow_matching_loss": 0.020515230670571327, + "weighted_rkd_loss": 0.008126647351309657 + }, + { + "epoch": 7.838231015420424, + "flow_matching_loss": 0.0185715195722878, + "grad_norm": 0.16132400929927826, + "learning_rate": 1.8731149063317927e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005954448878765106, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025492797372862696, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01445817742496729, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5798988461494445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26940, + "weighted_flow_matching_loss": 0.0185715195722878, + "weighted_rkd_loss": 0.00838416637852788 + }, + { + "epoch": 7.841140529531568, + "flow_matching_loss": 0.01702827773988247, + "grad_norm": 0.16822703182697296, + "learning_rate": 1.872314099522716e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00525900253560394, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002214840066153556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012732845172286034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5796403884887695, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26950, + "weighted_flow_matching_loss": 0.01702827773988247, + "weighted_rkd_loss": 0.007380537036806345 + }, + { + "epoch": 7.844050043642712, + "flow_matching_loss": 0.02146861478686333, + "grad_norm": 0.17810358107089996, + "learning_rate": 1.8715131796144466e-05, + "loss": 0.0302, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006179268937557936, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027267861180007456, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015085323993116618, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.579381912946701, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26960, + "weighted_flow_matching_loss": 0.02146861478686333, + "weighted_rkd_loss": 0.0087401382625103 + }, + { + "epoch": 7.846959557753856, + "flow_matching_loss": 0.01792115131393075, + "grad_norm": 0.13650308549404144, + "learning_rate": 1.870712146850282e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053211364429444075, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023577867192216217, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013000059593468904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5791234254837037, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26970, + "weighted_flow_matching_loss": 0.01792115131393075, + "weighted_rkd_loss": 0.007528667664155364 + }, + { + "epoch": 7.849869071864998, + "flow_matching_loss": 0.018346431478857995, + "grad_norm": 0.13668294250965118, + "learning_rate": 1.8699110014735554e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005962010193616152, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024637682596221565, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014387788623571396, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5788649082183838, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26980, + "weighted_flow_matching_loss": 0.018346431478857995, + "weighted_rkd_loss": 0.008328518783673643 + }, + { + "epoch": 7.852778585976142, + "flow_matching_loss": 0.019254637882113457, + "grad_norm": 0.1749672293663025, + "learning_rate": 1.8691097437276334e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006383823323994875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028563356027007103, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015623982530087233, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.578606367111206, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 26990, + "weighted_flow_matching_loss": 0.019254637882113457, + "weighted_rkd_loss": 0.009040056215599179 + }, + { + "epoch": 7.855688100087286, + "flow_matching_loss": 0.017964050639420748, + "grad_norm": 0.1714823991060257, + "learning_rate": 1.868308373855917e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005037163710221648, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00216874839970842, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012243075668811798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5783478081226349, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27000, + "weighted_flow_matching_loss": 0.017964050639420748, + "weighted_rkd_loss": 0.0070806802716106175 + }, + { + "epoch": 7.858597614198429, + "flow_matching_loss": 0.019197307433933018, + "grad_norm": 0.1532866656780243, + "learning_rate": 1.8675068921018415e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005653700581751764, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024362780386582017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013743679132312536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5780892312526703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27010, + "weighted_flow_matching_loss": 0.019197307433933018, + "weighted_rkd_loss": 0.00794512527063489 + }, + { + "epoch": 7.861507128309572, + "flow_matching_loss": 0.017628901451826096, + "grad_norm": 0.161074697971344, + "learning_rate": 1.8667052987088756e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005797619465738535, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002488323173020035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014083562046289444, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5778306305408478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27020, + "weighted_flow_matching_loss": 0.017628901451826096, + "weighted_rkd_loss": 0.008137936843559146 + }, + { + "epoch": 7.864416642420716, + "flow_matching_loss": 0.019637531042098998, + "grad_norm": 0.1656741052865982, + "learning_rate": 1.8659035939205224e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005646520061418414, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024174985359422863, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013710538763552904, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5775720059871674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27030, + "weighted_flow_matching_loss": 0.019637531042098998, + "weighted_rkd_loss": 0.00791874984279275 + }, + { + "epoch": 7.867326156531859, + "flow_matching_loss": 0.018066631443798543, + "grad_norm": 0.17013266682624817, + "learning_rate": 1.8651017779803178e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005309752933681011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002217766235116869, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012837271951138974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5773133635520935, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27040, + "weighted_flow_matching_loss": 0.018066631443798543, + "weighted_rkd_loss": 0.007411069376394153 + }, + { + "epoch": 7.870235670643003, + "flow_matching_loss": 0.017210197076201438, + "grad_norm": 0.1527140736579895, + "learning_rate": 1.8642998511318336e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00549804656766355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002244558639358729, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013240651786327362, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5770546793937683, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27050, + "weighted_flow_matching_loss": 0.017210197076201438, + "weighted_rkd_loss": 0.007640637131407857 + }, + { + "epoch": 7.873145184754146, + "flow_matching_loss": 0.01656512850895524, + "grad_norm": 0.14149023592472076, + "learning_rate": 1.863497813618673e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005594193935394287, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023382493294775487, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013526637200266122, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5767959952354431, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27060, + "weighted_flow_matching_loss": 0.01656512850895524, + "weighted_rkd_loss": 0.0078020811080932615 + }, + { + "epoch": 7.876054698865289, + "flow_matching_loss": 0.018181073665618896, + "grad_norm": 0.1368563026189804, + "learning_rate": 1.8626956656844746e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006349684251472354, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002674889052286744, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015374257601797581, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5765373110771179, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27070, + "weighted_flow_matching_loss": 0.018181073665618896, + "weighted_rkd_loss": 0.008863813895732164 + }, + { + "epoch": 7.878964212976433, + "flow_matching_loss": 0.018491281569004057, + "grad_norm": 0.16732074320316315, + "learning_rate": 1.861893407572909e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005756550189107657, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002516341698355973, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01402944205328822, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.576278567314148, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27080, + "weighted_flow_matching_loss": 0.018491281569004057, + "weighted_rkd_loss": 0.008084916416555643 + }, + { + "epoch": 7.8818737270875765, + "flow_matching_loss": 0.01958615444600582, + "grad_norm": 0.15918532013893127, + "learning_rate": 1.861091039527682e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057600173633545635, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025127686909399926, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014032803568989038, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.576019823551178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27090, + "weighted_flow_matching_loss": 0.01958615444600582, + "weighted_rkd_loss": 0.008083272911608218 + }, + { + "epoch": 7.88478324119872, + "flow_matching_loss": 0.018524029478430747, + "grad_norm": 0.14314542710781097, + "learning_rate": 1.8602885617925307e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005874858191236854, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024477550061419604, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014197471365332604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5757610499858856, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27100, + "weighted_flow_matching_loss": 0.018524029478430747, + "weighted_rkd_loss": 0.008174331765621901 + }, + { + "epoch": 7.887692755309863, + "flow_matching_loss": 0.017896097060292958, + "grad_norm": 0.14258132874965668, + "learning_rate": 1.8594859746112273e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006118463352322578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002779017051216215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015015943720936776, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5755022585391998, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27110, + "weighted_flow_matching_loss": 0.017896097060292958, + "weighted_rkd_loss": 0.00864173541776836 + }, + { + "epoch": 7.890602269421007, + "flow_matching_loss": 0.017234269343316556, + "grad_norm": 0.1445445418357849, + "learning_rate": 1.8586832782275768e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005219213012605905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002116663078777492, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012555088941007852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.575243455171585, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27120, + "weighted_flow_matching_loss": 0.017234269343316556, + "weighted_rkd_loss": 0.0072221873793751005 + }, + { + "epoch": 7.89351178353215, + "flow_matching_loss": 0.019127140566706658, + "grad_norm": 0.14548666775226593, + "learning_rate": 1.857880472885417e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053521548863500355, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022712231264449658, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012975532934069634, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5749846160411834, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27130, + "weighted_flow_matching_loss": 0.019127140566706658, + "weighted_rkd_loss": 0.007460807031020522 + }, + { + "epoch": 7.896421297643293, + "flow_matching_loss": 0.017761355824768545, + "grad_norm": 0.1415204107761383, + "learning_rate": 1.8570775588286185e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005483488226309419, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002162220678292215, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013129196967929602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.574725764989853, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27140, + "weighted_flow_matching_loss": 0.017761355824768545, + "weighted_rkd_loss": 0.007545649679377675 + }, + { + "epoch": 7.899330811754437, + "flow_matching_loss": 0.019682378601282836, + "grad_norm": 0.16532371938228607, + "learning_rate": 1.8562745363010854e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005966789973899722, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027515526162460446, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014685132540762424, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5744668960571289, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27150, + "weighted_flow_matching_loss": 0.019682378601282836, + "weighted_rkd_loss": 0.008436055621132255 + }, + { + "epoch": 7.902240325865581, + "flow_matching_loss": 0.017038043681532146, + "grad_norm": 0.1285293847322464, + "learning_rate": 1.8554714055467555e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005758974235504866, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002498007705435157, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014015956316143274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5742079973220825, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27160, + "weighted_flow_matching_loss": 0.017038043681532146, + "weighted_rkd_loss": 0.008048225613310932 + }, + { + "epoch": 7.905149839976724, + "flow_matching_loss": 0.017853710521012544, + "grad_norm": 0.15543483197689056, + "learning_rate": 1.8546681668095978e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00532946027815342, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002167528681457043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012826449237763882, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5739490866661072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27170, + "weighted_flow_matching_loss": 0.017853710521012544, + "weighted_rkd_loss": 0.007361654844135046 + }, + { + "epoch": 7.908059354087867, + "flow_matching_loss": 0.01880219206213951, + "grad_norm": 0.18915344774723053, + "learning_rate": 1.853864820333616e-05, + "loss": 0.0268, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005764169665053487, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002485066663939506, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01401340626180172, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5736901462078094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27180, + "weighted_flow_matching_loss": 0.01880219206213951, + "weighted_rkd_loss": 0.008039381681010127 + }, + { + "epoch": 7.910968868199011, + "flow_matching_loss": 0.017884033918380737, + "grad_norm": 0.1642964631319046, + "learning_rate": 1.8530613663628448e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005827941233292222, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025616817292757332, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014217564277350903, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5734311938285828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27190, + "weighted_flow_matching_loss": 0.017884033918380737, + "weighted_rkd_loss": 0.008152845688164235 + }, + { + "epoch": 7.913878382310154, + "flow_matching_loss": 0.0194832019507885, + "grad_norm": 0.16946455836296082, + "learning_rate": 1.8522578051413525e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00616703499108553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002654715452808887, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014988785237073898, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5731722116470337, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27200, + "weighted_flow_matching_loss": 0.0194832019507885, + "weighted_rkd_loss": 0.00859122546389699 + }, + { + "epoch": 7.916787896421297, + "flow_matching_loss": 0.019619043357670306, + "grad_norm": 0.1460336148738861, + "learning_rate": 1.85145413691324e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006179218459874391, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002708662743680179, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015067099407315255, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5729132235050202, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27210, + "weighted_flow_matching_loss": 0.019619043357670306, + "weighted_rkd_loss": 0.008632155088707804 + }, + { + "epoch": 7.919697410532441, + "flow_matching_loss": 0.017043671756982803, + "grad_norm": 0.14959490299224854, + "learning_rate": 1.8506503619226402e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0063167883083224295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026968637481331824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015330440551042556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5726542055606842, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27220, + "weighted_flow_matching_loss": 0.017043671756982803, + "weighted_rkd_loss": 0.008779031643643975 + }, + { + "epoch": 7.922606924643585, + "flow_matching_loss": 0.017997770197689535, + "grad_norm": 0.14515097439289093, + "learning_rate": 1.849846480413719e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006175470980815589, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025375147815793753, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01488845655694604, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5723951756954193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27230, + "weighted_flow_matching_loss": 0.017997770197689535, + "weighted_rkd_loss": 0.008522140514105558 + }, + { + "epoch": 7.925516438754728, + "flow_matching_loss": 0.01680589746683836, + "grad_norm": 0.17687100172042847, + "learning_rate": 1.8490424926306745e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006285207625478506, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002753635379485786, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015324050653725862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.572136116027832, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27240, + "weighted_flow_matching_loss": 0.01680589746683836, + "weighted_rkd_loss": 0.008767398074269295 + }, + { + "epoch": 7.928425952865871, + "flow_matching_loss": 0.018401913158595563, + "grad_norm": 0.1980612874031067, + "learning_rate": 1.8482383988177362e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005286637926474214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022157146129757168, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012788990512490273, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5718770444393158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27250, + "weighted_flow_matching_loss": 0.018401913158595563, + "weighted_rkd_loss": 0.007313755620270967 + }, + { + "epoch": 7.931335466977015, + "flow_matching_loss": 0.017019593436270953, + "grad_norm": 0.1523139327764511, + "learning_rate": 1.8474341992191666e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006284312717616558, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026265536667779087, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015195179078727961, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5716179490089417, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27260, + "weighted_flow_matching_loss": 0.017019593436270953, + "weighted_rkd_loss": 0.008685864415019751 + }, + { + "epoch": 7.934244981088158, + "flow_matching_loss": 0.019958732090890406, + "grad_norm": 0.15856602787971497, + "learning_rate": 1.8466298940792614e-05, + "loss": 0.0281, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00586314001120627, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002459817740600556, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01418609768152237, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5713588297367096, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27270, + "weighted_flow_matching_loss": 0.019958732090890406, + "weighted_rkd_loss": 0.008105324162170292 + }, + { + "epoch": 7.937154495199302, + "flow_matching_loss": 0.018971712421625854, + "grad_norm": 0.22609780728816986, + "learning_rate": 1.8458254836423467e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00610201209783554, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026758941472508012, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014879918284714222, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5710996985435486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27280, + "weighted_flow_matching_loss": 0.018971712421625854, + "weighted_rkd_loss": 0.008497868198901415 + }, + { + "epoch": 7.940064009310445, + "flow_matching_loss": 0.01727373553439975, + "grad_norm": 0.17822770774364471, + "learning_rate": 1.8450209681527808e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005769624281674624, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023138344287872313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013853082899004221, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5708405494689941, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27290, + "weighted_flow_matching_loss": 0.01727373553439975, + "weighted_rkd_loss": 0.007907913718372583 + }, + { + "epoch": 7.942973523421589, + "flow_matching_loss": 0.018853843118995427, + "grad_norm": 0.1477428823709488, + "learning_rate": 1.8442163478549546e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005893772840499878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002452231198549271, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014239776879549026, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5705813765525818, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27300, + "weighted_flow_matching_loss": 0.018853843118995427, + "weighted_rkd_loss": 0.008124940190464258 + }, + { + "epoch": 7.945883037532732, + "flow_matching_loss": 0.019335913192480802, + "grad_norm": 0.15721634030342102, + "learning_rate": 1.8434116229932905e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005712263891473413, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025206505320966243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013945178501307965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5703221797943115, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27310, + "weighted_flow_matching_loss": 0.019335913192480802, + "weighted_rkd_loss": 0.007953237183392048 + }, + { + "epoch": 7.9487925516438755, + "flow_matching_loss": 0.017579386290162802, + "grad_norm": 0.14251075685024261, + "learning_rate": 1.8426067938122423e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005594524694606662, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024684171890839933, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013657466415315867, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5700629711151123, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27320, + "weighted_flow_matching_loss": 0.017579386290162802, + "weighted_rkd_loss": 0.007785616163164377 + }, + { + "epoch": 7.951702065755019, + "flow_matching_loss": 0.017446045484393834, + "grad_norm": 0.15003152191638947, + "learning_rate": 1.841801860556296e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005322036612778902, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002321453345939517, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012965526711195708, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5698037445545197, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27330, + "weighted_flow_matching_loss": 0.017446045484393834, + "weighted_rkd_loss": 0.007387773739174009 + }, + { + "epoch": 7.954611579866162, + "flow_matching_loss": 0.018565264903008936, + "grad_norm": 0.15469525754451752, + "learning_rate": 1.8409968234699698e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005613294057548046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024992640828713775, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013725852128118277, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5695444941520691, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27340, + "weighted_flow_matching_loss": 0.018565264903008936, + "weighted_rkd_loss": 0.007817525789141655 + }, + { + "epoch": 7.957521093977306, + "flow_matching_loss": 0.018934213928878306, + "grad_norm": 0.15873569250106812, + "learning_rate": 1.840191682797812e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006213903846219182, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026341244112700222, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015061932150274515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5692852437496185, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27350, + "weighted_flow_matching_loss": 0.018934213928878306, + "weighted_rkd_loss": 0.00857451451011002 + }, + { + "epoch": 7.9604306080884495, + "flow_matching_loss": 0.018925976008176804, + "grad_norm": 0.14331834018230438, + "learning_rate": 1.839386438784403e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053365791216492655, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002374762191902846, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013047920539975166, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5690259456634521, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27360, + "weighted_flow_matching_loss": 0.018925976008176804, + "weighted_rkd_loss": 0.007424561399966478 + }, + { + "epoch": 7.963340122199592, + "flow_matching_loss": 0.01598933497443795, + "grad_norm": 0.15646608173847198, + "learning_rate": 1.838581091674355e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006419811910018325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027643117355182767, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015603935532271861, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5687666237354279, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27370, + "weighted_flow_matching_loss": 0.01598933497443795, + "weighted_rkd_loss": 0.008875062316656112 + }, + { + "epoch": 7.966249636310736, + "flow_matching_loss": 0.018934485409408806, + "grad_norm": 0.12897399067878723, + "learning_rate": 1.8377756417123108e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006000227807089686, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002692301454953849, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014692757092416287, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5685073018074036, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27380, + "weighted_flow_matching_loss": 0.018934485409408806, + "weighted_rkd_loss": 0.00835299170576036 + }, + { + "epoch": 7.96915915042188, + "flow_matching_loss": 0.018808884359896184, + "grad_norm": 0.14883410930633545, + "learning_rate": 1.8369700891429455e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006004219595342875, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00259273323463276, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014601172413676977, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5682479619979859, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27390, + "weighted_flow_matching_loss": 0.018808884359896184, + "weighted_rkd_loss": 0.008297112211585045 + }, + { + "epoch": 7.9720686645330225, + "flow_matching_loss": 0.0183613246306777, + "grad_norm": 0.14235052466392517, + "learning_rate": 1.8361644342109642e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00548218279145658, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002340803202241659, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013305168971419335, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5679885983467102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27400, + "weighted_flow_matching_loss": 0.0183613246306777, + "weighted_rkd_loss": 0.0075572171248495575 + }, + { + "epoch": 7.974978178644166, + "flow_matching_loss": 0.01891739321872592, + "grad_norm": 0.16032321751117706, + "learning_rate": 1.8353586771611038e-05, + "loss": 0.0277, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006432738853618503, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002601947414223105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015467424970120191, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5677292168140411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27410, + "weighted_flow_matching_loss": 0.01891739321872592, + "weighted_rkd_loss": 0.008781199296936392 + }, + { + "epoch": 7.97788769275531, + "flow_matching_loss": 0.02036676984280348, + "grad_norm": 0.1434301882982254, + "learning_rate": 1.8345528182381314e-05, + "loss": 0.0286, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059471516869962215, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025733465910889207, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01446764962747693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5674698173999786, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27420, + "weighted_flow_matching_loss": 0.02036676984280348, + "weighted_rkd_loss": 0.008209940511733294 + }, + { + "epoch": 7.980797206866454, + "flow_matching_loss": 0.017452978808432818, + "grad_norm": 0.13665549457073212, + "learning_rate": 1.8337468576868466e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006069716112688184, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027088017435744406, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014848234225064516, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5672104001045227, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27430, + "weighted_flow_matching_loss": 0.017452978808432818, + "weighted_rkd_loss": 0.008422073256224395 + }, + { + "epoch": 7.9837067209775965, + "flow_matching_loss": 0.017951169144362213, + "grad_norm": 0.15507031977176666, + "learning_rate": 1.8329407957520786e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059680409729480745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026610178407281636, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01459709983319044, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5669509589672088, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27440, + "weighted_flow_matching_loss": 0.017951169144362213, + "weighted_rkd_loss": 0.00827575260773301 + }, + { + "epoch": 7.98661623508874, + "flow_matching_loss": 0.01789000853896141, + "grad_norm": 0.15801364183425903, + "learning_rate": 1.8321346326786872e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00595059571787715, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002520280983299017, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014421472605317832, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5666915059089661, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27450, + "weighted_flow_matching_loss": 0.01789000853896141, + "weighted_rkd_loss": 0.008172363415360451 + }, + { + "epoch": 7.989525749199884, + "flow_matching_loss": 0.017674290388822556, + "grad_norm": 0.14161650836467743, + "learning_rate": 1.831328368711564e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005545944953337312, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023748830775730313, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013466773182153701, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5664320409297943, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27460, + "weighted_flow_matching_loss": 0.017674290388822556, + "weighted_rkd_loss": 0.007628003368154168 + }, + { + "epoch": 7.992435263311027, + "flow_matching_loss": 0.018231282383203505, + "grad_norm": 0.15292546153068542, + "learning_rate": 1.830522004095631e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052508613094687465, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002191037347074598, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012692759837955236, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5661725401878357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27470, + "weighted_flow_matching_loss": 0.018231282383203505, + "weighted_rkd_loss": 0.007186302216723561 + }, + { + "epoch": 7.99534477742217, + "flow_matching_loss": 0.018399513978511094, + "grad_norm": 0.16617682576179504, + "learning_rate": 1.8297155390758394e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00646982560865581, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002579704811796546, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015519356168806554, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5659130275249481, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27480, + "weighted_flow_matching_loss": 0.018399513978511094, + "weighted_rkd_loss": 0.00878256601281464 + }, + { + "epoch": 7.998254291533314, + "flow_matching_loss": 0.0185995327308774, + "grad_norm": 0.13457171618938446, + "learning_rate": 1.828908973897173e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006090672453865409, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002571040892507881, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014752385672181845, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5656535029411316, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27490, + "weighted_flow_matching_loss": 0.0185995327308774, + "weighted_rkd_loss": 0.008344733715057373 + }, + { + "epoch": 8.001163805644458, + "flow_matching_loss": 0.016423931531608104, + "grad_norm": 0.11943697184324265, + "learning_rate": 1.828102308804645e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005489815725013614, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002183997689280659, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013163629081100225, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5653939545154572, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27500, + "weighted_flow_matching_loss": 0.016423931531608104, + "weighted_rkd_loss": 0.007442682608962059 + }, + { + "epoch": 8.004073319755602, + "flow_matching_loss": 0.018942934274673463, + "grad_norm": 0.17161345481872559, + "learning_rate": 1.827295544043299e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060604372527450325, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002635093801654875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014755968376994133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5651344001293183, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27510, + "weighted_flow_matching_loss": 0.018942934274673463, + "weighted_rkd_loss": 0.008339127572253347 + }, + { + "epoch": 8.006982833866743, + "flow_matching_loss": 0.018264892138540746, + "grad_norm": 0.13782952725887299, + "learning_rate": 1.826488679858207e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058431218843907114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002655868011061102, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014342111721634864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5648748099803924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27520, + "weighted_flow_matching_loss": 0.018264892138540746, + "weighted_rkd_loss": 0.008101483294740319 + }, + { + "epoch": 8.009892347977887, + "flow_matching_loss": 0.019869653228670358, + "grad_norm": 0.15751737356185913, + "learning_rate": 1.8256817164944756e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006137543218210339, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002593292819801718, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014868379384279252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5646152079105378, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27530, + "weighted_flow_matching_loss": 0.019869653228670358, + "weighted_rkd_loss": 0.008394959801807999 + }, + { + "epoch": 8.012801862089031, + "flow_matching_loss": 0.01661978419870138, + "grad_norm": 0.14493544399738312, + "learning_rate": 1.8248746541972377e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006330249411985278, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002713381382636726, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015373880229890346, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5643555998802186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27540, + "weighted_flow_matching_loss": 0.01661978419870138, + "weighted_rkd_loss": 0.008676315238699318 + }, + { + "epoch": 8.015711376200175, + "flow_matching_loss": 0.018141068890690802, + "grad_norm": 0.1615590751171112, + "learning_rate": 1.8240674932116575e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005415642773732543, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023696671123616397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013200952764600515, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5640959560871124, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27550, + "weighted_flow_matching_loss": 0.018141068890690802, + "weighted_rkd_loss": 0.0074465794954448935 + }, + { + "epoch": 8.018620890311318, + "flow_matching_loss": 0.01894797831773758, + "grad_norm": 0.16540345549583435, + "learning_rate": 1.8232602337829297e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006201226892881095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002753026457503438, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015155480150133372, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5638363063335419, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27560, + "weighted_flow_matching_loss": 0.01894797831773758, + "weighted_rkd_loss": 0.008545161737129093 + }, + { + "epoch": 8.021530404422462, + "flow_matching_loss": 0.01998403212055564, + "grad_norm": 0.13832968473434448, + "learning_rate": 1.8224528761562777e-05, + "loss": 0.0285, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006224500248208642, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026149201672524213, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01506392052397132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.563576626777649, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27570, + "weighted_flow_matching_loss": 0.01998403212055564, + "weighted_rkd_loss": 0.008489612909033894 + }, + { + "epoch": 8.024439918533606, + "flow_matching_loss": 0.0177412798628211, + "grad_norm": 0.17339898645877838, + "learning_rate": 1.821645420576956e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006018363405019045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002597668848466128, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01463439567014575, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.563316947221756, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27580, + "weighted_flow_matching_loss": 0.0177412798628211, + "weighted_rkd_loss": 0.008243667334318161 + }, + { + "epoch": 8.027349432644748, + "flow_matching_loss": 0.0166356835514307, + "grad_norm": 0.1697693020105362, + "learning_rate": 1.820837867290248e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005641348334029317, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002271668252069503, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013554365001618862, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5630572438240051, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27590, + "weighted_flow_matching_loss": 0.0166356835514307, + "weighted_rkd_loss": 0.007631915295496583 + }, + { + "epoch": 8.030258946755891, + "flow_matching_loss": 0.019731460697948934, + "grad_norm": 0.18019625544548035, + "learning_rate": 1.820030216541467e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006325228605419397, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027118757250718774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015362332947552204, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5627975165843964, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27600, + "weighted_flow_matching_loss": 0.019731460697948934, + "weighted_rkd_loss": 0.008645823085680604 + }, + { + "epoch": 8.033168460867035, + "flow_matching_loss": 0.018875350151211022, + "grad_norm": 0.151799276471138, + "learning_rate": 1.8192224685759562e-05, + "loss": 0.0276, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00642969012260437, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027179101482033728, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015577290393412112, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5625377833843231, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27610, + "weighted_flow_matching_loss": 0.018875350151211022, + "weighted_rkd_loss": 0.008762880600988865 + }, + { + "epoch": 8.036077974978179, + "flow_matching_loss": 0.01833266466856003, + "grad_norm": 0.13075272738933563, + "learning_rate": 1.8184146236390883e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0056504130596295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002415575506165624, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013716401532292366, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.562278026342392, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27620, + "weighted_flow_matching_loss": 0.01833266466856003, + "weighted_rkd_loss": 0.007712433999404311 + }, + { + "epoch": 8.038987489089322, + "flow_matching_loss": 0.017025368474423885, + "grad_norm": 0.14237959682941437, + "learning_rate": 1.8176066819762644e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005787671217694879, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024568594410084186, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014032202027738094, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5620182514190674, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27630, + "weighted_flow_matching_loss": 0.017025368474423885, + "weighted_rkd_loss": 0.007886332925409079 + }, + { + "epoch": 8.041897003200466, + "flow_matching_loss": 0.01707847137004137, + "grad_norm": 0.13036279380321503, + "learning_rate": 1.8167986438329167e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005700915260240436, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002476439659949392, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013878270238637924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5617584645748138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27640, + "weighted_flow_matching_loss": 0.01707847137004137, + "weighted_rkd_loss": 0.007796305371448398 + }, + { + "epoch": 8.044806517311608, + "flow_matching_loss": 0.01817059628665447, + "grad_norm": 0.17252853512763977, + "learning_rate": 1.815990509454506e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005386742576956749, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002171615173574537, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012945100292563438, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5614986538887023, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27650, + "weighted_flow_matching_loss": 0.01817059628665447, + "weighted_rkd_loss": 0.007268717139959335 + }, + { + "epoch": 8.047716031422752, + "flow_matching_loss": 0.018823745474219323, + "grad_norm": 0.2068101018667221, + "learning_rate": 1.8151822790865218e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005744692566804588, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002812109189108014, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01430149432271719, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5612388253211975, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27660, + "weighted_flow_matching_loss": 0.018823745474219323, + "weighted_rkd_loss": 0.008026503305882216 + }, + { + "epoch": 8.050625545533896, + "flow_matching_loss": 0.01968338629230857, + "grad_norm": 0.14476080238819122, + "learning_rate": 1.8143739529744828e-05, + "loss": 0.0284, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00638524666428566, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027913268189877273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015561820100992918, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5609790027141571, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27670, + "weighted_flow_matching_loss": 0.01968338629230857, + "weighted_rkd_loss": 0.008729776786640286 + }, + { + "epoch": 8.05353505964504, + "flow_matching_loss": 0.018125853035598993, + "grad_norm": 0.14091157913208008, + "learning_rate": 1.813565531363937e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058224840555340055, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025007133022882043, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014145681448280812, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5607191324234009, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27680, + "weighted_flow_matching_loss": 0.018125853035598993, + "weighted_rkd_loss": 0.007931828033179045 + }, + { + "epoch": 8.056444573756183, + "flow_matching_loss": 0.016484505496919155, + "grad_norm": 0.15177027881145477, + "learning_rate": 1.8127570145004628e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005229171016253531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002170522720552981, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012628864683210849, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5604592561721802, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27690, + "weighted_flow_matching_loss": 0.016484505496919155, + "weighted_rkd_loss": 0.007077995548024773 + }, + { + "epoch": 8.059354087867327, + "flow_matching_loss": 0.016907614190131426, + "grad_norm": 0.13517534732818604, + "learning_rate": 1.811948402629665e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00585612584836781, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024427293450571597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014154980983585119, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5601993799209595, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27700, + "weighted_flow_matching_loss": 0.016907614190131426, + "weighted_rkd_loss": 0.007929659821093083 + }, + { + "epoch": 8.06226360197847, + "flow_matching_loss": 0.018528841249644757, + "grad_norm": 0.1347648650407791, + "learning_rate": 1.811139695997179e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005574877234175801, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023651465657167138, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013514900952577591, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5599394500255584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27710, + "weighted_flow_matching_loss": 0.018528841249644757, + "weighted_rkd_loss": 0.0075675792060792444 + }, + { + "epoch": 8.065173116089612, + "flow_matching_loss": 0.019642400462180375, + "grad_norm": 0.13984528183937073, + "learning_rate": 1.810330894848669e-05, + "loss": 0.0271, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00551343853585422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023480193340219557, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013374896440654992, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5596795380115509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27720, + "weighted_flow_matching_loss": 0.019642400462180375, + "weighted_rkd_loss": 0.007485721027478576 + }, + { + "epoch": 8.068082630200756, + "flow_matching_loss": 0.017669375613331796, + "grad_norm": 0.16886097192764282, + "learning_rate": 1.8095219994298266e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057303785579279065, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025683608371764423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014029118046164513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.559419596195221, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27730, + "weighted_flow_matching_loss": 0.017669375613331796, + "weighted_rkd_loss": 0.007848195731639862 + }, + { + "epoch": 8.0709921443119, + "flow_matching_loss": 0.018755425233393908, + "grad_norm": 0.1622002124786377, + "learning_rate": 1.8087130099863728e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005581713374704122, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022907777572982015, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013454204238951207, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.559159642457962, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27740, + "weighted_flow_matching_loss": 0.018755425233393908, + "weighted_rkd_loss": 0.007522983010858297 + }, + { + "epoch": 8.073901658423043, + "flow_matching_loss": 0.017004831973463298, + "grad_norm": 0.15366409718990326, + "learning_rate": 1.8079039267640583e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005592918395996094, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023550822981633244, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013540919218212367, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5588996708393097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27750, + "weighted_flow_matching_loss": 0.017004831973463298, + "weighted_rkd_loss": 0.007567998068407178 + }, + { + "epoch": 8.076811172534187, + "flow_matching_loss": 0.020321018528193235, + "grad_norm": 0.15419134497642517, + "learning_rate": 1.80709475000866e-05, + "loss": 0.0283, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005802111746743321, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002617594390176237, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014221817906945943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5586396753787994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27760, + "weighted_flow_matching_loss": 0.020321018528193235, + "weighted_rkd_loss": 0.007944811461493373 + }, + { + "epoch": 8.07972068664533, + "flow_matching_loss": 0.018782352842390538, + "grad_norm": 0.17427898943424225, + "learning_rate": 1.8062854799659855e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005938986456021667, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002614066272508353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014492039289325475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5583796679973603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27770, + "weighted_flow_matching_loss": 0.018782352842390538, + "weighted_rkd_loss": 0.008092076424509287 + }, + { + "epoch": 8.082630200756475, + "flow_matching_loss": 0.02137062679976225, + "grad_norm": 0.18281987309455872, + "learning_rate": 1.805476116881869e-05, + "loss": 0.029, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005615539336577058, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002435024525038898, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01366610312834382, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5581196546554565, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27780, + "weighted_flow_matching_loss": 0.02137062679976225, + "weighted_rkd_loss": 0.00762718846090138 + }, + { + "epoch": 8.085539714867616, + "flow_matching_loss": 0.01688263099640608, + "grad_norm": 0.19299069046974182, + "learning_rate": 1.804666661002174e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059490683255717155, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025997310993261637, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014497867599129678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5578596115112304, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27790, + "weighted_flow_matching_loss": 0.01688263099640608, + "weighted_rkd_loss": 0.008087697578594089 + }, + { + "epoch": 8.08844922897876, + "flow_matching_loss": 0.017013002093881367, + "grad_norm": 0.13704021275043488, + "learning_rate": 1.8038571125727917e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005387632921338081, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002406135608907789, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013181401509791612, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5575995564460754, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27800, + "weighted_flow_matching_loss": 0.017013002093881367, + "weighted_rkd_loss": 0.007349863974377513 + }, + { + "epoch": 8.091358743089904, + "flow_matching_loss": 0.01835826048627496, + "grad_norm": 0.15832480788230896, + "learning_rate": 1.8030474718396414e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005078264186158777, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002171207615174353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01232773596420884, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5573395013809204, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27810, + "weighted_flow_matching_loss": 0.01835826048627496, + "weighted_rkd_loss": 0.006870693946257234 + }, + { + "epoch": 8.094268257201048, + "flow_matching_loss": 0.01831035166978836, + "grad_norm": 0.12348219752311707, + "learning_rate": 1.8022377390486704e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005428559938445687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025411387672647835, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013398258574306965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.557079404592514, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27820, + "weighted_flow_matching_loss": 0.01831035166978836, + "weighted_rkd_loss": 0.007463913597166538 + }, + { + "epoch": 8.097177771312191, + "flow_matching_loss": 0.01788035938516259, + "grad_norm": 0.1869650036096573, + "learning_rate": 1.8014279144458546e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00553689356893301, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002533601701725274, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013607388827949762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5568193078041077, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27830, + "weighted_flow_matching_loss": 0.01788035938516259, + "weighted_rkd_loss": 0.007576790358871222 + }, + { + "epoch": 8.100087285423335, + "flow_matching_loss": 0.019702151976525785, + "grad_norm": 0.15128043293952942, + "learning_rate": 1.800617998277197e-05, + "loss": 0.0273, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005691541731357575, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023532146122306587, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013736298028379678, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5565591931343079, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27840, + "weighted_flow_matching_loss": 0.019702151976525785, + "weighted_rkd_loss": 0.007645009923726321 + }, + { + "epoch": 8.102996799534477, + "flow_matching_loss": 0.01833704151213169, + "grad_norm": 0.15171405673027039, + "learning_rate": 1.799807990788729e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005988537659868598, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023893473437055944, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014366422593593598, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5562990605831146, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27850, + "weighted_flow_matching_loss": 0.01833704151213169, + "weighted_rkd_loss": 0.007992072077468038 + }, + { + "epoch": 8.10590631364562, + "flow_matching_loss": 0.01771839940920472, + "grad_norm": 0.16245247423648834, + "learning_rate": 1.7989978922265094e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052176931407302615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023103593150153754, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012745745666325092, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5560389161109924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27860, + "weighted_flow_matching_loss": 0.01771839940920472, + "weighted_rkd_loss": 0.007087222766131163 + }, + { + "epoch": 8.108815827756764, + "flow_matching_loss": 0.018888194113969803, + "grad_norm": 0.15443430840969086, + "learning_rate": 1.7981877028366248e-05, + "loss": 0.0272, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006105303019285202, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026565287727862595, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01486713457852602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5557787418365479, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27870, + "weighted_flow_matching_loss": 0.018888194113969803, + "weighted_rkd_loss": 0.008262913441285491 + }, + { + "epoch": 8.111725341867908, + "flow_matching_loss": 0.017503321170806885, + "grad_norm": 0.14817912876605988, + "learning_rate": 1.797377422865189e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051529652439057825, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002226668200455606, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01253259889781475, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5555185675621033, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27880, + "weighted_flow_matching_loss": 0.017503321170806885, + "weighted_rkd_loss": 0.0069620456546545025 + }, + { + "epoch": 8.114634855979052, + "flow_matching_loss": 0.017745085060596466, + "grad_norm": 0.14425009489059448, + "learning_rate": 1.796567052558344e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005733727104961872, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002446580259129405, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013914034608751536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5552583813667298, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27890, + "weighted_flow_matching_loss": 0.017745085060596466, + "weighted_rkd_loss": 0.007725891191512346 + }, + { + "epoch": 8.117544370090195, + "flow_matching_loss": 0.018065816070884467, + "grad_norm": 0.1353309452533722, + "learning_rate": 1.795756592162259e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005222270800732076, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024487491580657663, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012893290817737579, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5549981772899628, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27900, + "weighted_flow_matching_loss": 0.018065816070884467, + "weighted_rkd_loss": 0.007155857095494866 + }, + { + "epoch": 8.12045388420134, + "flow_matching_loss": 0.018669821973890065, + "grad_norm": 0.17423951625823975, + "learning_rate": 1.79494604192313e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006149715138599276, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002746721066068858, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015046151168644429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5547379553318024, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27910, + "weighted_flow_matching_loss": 0.018669821973890065, + "weighted_rkd_loss": 0.008346575032919645 + }, + { + "epoch": 8.123363398312481, + "flow_matching_loss": 0.016483797412365674, + "grad_norm": 0.17945049703121185, + "learning_rate": 1.7941354020871815e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005702129565179348, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024294940987601875, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013833753019571304, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5544777154922486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27920, + "weighted_flow_matching_loss": 0.016483797412365674, + "weighted_rkd_loss": 0.007670312700793147 + }, + { + "epoch": 8.126272912423625, + "flow_matching_loss": 0.018050482124090196, + "grad_norm": 0.16916781663894653, + "learning_rate": 1.793324672900664e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00545189967378974, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002268457692116499, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013172257039695979, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5542174577713013, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27930, + "weighted_flow_matching_loss": 0.018050482124090196, + "weighted_rkd_loss": 0.007300296844914555 + }, + { + "epoch": 8.129182426534769, + "flow_matching_loss": 0.01947151720523834, + "grad_norm": 0.15993517637252808, + "learning_rate": 1.7925138546098553e-05, + "loss": 0.027, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055655706208199264, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024215572164393963, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013552698399871587, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.553957200050354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27940, + "weighted_flow_matching_loss": 0.01947151720523834, + "weighted_rkd_loss": 0.007507579447701573 + }, + { + "epoch": 8.132091940645912, + "flow_matching_loss": 0.01797965094447136, + "grad_norm": 0.1615288257598877, + "learning_rate": 1.791702947461061e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059544707648456095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025129902292974292, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014421931840479374, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5536969125270843, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27950, + "weighted_flow_matching_loss": 0.01797965094447136, + "weighted_rkd_loss": 0.007985474402084946 + }, + { + "epoch": 8.135001454757056, + "flow_matching_loss": 0.01779366107657552, + "grad_norm": 0.15825316309928894, + "learning_rate": 1.7908919517006135e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005480962200090289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002361613849643618, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013323538191616534, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5534366130828857, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27960, + "weighted_flow_matching_loss": 0.01779366107657552, + "weighted_rkd_loss": 0.007373742247000337 + }, + { + "epoch": 8.1379109688682, + "flow_matching_loss": 0.01743338108062744, + "grad_norm": 0.15271131694316864, + "learning_rate": 1.7900808675748714e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005316748702898622, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022533992887474597, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0128868967294693, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5531763017177582, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27970, + "weighted_flow_matching_loss": 0.01743338108062744, + "weighted_rkd_loss": 0.007128829602152109 + }, + { + "epoch": 8.140820482979342, + "flow_matching_loss": 0.01883281348273158, + "grad_norm": 0.13395603001117706, + "learning_rate": 1.7892696953302214e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005113202729262412, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002245685365051031, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012472090870141983, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5529159724712371, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27980, + "weighted_flow_matching_loss": 0.01883281348273158, + "weighted_rkd_loss": 0.006896077515557408 + }, + { + "epoch": 8.143729997090485, + "flow_matching_loss": 0.019661706499755384, + "grad_norm": 0.17230984568595886, + "learning_rate": 1.7884584352130747e-05, + "loss": 0.0275, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005811812682077288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024849206558428703, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014108545798808336, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5526556372642517, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 27990, + "weighted_flow_matching_loss": 0.019661706499755384, + "weighted_rkd_loss": 0.007797052199020982 + }, + { + "epoch": 8.146639511201629, + "flow_matching_loss": 0.019314318802207707, + "grad_norm": 0.14762058854103088, + "learning_rate": 1.7876470874698722e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005271973507478833, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002210615563672036, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012754562497138976, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5523952841758728, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28000, + "weighted_flow_matching_loss": 0.019314318802207707, + "weighted_rkd_loss": 0.0070455704350024465 + }, + { + "epoch": 8.149549025312773, + "flow_matching_loss": 0.01852191165089607, + "grad_norm": 0.13856002688407898, + "learning_rate": 1.7868356523470796e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005670745484530925, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024931875872425736, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013834678567945957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5521349132061004, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28010, + "weighted_flow_matching_loss": 0.01852191165089607, + "weighted_rkd_loss": 0.007638674462214112 + }, + { + "epoch": 8.152458539423916, + "flow_matching_loss": 0.01925685312598944, + "grad_norm": 0.16464678943157196, + "learning_rate": 1.7860241300911895e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054843316320329905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024067307473160326, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013375393953174353, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5518745243549347, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28020, + "weighted_flow_matching_loss": 0.01925685312598944, + "weighted_rkd_loss": 0.007381500629708171 + }, + { + "epoch": 8.15536805353506, + "flow_matching_loss": 0.01881208214908838, + "grad_norm": 0.1724558174610138, + "learning_rate": 1.7852125209487202e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005165234860032797, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022763351444154978, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012606804817914962, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5516141295433045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28030, + "weighted_flow_matching_loss": 0.01881208214908838, + "weighted_rkd_loss": 0.006954096630215645 + }, + { + "epoch": 8.158277567646204, + "flow_matching_loss": 0.01975794481113553, + "grad_norm": 0.14070604741573334, + "learning_rate": 1.7844008251662175e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005701589654199779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023876601946540175, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01379083963111043, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5513537228107452, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28040, + "weighted_flow_matching_loss": 0.01975794481113553, + "weighted_rkd_loss": 0.007603758107870817 + }, + { + "epoch": 8.161187081757346, + "flow_matching_loss": 0.01817984776571393, + "grad_norm": 0.13312532007694244, + "learning_rate": 1.7835890429902535e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005926143564283848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026044576661661267, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014456744771450758, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5510932981967926, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28050, + "weighted_flow_matching_loss": 0.01817984776571393, + "weighted_rkd_loss": 0.007967014983296394 + }, + { + "epoch": 8.16409659586849, + "flow_matching_loss": 0.01765211671590805, + "grad_norm": 0.15652117133140564, + "learning_rate": 1.7827771746674257e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005895870551466942, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023929179180413483, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014184659346938133, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5508328378200531, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28060, + "weighted_flow_matching_loss": 0.01765211671590805, + "weighted_rkd_loss": 0.007813284639269114 + }, + { + "epoch": 8.167006109979633, + "flow_matching_loss": 0.018760336376726628, + "grad_norm": 0.16330385208129883, + "learning_rate": 1.7819652204443593e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005712294252589345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002321987459436059, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013746576104313134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5505724072456359, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28070, + "weighted_flow_matching_loss": 0.018760336376726628, + "weighted_rkd_loss": 0.007568664010614156 + }, + { + "epoch": 8.169915624090777, + "flow_matching_loss": 0.02089480198919773, + "grad_norm": 0.1686992049217224, + "learning_rate": 1.7811531805677033e-05, + "loss": 0.0278, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005110324406996369, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002291824237909168, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01251247301697731, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.550311952829361, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28080, + "weighted_flow_matching_loss": 0.02089480198919773, + "weighted_rkd_loss": 0.006885804608464241 + }, + { + "epoch": 8.17282513820192, + "flow_matching_loss": 0.017570343054831027, + "grad_norm": 0.14245367050170898, + "learning_rate": 1.7803410552841346e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005630872678011656, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002378566435072571, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013640311919152737, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5500514388084412, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28090, + "weighted_flow_matching_loss": 0.017570343054831027, + "weighted_rkd_loss": 0.007502743182703853 + }, + { + "epoch": 8.175734652313064, + "flow_matching_loss": 0.019184466172009706, + "grad_norm": 0.1775268316268921, + "learning_rate": 1.7795288448403548e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005286282626911998, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021677242475561798, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01274028941988945, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5497909486293793, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28100, + "weighted_flow_matching_loss": 0.019184466172009706, + "weighted_rkd_loss": 0.007004445884376764 + }, + { + "epoch": 8.178644166424208, + "flow_matching_loss": 0.01761251511052251, + "grad_norm": 0.1586112231016159, + "learning_rate": 1.7787165494830928e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0061512105632573364, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002694578457158059, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014996999688446522, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.549530440568924, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28110, + "weighted_flow_matching_loss": 0.01761251511052251, + "weighted_rkd_loss": 0.008241349644958974 + }, + { + "epoch": 8.18155368053535, + "flow_matching_loss": 0.018541836366057395, + "grad_norm": 0.182094544172287, + "learning_rate": 1.7779041694591024e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005993602843955159, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0028835185104981063, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01487072417512536, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5492699205875397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28120, + "weighted_flow_matching_loss": 0.018541836366057395, + "weighted_rkd_loss": 0.008168120682239533 + }, + { + "epoch": 8.184463194646494, + "flow_matching_loss": 0.020588031318038703, + "grad_norm": 0.14003780484199524, + "learning_rate": 1.7770917050151628e-05, + "loss": 0.0282, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005631858203560114, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002587407163809985, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01385112376883626, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.549009382724762, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28130, + "weighted_flow_matching_loss": 0.020588031318038703, + "weighted_rkd_loss": 0.007604336692020297 + }, + { + "epoch": 8.187372708757637, + "flow_matching_loss": 0.01806390406563878, + "grad_norm": 0.18672750890254974, + "learning_rate": 1.7762791563980795e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005733011336997151, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025107258348725735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013976748567074537, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5487488389015198, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28140, + "weighted_flow_matching_loss": 0.01806390406563878, + "weighted_rkd_loss": 0.007669674791395664 + }, + { + "epoch": 8.190282222868781, + "flow_matching_loss": 0.02093637939542532, + "grad_norm": 0.15563127398490906, + "learning_rate": 1.7754665238546828e-05, + "loss": 0.0288, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0059068446513265375, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025322120054624975, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01434590145945549, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5484882652759552, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28150, + "weighted_flow_matching_loss": 0.02093637939542532, + "weighted_rkd_loss": 0.007868660753592849 + }, + { + "epoch": 8.193191736979925, + "flow_matching_loss": 0.01892671110108495, + "grad_norm": 0.16033096611499786, + "learning_rate": 1.77465380763183e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006058169063180685, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025132067152298988, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014629544876515865, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5482276916503906, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28160, + "weighted_flow_matching_loss": 0.01892671110108495, + "weighted_rkd_loss": 0.008020327473059297 + }, + { + "epoch": 8.196101251091068, + "flow_matching_loss": 0.019796624965965747, + "grad_norm": 0.17266049981117249, + "learning_rate": 1.773841007976402e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004735764116048813, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002012935758102685, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011484464071691037, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5479671061038971, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28170, + "weighted_flow_matching_loss": 0.019796624965965747, + "weighted_rkd_loss": 0.0062931391410529615 + }, + { + "epoch": 8.19901076520221, + "flow_matching_loss": 0.01765846712514758, + "grad_norm": 0.15931856632232666, + "learning_rate": 1.773028125135306e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005671525048092008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023137743468396367, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0136568246409297, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5477065026760102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28180, + "weighted_flow_matching_loss": 0.01765846712514758, + "weighted_rkd_loss": 0.007479917304590344 + }, + { + "epoch": 8.201920279313354, + "flow_matching_loss": 0.019279824662953617, + "grad_norm": 0.17860153317451477, + "learning_rate": 1.7722151593554743e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005378661211580038, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002372277167160064, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01312959948554635, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5474458873271942, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28190, + "weighted_flow_matching_loss": 0.019279824662953617, + "weighted_rkd_loss": 0.007187625207006931 + }, + { + "epoch": 8.204829793424498, + "flow_matching_loss": 0.018069934006780385, + "grad_norm": 0.2031845897436142, + "learning_rate": 1.7714021108838652e-05, + "loss": 0.026, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005910530127584935, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026702597038820385, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014491319842636585, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5471852600574494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28200, + "weighted_flow_matching_loss": 0.018069934006780385, + "weighted_rkd_loss": 0.007929409993812442 + }, + { + "epoch": 8.207739307535642, + "flow_matching_loss": 0.01672805128619075, + "grad_norm": 0.15112340450286865, + "learning_rate": 1.77058897996746e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005714495293796062, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023003393202088774, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01372932968661189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5469246149063111, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28210, + "weighted_flow_matching_loss": 0.01672805128619075, + "weighted_rkd_loss": 0.007508916687220335 + }, + { + "epoch": 8.210648821646785, + "flow_matching_loss": 0.017600644286721943, + "grad_norm": 0.14685983955860138, + "learning_rate": 1.7697757668532677e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0049965869635343555, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021069739595986903, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012100147735327482, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5466639578342438, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28220, + "weighted_flow_matching_loss": 0.017600644286721943, + "weighted_rkd_loss": 0.006614677980542183 + }, + { + "epoch": 8.213558335757929, + "flow_matching_loss": 0.018502644170075654, + "grad_norm": 0.1508309543132782, + "learning_rate": 1.7689624717883204e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00542671037837863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022554442984983325, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013108865078538656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5464032888412476, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28230, + "weighted_flow_matching_loss": 0.018502644170075654, + "weighted_rkd_loss": 0.007162793586030602 + }, + { + "epoch": 8.216467849869073, + "flow_matching_loss": 0.018823699187487365, + "grad_norm": 0.12148908525705338, + "learning_rate": 1.768149095019676e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051061921054497365, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021637280005961657, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01237611211836338, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5461426138877868, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28240, + "weighted_flow_matching_loss": 0.018823699187487365, + "weighted_rkd_loss": 0.006759209744632244 + }, + { + "epoch": 8.219377363980215, + "flow_matching_loss": 0.016854461282491684, + "grad_norm": 0.13472281396389008, + "learning_rate": 1.7673356367944162e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004877622751519084, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002156140422448516, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011911385972052813, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5458819210529328, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28250, + "weighted_flow_matching_loss": 0.016854461282491684, + "weighted_rkd_loss": 0.006502284808084369 + }, + { + "epoch": 8.222286878091358, + "flow_matching_loss": 0.015980897657573222, + "grad_norm": 0.1409446746110916, + "learning_rate": 1.7665220973596484e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005141774844378233, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020626502227969468, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012346199993044139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5456212103366852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28260, + "weighted_flow_matching_loss": 0.015980897657573222, + "weighted_rkd_loss": 0.00673641418106854 + }, + { + "epoch": 8.225196392202502, + "flow_matching_loss": 0.017945366725325585, + "grad_norm": 0.13198821246623993, + "learning_rate": 1.7657084769625046e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005091920890845358, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002306111855432391, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01248995354399085, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5453604936599732, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28270, + "weighted_flow_matching_loss": 0.017945366725325585, + "weighted_rkd_loss": 0.00681156269274652 + }, + { + "epoch": 8.228105906313646, + "flow_matching_loss": 0.017628260515630247, + "grad_norm": 0.13897787034511566, + "learning_rate": 1.7648947758501414e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004882859322242439, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002092004183214158, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011857722979038953, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5450997650623322, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28280, + "weighted_flow_matching_loss": 0.017628260515630247, + "weighted_rkd_loss": 0.006463600602000952 + }, + { + "epoch": 8.23101542042479, + "flow_matching_loss": 0.018714917171746493, + "grad_norm": 0.11136133968830109, + "learning_rate": 1.7640809942697394e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005538347084075213, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024588693398982287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013535563368350267, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5448390245437622, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28290, + "weighted_flow_matching_loss": 0.018714917171746493, + "weighted_rkd_loss": 0.007374636549502611 + }, + { + "epoch": 8.233924934535933, + "flow_matching_loss": 0.018575326725840568, + "grad_norm": 0.12780332565307617, + "learning_rate": 1.7632671324685037e-05, + "loss": 0.0269, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006215603183954954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002824629214592278, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015255835745483637, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5445782721042634, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28300, + "weighted_flow_matching_loss": 0.018575326725840568, + "weighted_rkd_loss": 0.008308177208527922 + }, + { + "epoch": 8.236834448647077, + "flow_matching_loss": 0.01892259605228901, + "grad_norm": 0.14443810284137726, + "learning_rate": 1.7624531906936647e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0058892540168017146, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002538813534192741, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01431732140481472, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5443175137042999, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28310, + "weighted_flow_matching_loss": 0.01892259605228901, + "weighted_rkd_loss": 0.007793150842189789 + }, + { + "epoch": 8.239743962758219, + "flow_matching_loss": 0.01608937531709671, + "grad_norm": 0.11965910345315933, + "learning_rate": 1.761639169192476e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005330631393007934, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022171016316860914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01287836441770196, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5440567135810852, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28320, + "weighted_flow_matching_loss": 0.01608937531709671, + "weighted_rkd_loss": 0.007006629835814238 + }, + { + "epoch": 8.242653476869362, + "flow_matching_loss": 0.018592481035739183, + "grad_norm": 0.17933817207813263, + "learning_rate": 1.7608250682122157e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005942048504948616, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002560592337977141, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01444468917325139, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.543795931339264, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28330, + "weighted_flow_matching_loss": 0.018592481035739183, + "weighted_rkd_loss": 0.007854876900091767 + }, + { + "epoch": 8.245562990980506, + "flow_matching_loss": 0.017662258353084326, + "grad_norm": 0.1328602433204651, + "learning_rate": 1.760010888000186e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005422220379114151, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023313081357628105, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013175748754292727, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5435351252555847, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28340, + "weighted_flow_matching_loss": 0.017662258353084326, + "weighted_rkd_loss": 0.007161527359858155 + }, + { + "epoch": 8.24847250509165, + "flow_matching_loss": 0.01769960215315223, + "grad_norm": 0.16827373206615448, + "learning_rate": 1.7591966288037142e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004844316863454878, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020620646537281573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01175069846212864, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.543274313211441, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28350, + "weighted_flow_matching_loss": 0.01769960215315223, + "weighted_rkd_loss": 0.006383761949837208 + }, + { + "epoch": 8.251382019202794, + "flow_matching_loss": 0.017927603889256717, + "grad_norm": 0.14640094339847565, + "learning_rate": 1.75838229087015e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005537227587774396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002296432573348284, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013370887842029333, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5430134892463684, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28360, + "weighted_flow_matching_loss": 0.017927603889256717, + "weighted_rkd_loss": 0.007260522106662393 + }, + { + "epoch": 8.254291533313937, + "flow_matching_loss": 0.017341568134725094, + "grad_norm": 0.12966853380203247, + "learning_rate": 1.7575678744468684e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005965153127908707, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025118020363152026, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014442108292132615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5427526473999024, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28370, + "weighted_flow_matching_loss": 0.017341568134725094, + "weighted_rkd_loss": 0.007838409580290317 + }, + { + "epoch": 8.25720104742508, + "flow_matching_loss": 0.019173252303153275, + "grad_norm": 0.15546543896198273, + "learning_rate": 1.756753379781267e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005385816283524037, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002522651304025203, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013294283952564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5424917936325073, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28380, + "weighted_flow_matching_loss": 0.019173252303153275, + "weighted_rkd_loss": 0.00721204555593431 + }, + { + "epoch": 8.260110561536223, + "flow_matching_loss": 0.018043152149766685, + "grad_norm": 0.17763738334178925, + "learning_rate": 1.7559388071207682e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005574048543348908, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024053176632151006, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01355341486632824, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5422309339046478, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28390, + "weighted_flow_matching_loss": 0.018043152149766685, + "weighted_rkd_loss": 0.007348966179415583 + }, + { + "epoch": 8.263020075647367, + "flow_matching_loss": 0.017486599460244178, + "grad_norm": 0.15423491597175598, + "learning_rate": 1.7551241567128173e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00549119277857244, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002214767993427813, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013197153620421887, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5419700622558594, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28400, + "weighted_flow_matching_loss": 0.017486599460244178, + "weighted_rkd_loss": 0.007152406824752688 + }, + { + "epoch": 8.26592958975851, + "flow_matching_loss": 0.0192324654199183, + "grad_norm": 0.18565082550048828, + "learning_rate": 1.7543094288048836e-05, + "loss": 0.0274, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006170743051916361, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0027254466549493372, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015066932793706656, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5417091727256775, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28410, + "weighted_flow_matching_loss": 0.0192324654199183, + "weighted_rkd_loss": 0.008161944663152098 + }, + { + "epoch": 8.268839103869654, + "flow_matching_loss": 0.017213054560124876, + "grad_norm": 0.1504427045583725, + "learning_rate": 1.7534946236444602e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004954738402739167, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021766503574326634, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012086127139627933, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5414482772350311, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28420, + "weighted_flow_matching_loss": 0.017213054560124876, + "weighted_rkd_loss": 0.006544056721031666 + }, + { + "epoch": 8.271748617980798, + "flow_matching_loss": 0.017090861964970828, + "grad_norm": 0.14611780643463135, + "learning_rate": 1.7526797414790633e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005278521962463856, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002364107046741992, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01292115096002817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5411873638629914, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28430, + "weighted_flow_matching_loss": 0.017090861964970828, + "weighted_rkd_loss": 0.00699269687756896 + }, + { + "epoch": 8.274658132091941, + "flow_matching_loss": 0.018161735963076353, + "grad_norm": 0.16649673879146576, + "learning_rate": 1.751864782556233e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006023782561533153, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026400695787742734, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014687634631991386, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5409264445304871, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28440, + "weighted_flow_matching_loss": 0.018161735963076353, + "weighted_rkd_loss": 0.0079448944889009 + }, + { + "epoch": 8.277567646203083, + "flow_matching_loss": 0.017797357216477395, + "grad_norm": 0.12833812832832336, + "learning_rate": 1.7510497471235314e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005562627059407532, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002429536287672818, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01355479033663869, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5406655132770538, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28450, + "weighted_flow_matching_loss": 0.017797357216477395, + "weighted_rkd_loss": 0.007328656688332558 + }, + { + "epoch": 8.280477160314227, + "flow_matching_loss": 0.01856513125821948, + "grad_norm": 0.1740860939025879, + "learning_rate": 1.7502346354285455e-05, + "loss": 0.0266, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060403968673199415, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002755418047308922, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01483621196821332, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5404045760631562, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28460, + "weighted_flow_matching_loss": 0.01856513125821948, + "weighted_rkd_loss": 0.0080176190007478 + }, + { + "epoch": 8.28338667442537, + "flow_matching_loss": 0.018118360731750726, + "grad_norm": 0.19071970880031586, + "learning_rate": 1.7494194477188838e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005056081898510456, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020420412882231178, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012154205050319434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.540143620967865, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28470, + "weighted_flow_matching_loss": 0.018118360731750726, + "weighted_rkd_loss": 0.006564914388582111 + }, + { + "epoch": 8.286296188536515, + "flow_matching_loss": 0.017440115846693516, + "grad_norm": 0.13904842734336853, + "learning_rate": 1.74860418424218e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005817355588078499, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025468967971391974, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014181607961654663, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5398826599121094, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28480, + "weighted_flow_matching_loss": 0.017440115846693516, + "weighted_rkd_loss": 0.007656399486586451 + }, + { + "epoch": 8.289205702647658, + "flow_matching_loss": 0.018307030014693738, + "grad_norm": 0.17453151941299438, + "learning_rate": 1.7477888452460885e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052722831256687645, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021532770479097962, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012697843462228775, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5396216809749603, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28490, + "weighted_flow_matching_loss": 0.018307030014693738, + "weighted_rkd_loss": 0.00685193189419806 + }, + { + "epoch": 8.292115216758802, + "flow_matching_loss": 0.01709258509799838, + "grad_norm": 0.14052030444145203, + "learning_rate": 1.746973430978289e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005388972163200378, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022910682018846273, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013069012574851513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5393606901168824, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28500, + "weighted_flow_matching_loss": 0.01709258509799838, + "weighted_rkd_loss": 0.007048869039863348 + }, + { + "epoch": 8.295024730869944, + "flow_matching_loss": 0.017045955266803504, + "grad_norm": 0.16987714171409607, + "learning_rate": 1.7461579416864814e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005492384429089725, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022791844792664053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013263953477144241, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5390997052192688, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28510, + "weighted_flow_matching_loss": 0.017045955266803504, + "weighted_rkd_loss": 0.00715054371394217 + }, + { + "epoch": 8.297934244981088, + "flow_matching_loss": 0.01870405040681362, + "grad_norm": 0.14459393918514252, + "learning_rate": 1.7453423776183907e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005315122893080115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002388841041829437, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013019086793065071, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5388386905193329, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28520, + "weighted_flow_matching_loss": 0.01870405040681362, + "weighted_rkd_loss": 0.00701514845713973 + }, + { + "epoch": 8.300843759092231, + "flow_matching_loss": 0.01748373033478856, + "grad_norm": 0.13097330927848816, + "learning_rate": 1.744526739021764e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005483200447633863, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002184359461534768, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013150760252028704, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.538577675819397, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28530, + "weighted_flow_matching_loss": 0.01748373033478856, + "weighted_rkd_loss": 0.007082667201757431 + }, + { + "epoch": 8.303753273203375, + "flow_matching_loss": 0.017641987279057504, + "grad_norm": 0.15572942793369293, + "learning_rate": 1.74371102614437e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005208878521807491, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023025481961667536, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012720305193215608, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5383166551589966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28540, + "weighted_flow_matching_loss": 0.017641987279057504, + "weighted_rkd_loss": 0.0068475714884698394 + }, + { + "epoch": 8.306662787314519, + "flow_matching_loss": 0.019323653541505337, + "grad_norm": 0.16534344851970673, + "learning_rate": 1.742895239234001e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005532615561969578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023436152376234533, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013408846408128738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5380555987358093, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28550, + "weighted_flow_matching_loss": 0.019323653541505337, + "weighted_rkd_loss": 0.00721465521492064 + }, + { + "epoch": 8.309572301425662, + "flow_matching_loss": 0.01849910169839859, + "grad_norm": 0.15938957035541534, + "learning_rate": 1.742079378538472e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055645288899540905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023699272191151977, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013498985208570957, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5377945482730866, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28560, + "weighted_flow_matching_loss": 0.01849910169839859, + "weighted_rkd_loss": 0.007259561587125063 + }, + { + "epoch": 8.312481815536806, + "flow_matching_loss": 0.018678080476820468, + "grad_norm": 0.17211319506168365, + "learning_rate": 1.7412634443056192e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005813883151859045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025684151332825422, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014196181669831276, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5375335216522217, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28570, + "weighted_flow_matching_loss": 0.018678080476820468, + "weighted_rkd_loss": 0.00763081219047308 + }, + { + "epoch": 8.315391329647948, + "flow_matching_loss": 0.018375434912741184, + "grad_norm": 0.146916002035141, + "learning_rate": 1.7404474367833024e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006079678423702717, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002635577879846096, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014794934913516045, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5372724533081055, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28580, + "weighted_flow_matching_loss": 0.018375434912741184, + "weighted_rkd_loss": 0.007948777265846729 + }, + { + "epoch": 8.318300843759092, + "flow_matching_loss": 0.01806908929720521, + "grad_norm": 0.1667270064353943, + "learning_rate": 1.739631356219403e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005506563698872924, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023015141603536906, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013314641453325749, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5370113253593445, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28590, + "weighted_flow_matching_loss": 0.01806908929720521, + "weighted_rkd_loss": 0.007150087738409638 + }, + { + "epoch": 8.321210357870235, + "flow_matching_loss": 0.018021164555102588, + "grad_norm": 0.13854651153087616, + "learning_rate": 1.7388152028618244e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005034510465338826, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002106945402920246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01217596624046564, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5367502570152283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28600, + "weighted_flow_matching_loss": 0.018021164555102588, + "weighted_rkd_loss": 0.0065354571212083105 + }, + { + "epoch": 8.32411987198138, + "flow_matching_loss": 0.018327942676842214, + "grad_norm": 0.14353232085704803, + "learning_rate": 1.737998976958493e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005332385003566742, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002438408264424652, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01310317823663354, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5364891529083252, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28610, + "weighted_flow_matching_loss": 0.018327942676842214, + "weighted_rkd_loss": 0.0070296102203428745 + }, + { + "epoch": 8.327029386092523, + "flow_matching_loss": 0.015916460659354925, + "grad_norm": 0.15625643730163574, + "learning_rate": 1.7371826787573562e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057656255317851905, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024229408241808414, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013954191841185093, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5362280488014222, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28620, + "weighted_flow_matching_loss": 0.015916460659354925, + "weighted_rkd_loss": 0.007482715090736747 + }, + { + "epoch": 8.329938900203667, + "flow_matching_loss": 0.018636116292327642, + "grad_norm": 0.1384250521659851, + "learning_rate": 1.736366308506384e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005053454544395208, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021337181446142496, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012240627221763134, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5359669268131256, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28630, + "weighted_flow_matching_loss": 0.018636116292327642, + "weighted_rkd_loss": 0.006560580804944038 + }, + { + "epoch": 8.33284841431481, + "flow_matching_loss": 0.018515570275485515, + "grad_norm": 0.1373405009508133, + "learning_rate": 1.7355498664535682e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005334201641380787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002330188569612801, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012998591735959053, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5357057929039002, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28640, + "weighted_flow_matching_loss": 0.018515570275485515, + "weighted_rkd_loss": 0.006963426247239113 + }, + { + "epoch": 8.335757928425952, + "flow_matching_loss": 0.019481811113655566, + "grad_norm": 0.15096528828144073, + "learning_rate": 1.7347333528469225e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004506700369529426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.001975205191411078, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.010988606046885252, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5354446530342102, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28650, + "weighted_flow_matching_loss": 0.019481811113655566, + "weighted_rkd_loss": 0.005883778445422649 + }, + { + "epoch": 8.338667442537096, + "flow_matching_loss": 0.018851420003920795, + "grad_norm": 0.14966876804828644, + "learning_rate": 1.733916767934482e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005293494882062077, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022168046096339824, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012803794350475073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5351835072040558, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28660, + "weighted_flow_matching_loss": 0.018851420003920795, + "weighted_rkd_loss": 0.006852315785363316 + }, + { + "epoch": 8.34157695664824, + "flow_matching_loss": 0.01839433591812849, + "grad_norm": 0.15752604603767395, + "learning_rate": 1.7331001119643034e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00574185042642057, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024023089557886125, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013886009901762008, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5349223375320434, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28670, + "weighted_flow_matching_loss": 0.01839433591812849, + "weighted_rkd_loss": 0.007428015721961856 + }, + { + "epoch": 8.344486470759383, + "flow_matching_loss": 0.01938432492315769, + "grad_norm": 0.15789580345153809, + "learning_rate": 1.7322833851844653e-05, + "loss": 0.0263, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005251877312548459, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002385056484490633, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012888811063021421, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5346611738204956, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28680, + "weighted_flow_matching_loss": 0.01938432492315769, + "weighted_rkd_loss": 0.006891215639188886 + }, + { + "epoch": 8.347395984870527, + "flow_matching_loss": 0.017452105507254602, + "grad_norm": 0.15059256553649902, + "learning_rate": 1.731466587843068e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005063460953533649, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022057474008761345, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012332669366151095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5343999922275543, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28690, + "weighted_flow_matching_loss": 0.017452105507254602, + "weighted_rkd_loss": 0.006590651627629995 + }, + { + "epoch": 8.35030549898167, + "flow_matching_loss": 0.01839852211996913, + "grad_norm": 0.14602626860141754, + "learning_rate": 1.7306497201882334e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00555762923322618, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002438142686150968, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013553401362150908, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5341388046741485, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28700, + "weighted_flow_matching_loss": 0.01839852211996913, + "weighted_rkd_loss": 0.007239366136491299 + }, + { + "epoch": 8.353215013092813, + "flow_matching_loss": 0.018268846441060305, + "grad_norm": 0.14614860713481903, + "learning_rate": 1.7298327824681032e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005257867369800806, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022119403583928943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012727675028145313, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5338776111602783, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28710, + "weighted_flow_matching_loss": 0.018268846441060305, + "weighted_rkd_loss": 0.0067949641030281786 + }, + { + "epoch": 8.356124527203956, + "flow_matching_loss": 0.016883081384003164, + "grad_norm": 0.16330018639564514, + "learning_rate": 1.729015774930842e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057659693993628025, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024523438303731384, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013984282594174146, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5336164116859436, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28720, + "weighted_flow_matching_loss": 0.016883081384003164, + "weighted_rkd_loss": 0.007462309952825308 + }, + { + "epoch": 8.3590340413151, + "flow_matching_loss": 0.01805162550881505, + "grad_norm": 0.17637225985527039, + "learning_rate": 1.7281986978246353e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005535658495500684, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023137270472943783, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013385044038295746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5333551883697509, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28730, + "weighted_flow_matching_loss": 0.01805162550881505, + "weighted_rkd_loss": 0.007138940971344709 + }, + { + "epoch": 8.361943555426244, + "flow_matching_loss": 0.017495205719023944, + "grad_norm": 0.1473555713891983, + "learning_rate": 1.7273815513976894e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005106620141305029, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021239158581011, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012337156245484949, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5330939650535583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28740, + "weighted_flow_matching_loss": 0.017495205719023944, + "weighted_rkd_loss": 0.006576896063052118 + }, + { + "epoch": 8.364853069537387, + "flow_matching_loss": 0.019343974068760872, + "grad_norm": 0.20138372480869293, + "learning_rate": 1.7265643358982318e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00556698446162045, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022647281992249193, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013398697134107351, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5328327298164368, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28750, + "weighted_flow_matching_loss": 0.019343974068760872, + "weighted_rkd_loss": 0.007139215245842934 + }, + { + "epoch": 8.367762583648531, + "flow_matching_loss": 0.016685434617102146, + "grad_norm": 0.14806559681892395, + "learning_rate": 1.7257470515745117e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006327710766345263, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002798857935704291, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015454279724508524, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5325714826583863, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28760, + "weighted_flow_matching_loss": 0.016685434617102146, + "weighted_rkd_loss": 0.00823056916706264 + }, + { + "epoch": 8.370672097759675, + "flow_matching_loss": 0.01690856283530593, + "grad_norm": 0.1403331160545349, + "learning_rate": 1.724929698674797e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005195441003888845, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021990379435010256, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0125899201259017, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5323102355003357, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28770, + "weighted_flow_matching_loss": 0.01690856283530593, + "weighted_rkd_loss": 0.006701765628531575 + }, + { + "epoch": 8.373581611870817, + "flow_matching_loss": 0.017763386573642494, + "grad_norm": 0.14196759462356567, + "learning_rate": 1.7241122774473788e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005593712301924824, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023935319972224534, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013580956682562827, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5320489704608917, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28780, + "weighted_flow_matching_loss": 0.017763386573642494, + "weighted_rkd_loss": 0.007225742191076278 + }, + { + "epoch": 8.37649112598196, + "flow_matching_loss": 0.017915335297584534, + "grad_norm": 0.16744346916675568, + "learning_rate": 1.7232947881405684e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005398342572152615, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023684654384851455, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013165150582790375, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5317877054214477, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28790, + "weighted_flow_matching_loss": 0.017915335297584534, + "weighted_rkd_loss": 0.00700098043307662 + }, + { + "epoch": 8.379400640093104, + "flow_matching_loss": 0.016601168736815452, + "grad_norm": 0.14260564744472504, + "learning_rate": 1.722477231002697e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005504610016942024, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002110850589815527, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0131200703792274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5315264225006103, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28800, + "weighted_flow_matching_loss": 0.016601168736815452, + "weighted_rkd_loss": 0.006973687652498483 + }, + { + "epoch": 8.382310154204248, + "flow_matching_loss": 0.017885793559253217, + "grad_norm": 0.13340122997760773, + "learning_rate": 1.721659606282117e-05, + "loss": 0.0257, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006106545217335224, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025877795997075737, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014800870046019553, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.531265139579773, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28810, + "weighted_flow_matching_loss": 0.017885793559253217, + "weighted_rkd_loss": 0.007863119523972274 + }, + { + "epoch": 8.385219668315392, + "flow_matching_loss": 0.018100434076040985, + "grad_norm": 0.14656855165958405, + "learning_rate": 1.7208419142272012e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00549245229922235, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.00241539393318817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013400298357009888, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5310038447380065, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28820, + "weighted_flow_matching_loss": 0.018100434076040985, + "weighted_rkd_loss": 0.007115718955174088 + }, + { + "epoch": 8.388129182426535, + "flow_matching_loss": 0.015683444682508707, + "grad_norm": 0.12398482859134674, + "learning_rate": 1.720024155086343e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005787245510146022, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002339360711630434, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013913851696997881, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5307425379753112, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28830, + "weighted_flow_matching_loss": 0.015683444682508707, + "weighted_rkd_loss": 0.007384662190452218 + }, + { + "epoch": 8.391038696537677, + "flow_matching_loss": 0.017946261540055276, + "grad_norm": 0.15383277833461761, + "learning_rate": 1.7192063291079555e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005852848663926125, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002411861123982817, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01411755857989192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5304812252521515, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28840, + "weighted_flow_matching_loss": 0.017946261540055276, + "weighted_rkd_loss": 0.007489058654755354 + }, + { + "epoch": 8.393948210648821, + "flow_matching_loss": 0.0175225667655468, + "grad_norm": 0.1620333194732666, + "learning_rate": 1.7183884365404738e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00524941913317889, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002257873257622123, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012756711710244418, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5302199065685272, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28850, + "weighted_flow_matching_loss": 0.0175225667655468, + "weighted_rkd_loss": 0.006763895833864808 + }, + { + "epoch": 8.396857724759965, + "flow_matching_loss": 0.016358553897589445, + "grad_norm": 0.1394021362066269, + "learning_rate": 1.7175704776323504e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005222284351475537, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022433064063079656, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012687875237315894, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5299585700035095, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28860, + "weighted_flow_matching_loss": 0.016358553897589445, + "weighted_rkd_loss": 0.006724090455099941 + }, + { + "epoch": 8.399767238871108, + "flow_matching_loss": 0.01871710065752268, + "grad_norm": 0.16654133796691895, + "learning_rate": 1.7167524526320612e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005371294636279345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022383625386282802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012980951648205518, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5296972393989563, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28870, + "weighted_flow_matching_loss": 0.01871710065752268, + "weighted_rkd_loss": 0.006875943299382925 + }, + { + "epoch": 8.402676752982252, + "flow_matching_loss": 0.016346711479127408, + "grad_norm": 0.14819689095020294, + "learning_rate": 1.7159343617881e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005457684444263577, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002279577439185232, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01319494629278779, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5294358909130097, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28880, + "weighted_flow_matching_loss": 0.016346711479127408, + "weighted_rkd_loss": 0.0069858727511018515 + }, + { + "epoch": 8.405586267093396, + "flow_matching_loss": 0.018146944511681796, + "grad_norm": 0.15030115842819214, + "learning_rate": 1.71511620534898e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005872429558075964, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026524073793552814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014397266414016485, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5291745364665985, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28890, + "weighted_flow_matching_loss": 0.018146944511681796, + "weighted_rkd_loss": 0.007618577731773257 + }, + { + "epoch": 8.40849578120454, + "flow_matching_loss": 0.01804948868229985, + "grad_norm": 0.13419181108474731, + "learning_rate": 1.7142979835632376e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00621129646897316, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002780520461965352, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015203113574534655, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5289131700992584, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28900, + "weighted_flow_matching_loss": 0.01804948868229985, + "weighted_rkd_loss": 0.008041094429790974 + }, + { + "epoch": 8.411405295315681, + "flow_matching_loss": 0.015773747488856314, + "grad_norm": 0.16549058258533478, + "learning_rate": 1.7134796966794254e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005546209844760596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023638554266653955, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013456275314092636, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5286518037319183, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28910, + "weighted_flow_matching_loss": 0.015773747488856314, + "weighted_rkd_loss": 0.0071136432234197855 + }, + { + "epoch": 8.414314809426825, + "flow_matching_loss": 0.018205551244318486, + "grad_norm": 0.1284068375825882, + "learning_rate": 1.7126613449461186e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00512586859986186, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020828265929594635, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012334563862532377, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5283904373645782, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28920, + "weighted_flow_matching_loss": 0.018205551244318486, + "weighted_rkd_loss": 0.006517468811944127 + }, + { + "epoch": 8.417224323537969, + "flow_matching_loss": 0.017600841727107762, + "grad_norm": 0.15833412110805511, + "learning_rate": 1.7118429286119098e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00553180011920631, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023206750629469752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013384275138378143, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5281290411949158, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28930, + "weighted_flow_matching_loss": 0.017600841727107762, + "weighted_rkd_loss": 0.007068550074473023 + }, + { + "epoch": 8.420133837649113, + "flow_matching_loss": 0.01682548439130187, + "grad_norm": 0.1393669694662094, + "learning_rate": 1.7110244479254126e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005537864752113819, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022465199464932082, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013322249427437783, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5278676509857178, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28940, + "weighted_flow_matching_loss": 0.01682548439130187, + "weighted_rkd_loss": 0.007032381184399128 + }, + { + "epoch": 8.423043351760256, + "flow_matching_loss": 0.017011613491922616, + "grad_norm": 0.16566741466522217, + "learning_rate": 1.71020590313526e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005718113295733929, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002362838410772383, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013799065072089434, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5276062488555908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28950, + "weighted_flow_matching_loss": 0.017011613491922616, + "weighted_rkd_loss": 0.007280422234907746 + }, + { + "epoch": 8.4259528658714, + "flow_matching_loss": 0.017449134029448032, + "grad_norm": 0.1786823570728302, + "learning_rate": 1.7093872944901044e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00593115477822721, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002526059118099511, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014388368558138609, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5273448407649994, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28960, + "weighted_flow_matching_loss": 0.017449134029448032, + "weighted_rkd_loss": 0.007587532978504896 + }, + { + "epoch": 8.428862379982544, + "flow_matching_loss": 0.018228058330714704, + "grad_norm": 0.12367717176675797, + "learning_rate": 1.7085686222386172e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005228719068691134, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021181610762141647, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012575599271804095, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5270834267139435, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28970, + "weighted_flow_matching_loss": 0.018228058330714704, + "weighted_rkd_loss": 0.006628392683342099 + }, + { + "epoch": 8.431771894093686, + "flow_matching_loss": 0.01635031895712018, + "grad_norm": 0.1224222183227539, + "learning_rate": 1.7077498866294895e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005885219015181064, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024812669842503967, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014251704886555671, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5268220007419586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28980, + "weighted_flow_matching_loss": 0.01635031895712018, + "weighted_rkd_loss": 0.007508278405293822 + }, + { + "epoch": 8.43468140820483, + "flow_matching_loss": 0.017190948221832515, + "grad_norm": 0.16106492280960083, + "learning_rate": 1.7069310879114313e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005465241009369492, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022853640373796226, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013215846102684736, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5265605747699738, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 28990, + "weighted_flow_matching_loss": 0.017190948221832515, + "weighted_rkd_loss": 0.006958965724334121 + }, + { + "epoch": 8.437590922315973, + "flow_matching_loss": 0.019304662570357324, + "grad_norm": 0.17039871215820312, + "learning_rate": 1.7061122263331727e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005577739700675011, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023082297760993243, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013463709224015474, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5262991309165954, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29000, + "weighted_flow_matching_loss": 0.019304662570357324, + "weighted_rkd_loss": 0.007086002500727773 + }, + { + "epoch": 8.440500436427117, + "flow_matching_loss": 0.018287589028477668, + "grad_norm": 0.12906603515148163, + "learning_rate": 1.7052933021434617e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005342049524188042, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024735559476539493, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013157655019313097, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5260376930236816, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29010, + "weighted_flow_matching_loss": 0.018287589028477668, + "weighted_rkd_loss": 0.006921420805156231 + }, + { + "epoch": 8.44340995053826, + "flow_matching_loss": 0.018052821792662144, + "grad_norm": 0.1694699227809906, + "learning_rate": 1.7044743155910658e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005427856161259115, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002296269161161035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013151981495320797, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5257762372493744, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29020, + "weighted_flow_matching_loss": 0.018052821792662144, + "weighted_rkd_loss": 0.006915087578818202 + }, + { + "epoch": 8.446319464649404, + "flow_matching_loss": 0.016796760261058807, + "grad_norm": 0.1292317807674408, + "learning_rate": 1.703655266924772e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005359846074134112, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022923743003048004, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01301206648349762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5255147814750671, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29030, + "weighted_flow_matching_loss": 0.016796760261058807, + "weighted_rkd_loss": 0.006838119868189096 + }, + { + "epoch": 8.449228978760548, + "flow_matching_loss": 0.017701750341802835, + "grad_norm": 0.12750419974327087, + "learning_rate": 1.7028361563933846e-05, + "loss": 0.0249, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005684410966932773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002420036995317787, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013788858894258738, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.525253313779831, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29040, + "weighted_flow_matching_loss": 0.017701750341802835, + "weighted_rkd_loss": 0.007242667349055409 + }, + { + "epoch": 8.45213849287169, + "flow_matching_loss": 0.017371195275336503, + "grad_norm": 0.13946644961833954, + "learning_rate": 1.702016984245729e-05, + "loss": 0.024, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005213227379135788, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002170047035906464, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012596501875668764, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5249918401241302, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29050, + "weighted_flow_matching_loss": 0.017371195275336503, + "weighted_rkd_loss": 0.0066131643485277895 + }, + { + "epoch": 8.455048006982834, + "flow_matching_loss": 0.016322309616953135, + "grad_norm": 0.14527690410614014, + "learning_rate": 1.701197750730647e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0050582301802933214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002148089453112334, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012264549825340509, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5247303605079651, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29060, + "weighted_flow_matching_loss": 0.016322309616953135, + "weighted_rkd_loss": 0.006435547117143869 + }, + { + "epoch": 8.457957521093977, + "flow_matching_loss": 0.01856791591271758, + "grad_norm": 0.16979995369911194, + "learning_rate": 1.7003784560970013e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051216983934864405, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002196715783793479, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012440112605690955, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5244688749313354, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29070, + "weighted_flow_matching_loss": 0.01856791591271758, + "weighted_rkd_loss": 0.006524517387151718 + }, + { + "epoch": 8.460867035205121, + "flow_matching_loss": 0.019235055055469276, + "grad_norm": 0.14293785393238068, + "learning_rate": 1.6995591005936714e-05, + "loss": 0.0262, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005481254775077105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023337404010817407, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013296249881386757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5242073833942413, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29080, + "weighted_flow_matching_loss": 0.019235055055469276, + "weighted_rkd_loss": 0.00697011393494904 + }, + { + "epoch": 8.463776549316265, + "flow_matching_loss": 0.017201070580631493, + "grad_norm": 0.13509288430213928, + "learning_rate": 1.6987396844695555e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005758233647793531, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002229586453177035, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013746053725481034, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5239458799362182, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29090, + "weighted_flow_matching_loss": 0.017201070580631493, + "weighted_rkd_loss": 0.0072021727915853265 + }, + { + "epoch": 8.466686063427408, + "flow_matching_loss": 0.01693038297817111, + "grad_norm": 0.1552550047636032, + "learning_rate": 1.6979202079735708e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005457927333191037, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002284330071415752, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01320018470287323, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5236843705177308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29100, + "weighted_flow_matching_loss": 0.01693038297817111, + "weighted_rkd_loss": 0.006912661530077457 + }, + { + "epoch": 8.46959557753855, + "flow_matching_loss": 0.015605085901916027, + "grad_norm": 0.16722050309181213, + "learning_rate": 1.697100671354653e-05, + "loss": 0.0233, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006005060812458396, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002647354395594448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0146574760787189, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5234228670597076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29110, + "weighted_flow_matching_loss": 0.015605085901916027, + "weighted_rkd_loss": 0.007672002073377371 + }, + { + "epoch": 8.472505091649694, + "flow_matching_loss": 0.019099933840334416, + "grad_norm": 0.14148572087287903, + "learning_rate": 1.6962810748617548e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005132556683383882, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021702843136154116, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012435397785156965, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5231613337993621, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29120, + "weighted_flow_matching_loss": 0.019099933840334416, + "weighted_rkd_loss": 0.006505747279152274 + }, + { + "epoch": 8.475414605760838, + "flow_matching_loss": 0.018269923515617847, + "grad_norm": 0.13963578641414642, + "learning_rate": 1.695461418743849e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0054974313359707596, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022617727518081665, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01325663523748517, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5228998184204101, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29130, + "weighted_flow_matching_loss": 0.018269923515617847, + "weighted_rkd_loss": 0.006931900978088379 + }, + { + "epoch": 8.478324119871981, + "flow_matching_loss": 0.018172085657715796, + "grad_norm": 0.15426263213157654, + "learning_rate": 1.6946417032499252e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00564453424885869, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024163992609828712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01370546780526638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5226382851600647, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29140, + "weighted_flow_matching_loss": 0.018172085657715796, + "weighted_rkd_loss": 0.007162957498803735 + }, + { + "epoch": 8.481233633983125, + "flow_matching_loss": 0.017976687289774418, + "grad_norm": 0.12753430008888245, + "learning_rate": 1.6938219286289908e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005068359989672899, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022900487296283246, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012426768895238638, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5223767399787903, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29150, + "weighted_flow_matching_loss": 0.017976687289774418, + "weighted_rkd_loss": 0.006491498090326786 + }, + { + "epoch": 8.484143148094269, + "flow_matching_loss": 0.017707299999892713, + "grad_norm": 0.11975529044866562, + "learning_rate": 1.693002095130072e-05, + "loss": 0.0258, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006412539654411376, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002740513440221548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.015565592888742685, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5221151947975159, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29160, + "weighted_flow_matching_loss": 0.017707299999892713, + "weighted_rkd_loss": 0.008127049589529634 + }, + { + "epoch": 8.487052662205413, + "flow_matching_loss": 0.016710403747856616, + "grad_norm": 0.13435962796211243, + "learning_rate": 1.6921822030022132e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0050897769629955295, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002127408760134131, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012306962627917528, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5218536496162415, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29170, + "weighted_flow_matching_loss": 0.016710403747856616, + "weighted_rkd_loss": 0.006422344129532576 + }, + { + "epoch": 8.489962176316554, + "flow_matching_loss": 0.016765731014311315, + "grad_norm": 0.1549612283706665, + "learning_rate": 1.691362252494475e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0055394134484231475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024047151673585176, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013483542017638683, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5215920865535736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29180, + "weighted_flow_matching_loss": 0.016765731014311315, + "weighted_rkd_loss": 0.00703293108381331 + }, + { + "epoch": 8.492871690427698, + "flow_matching_loss": 0.01664173062890768, + "grad_norm": 0.1457185298204422, + "learning_rate": 1.690542243855937e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005525603471323848, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002457617549225688, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013508824538439513, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5213305294513703, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29190, + "weighted_flow_matching_loss": 0.01664173062890768, + "weighted_rkd_loss": 0.007042481028474867 + }, + { + "epoch": 8.495781204538842, + "flow_matching_loss": 0.01743197292089462, + "grad_norm": 0.13031888008117676, + "learning_rate": 1.6897221773356963e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005073536839336157, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019975609029643238, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01214463459327817, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5210689604282379, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29200, + "weighted_flow_matching_loss": 0.01743197292089462, + "weighted_rkd_loss": 0.006328174797818065 + }, + { + "epoch": 8.498690718649986, + "flow_matching_loss": 0.0179687705822289, + "grad_norm": 0.15731093287467957, + "learning_rate": 1.6889020531828676e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052162489853799345, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022733939928002655, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012705891951918602, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5208073854446411, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29210, + "weighted_flow_matching_loss": 0.0179687705822289, + "weighted_rkd_loss": 0.006617274158634246 + }, + { + "epoch": 8.50160023276113, + "flow_matching_loss": 0.01723323715850711, + "grad_norm": 0.15091882646083832, + "learning_rate": 1.688081871646583e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00557273724116385, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002463033376261592, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01360850790515542, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5205458045005799, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29220, + "weighted_flow_matching_loss": 0.01723323715850711, + "weighted_rkd_loss": 0.007083966024219989 + }, + { + "epoch": 8.504509746872273, + "flow_matching_loss": 0.017109508253633977, + "grad_norm": 0.12977993488311768, + "learning_rate": 1.6872616329759917e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005344223557040095, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002113963826559484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012802410963922739, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5202842235565186, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29230, + "weighted_flow_matching_loss": 0.017109508253633977, + "weighted_rkd_loss": 0.0066609310451895 + }, + { + "epoch": 8.507419260983415, + "flow_matching_loss": 0.01870080903172493, + "grad_norm": 0.1510881930589676, + "learning_rate": 1.6864413374202605e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005192783195525408, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002183430304285139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012568996753543615, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5200226306915283, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29240, + "weighted_flow_matching_loss": 0.01870080903172493, + "weighted_rkd_loss": 0.006536235613748431 + }, + { + "epoch": 8.510328775094559, + "flow_matching_loss": 0.017997883912175893, + "grad_norm": 0.12468370795249939, + "learning_rate": 1.685620985228574e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005595677811652422, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023339818697422745, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013525337539613247, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5197610318660736, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29250, + "weighted_flow_matching_loss": 0.017997883912175893, + "weighted_rkd_loss": 0.007029981259256602 + }, + { + "epoch": 8.513238289205702, + "flow_matching_loss": 0.019721895270049573, + "grad_norm": 0.13463227450847626, + "learning_rate": 1.6848005766501326e-05, + "loss": 0.0267, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005587565060704946, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002342844964005053, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01351797515526414, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5194994330406189, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29260, + "weighted_flow_matching_loss": 0.019721895270049573, + "weighted_rkd_loss": 0.007022728538140654 + }, + { + "epoch": 8.516147803316846, + "flow_matching_loss": 0.01610722616314888, + "grad_norm": 0.14491532742977142, + "learning_rate": 1.683980111934156e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005304523464292288, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002234984515234828, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012844031397253274, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5192378282546997, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29270, + "weighted_flow_matching_loss": 0.01610722616314888, + "weighted_rkd_loss": 0.006669075461104512 + }, + { + "epoch": 8.51905731742799, + "flow_matching_loss": 0.01862049028277397, + "grad_norm": 0.17435108125209808, + "learning_rate": 1.6831595913298787e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00544232320971787, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022904964862391353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013175142835825682, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5189762115478516, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29280, + "weighted_flow_matching_loss": 0.01862049028277397, + "weighted_rkd_loss": 0.006837605452165008 + }, + { + "epoch": 8.521966831539133, + "flow_matching_loss": 0.016597551852464677, + "grad_norm": 0.12215778231620789, + "learning_rate": 1.682339015086554e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005069669196382165, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022321759606711566, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012371514271944762, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5187145888805389, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29290, + "weighted_flow_matching_loss": 0.016597551852464677, + "weighted_rkd_loss": 0.006417261390015483 + }, + { + "epoch": 8.524876345650277, + "flow_matching_loss": 0.017737158853560687, + "grad_norm": 0.15316689014434814, + "learning_rate": 1.6815183834534507e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006122893281280994, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024789507035166023, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014724737405776978, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5184529721736908, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29300, + "weighted_flow_matching_loss": 0.017737158853560687, + "weighted_rkd_loss": 0.007634056778624654 + }, + { + "epoch": 8.52778585976142, + "flow_matching_loss": 0.019563950318843126, + "grad_norm": 0.17096024751663208, + "learning_rate": 1.680697696679855e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005532979546114802, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023034659447148443, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013369425013661385, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5181913495063781, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29310, + "weighted_flow_matching_loss": 0.019563950318843126, + "weighted_rkd_loss": 0.006927932519465685 + }, + { + "epoch": 8.530695373872563, + "flow_matching_loss": 0.01851632967591286, + "grad_norm": 0.16496874392032623, + "learning_rate": 1.6798769550150702e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005339477816596628, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002400065481197089, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013079021126031876, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5179297029972076, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29320, + "weighted_flow_matching_loss": 0.01851632967591286, + "weighted_rkd_loss": 0.006773885525763035 + }, + { + "epoch": 8.533604887983707, + "flow_matching_loss": 0.017538397293537854, + "grad_norm": 0.1405603289604187, + "learning_rate": 1.679056158708416e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0060003690887242556, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0026694857398979364, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01467022392898798, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.517668080329895, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29330, + "weighted_flow_matching_loss": 0.017538397293537854, + "weighted_rkd_loss": 0.007594365207478404 + }, + { + "epoch": 8.53651440209485, + "flow_matching_loss": 0.017386389803141354, + "grad_norm": 0.12398547679185867, + "learning_rate": 1.6782353080092283e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005145271727815271, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021710695582441986, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012461612932384014, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5174064338207245, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29340, + "weighted_flow_matching_loss": 0.017386389803141354, + "weighted_rkd_loss": 0.006447722902521491 + }, + { + "epoch": 8.539423916205994, + "flow_matching_loss": 0.01738954018801451, + "grad_norm": 0.13558433949947357, + "learning_rate": 1.6774144031668603e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0051479526096954945, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002268227783497423, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012564132921397686, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5171447694301605, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29350, + "weighted_flow_matching_loss": 0.01738954018801451, + "weighted_rkd_loss": 0.006497556995600462 + }, + { + "epoch": 8.542333430317138, + "flow_matching_loss": 0.018502697255462407, + "grad_norm": 0.12883007526397705, + "learning_rate": 1.6765934444306804e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005096045602113008, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020849520922638478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012277043331414461, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5168831169605255, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29360, + "weighted_flow_matching_loss": 0.018502697255462407, + "weighted_rkd_loss": 0.006345890974625945 + }, + { + "epoch": 8.545242944428281, + "flow_matching_loss": 0.01939577264711261, + "grad_norm": 0.14462637901306152, + "learning_rate": 1.6757724320500753e-05, + "loss": 0.0264, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005563215119764209, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024874322582036255, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013613862637430429, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5166215002536774, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29370, + "weighted_flow_matching_loss": 0.01939577264711261, + "weighted_rkd_loss": 0.007033217698335648 + }, + { + "epoch": 8.548152458539423, + "flow_matching_loss": 0.018849209975451232, + "grad_norm": 0.13215011358261108, + "learning_rate": 1.674951366274447e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005235691135749221, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022053785738535225, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012676760833710433, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5163598358631134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29380, + "weighted_flow_matching_loss": 0.018849209975451232, + "weighted_rkd_loss": 0.006545741809532046 + }, + { + "epoch": 8.551061972650567, + "flow_matching_loss": 0.01810924503952265, + "grad_norm": 0.140000581741333, + "learning_rate": 1.674130247353212e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005074948794208467, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021010084659792484, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01225090604275465, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5160981714725494, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29390, + "weighted_flow_matching_loss": 0.01810924503952265, + "weighted_rkd_loss": 0.006322733126580715 + }, + { + "epoch": 8.55397148676171, + "flow_matching_loss": 0.01701024705544114, + "grad_norm": 0.17664602398872375, + "learning_rate": 1.6733090755358056e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005124087305739522, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022724528098478913, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012520627491176129, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5158364832401275, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29400, + "weighted_flow_matching_loss": 0.01701024705544114, + "weighted_rkd_loss": 0.006458519073203206 + }, + { + "epoch": 8.556881000872854, + "flow_matching_loss": 0.01673349542543292, + "grad_norm": 0.14795365929603577, + "learning_rate": 1.672487851071678e-05, + "loss": 0.0234, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005316674523055553, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002296760678291321, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012930109910666943, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5155747830867767, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29410, + "weighted_flow_matching_loss": 0.01673349542543292, + "weighted_rkd_loss": 0.006666430272161961 + }, + { + "epoch": 8.559790514983998, + "flow_matching_loss": 0.018800613284111024, + "grad_norm": 0.14751815795898438, + "learning_rate": 1.6716665742102956e-05, + "loss": 0.0253, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00524653228931129, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020369697362184525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012530034501105547, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5153131186962128, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29420, + "weighted_flow_matching_loss": 0.018800613284111024, + "weighted_rkd_loss": 0.006456864252686501 + }, + { + "epoch": 8.562700029095142, + "flow_matching_loss": 0.017763999104499818, + "grad_norm": 0.11895459145307541, + "learning_rate": 1.670845245201141e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052709277020767335, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021768483449704943, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.0127187036909163, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5150514543056488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29430, + "weighted_flow_matching_loss": 0.017763999104499818, + "weighted_rkd_loss": 0.006550808064639568 + }, + { + "epoch": 8.565609543206284, + "flow_matching_loss": 0.018130759336054326, + "grad_norm": 0.12989358603954315, + "learning_rate": 1.670023864293711e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004793899483047426, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019736876944079994, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011561486776918173, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5147897481918335, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29440, + "weighted_flow_matching_loss": 0.018130759336054326, + "weighted_rkd_loss": 0.005951721826568246 + }, + { + "epoch": 8.568519057317427, + "flow_matching_loss": 0.01799194039776921, + "grad_norm": 0.2096477448940277, + "learning_rate": 1.669202431737521e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057959240861237046, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002476219739764929, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014068068005144596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5145280659198761, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29450, + "weighted_flow_matching_loss": 0.01799194039776921, + "weighted_rkd_loss": 0.007238434394821525 + }, + { + "epoch": 8.571428571428571, + "flow_matching_loss": 0.016711958684027195, + "grad_norm": 0.132594496011734, + "learning_rate": 1.6683809477820995e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005462027201429009, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0023464564932510258, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013270510919392109, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5142663657665253, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29460, + "weighted_flow_matching_loss": 0.016711958684027195, + "weighted_rkd_loss": 0.006824588309973479 + }, + { + "epoch": 8.574338085539715, + "flow_matching_loss": 0.01601647948846221, + "grad_norm": 0.15482595562934875, + "learning_rate": 1.6675594126769916e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005449651344679296, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002331322291865945, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013230625260621309, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5140046775341034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29470, + "weighted_flow_matching_loss": 0.01601647948846221, + "weighted_rkd_loss": 0.006800685683265329 + }, + { + "epoch": 8.577247599650859, + "flow_matching_loss": 0.01675174217671156, + "grad_norm": 0.15570585429668427, + "learning_rate": 1.6667378266717586e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005496794637292623, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002295621647499502, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013289210852235556, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5137429594993591, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29480, + "weighted_flow_matching_loss": 0.01675174217671156, + "weighted_rkd_loss": 0.006827152241021395 + }, + { + "epoch": 8.580157113762002, + "flow_matching_loss": 0.018523338809609415, + "grad_norm": 0.16891932487487793, + "learning_rate": 1.665916190015976e-05, + "loss": 0.0256, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005680111423134804, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024689418612979354, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01382916485890746, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5134812593460083, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29490, + "weighted_flow_matching_loss": 0.018523338809609415, + "weighted_rkd_loss": 0.0071009621024131775 + }, + { + "epoch": 8.583066627873146, + "flow_matching_loss": 0.01811918690800667, + "grad_norm": 0.16294215619564056, + "learning_rate": 1.665094502959236e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005327873351052404, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021548224380239845, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012810569163411856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.513219553232193, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29500, + "weighted_flow_matching_loss": 0.01811918690800667, + "weighted_rkd_loss": 0.006574710225686431 + }, + { + "epoch": 8.585976141984288, + "flow_matching_loss": 0.01864691013470292, + "grad_norm": 0.16181448101997375, + "learning_rate": 1.664272765751145e-05, + "loss": 0.025, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005128176650032401, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021342763095162807, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012390629574656486, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5129578351974488, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29510, + "weighted_flow_matching_loss": 0.01864691013470292, + "weighted_rkd_loss": 0.006355934450402856 + }, + { + "epoch": 8.588885656095432, + "flow_matching_loss": 0.01674150200560689, + "grad_norm": 0.14673179388046265, + "learning_rate": 1.6634509786413248e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005206554988399148, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022428821423090993, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012655992060899734, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5126961171627045, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29520, + "weighted_flow_matching_loss": 0.01674150200560689, + "weighted_rkd_loss": 0.006488642375916243 + }, + { + "epoch": 8.591795170206575, + "flow_matching_loss": 0.015757645200937986, + "grad_norm": 0.1603100597858429, + "learning_rate": 1.6626291418794135e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.006151844235137105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002825342584401369, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01512903105467558, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5124343931674957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29530, + "weighted_flow_matching_loss": 0.015757645200937986, + "weighted_rkd_loss": 0.007752620382234454 + }, + { + "epoch": 8.594704684317719, + "flow_matching_loss": 0.016940520703792573, + "grad_norm": 0.14220575988292694, + "learning_rate": 1.6618072557150625e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005086202896200121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002065643074456602, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012238048948347569, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5121726751327514, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29540, + "weighted_flow_matching_loss": 0.016940520703792573, + "weighted_rkd_loss": 0.006267971452325582 + }, + { + "epoch": 8.597614198428863, + "flow_matching_loss": 0.01781073259189725, + "grad_norm": 0.14604677259922028, + "learning_rate": 1.6609853203979398e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005433709779754281, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024723074515350164, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013339726999402047, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5119109392166138, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29550, + "weighted_flow_matching_loss": 0.01781073259189725, + "weighted_rkd_loss": 0.006828727805987001 + }, + { + "epoch": 8.600523712540006, + "flow_matching_loss": 0.019442737568169832, + "grad_norm": 0.1367872953414917, + "learning_rate": 1.6601633361777273e-05, + "loss": 0.0259, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005200985400006175, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022436051280237735, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012645575869828462, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5116492092609406, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29560, + "weighted_flow_matching_loss": 0.019442737568169832, + "weighted_rkd_loss": 0.006470043770968914 + }, + { + "epoch": 8.603433226651148, + "flow_matching_loss": 0.019257505517452957, + "grad_norm": 0.15901978313922882, + "learning_rate": 1.6593413033041215e-05, + "loss": 0.0252, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004827382229268551, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019563879119232296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011611152440309525, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5113874793052673, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29570, + "weighted_flow_matching_loss": 0.019257505517452957, + "weighted_rkd_loss": 0.0059377821162343025 + }, + { + "epoch": 8.606342740762292, + "flow_matching_loss": 0.017784151434898376, + "grad_norm": 0.14540539681911469, + "learning_rate": 1.6585192220268356e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005343259731307626, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022783891414292156, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012964908685535192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5111257433891296, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29580, + "weighted_flow_matching_loss": 0.017784151434898376, + "weighted_rkd_loss": 0.006626742053776979 + }, + { + "epoch": 8.609252254873436, + "flow_matching_loss": 0.016507871448993683, + "grad_norm": 0.14016112685203552, + "learning_rate": 1.6576970925955956e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005150186596438289, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021722910343669353, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012472664099186659, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.510864007472992, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29590, + "weighted_flow_matching_loss": 0.016507871448993683, + "weighted_rkd_loss": 0.006371835758909583 + }, + { + "epoch": 8.61216176898458, + "flow_matching_loss": 0.015703776199370623, + "grad_norm": 0.12710222601890564, + "learning_rate": 1.6568749152601422e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004808323457837105, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002006615651771426, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011623262520879507, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5106022536754609, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29600, + "weighted_flow_matching_loss": 0.015703776199370623, + "weighted_rkd_loss": 0.005935006123036146 + }, + { + "epoch": 8.615071283095723, + "flow_matching_loss": 0.01757068783044815, + "grad_norm": 0.1750188022851944, + "learning_rate": 1.6560526902702317e-05, + "loss": 0.0246, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005744726629927754, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002270491421222687, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013759944494813681, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5103405117988586, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29610, + "weighted_flow_matching_loss": 0.01757068783044815, + "weighted_rkd_loss": 0.007022208161652088 + }, + { + "epoch": 8.617980797206867, + "flow_matching_loss": 0.017190766148269178, + "grad_norm": 0.14011727273464203, + "learning_rate": 1.6552304178756337e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005657236138358712, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024550499860197307, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013769522309303284, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.510078763961792, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29620, + "weighted_flow_matching_loss": 0.017190766148269178, + "weighted_rkd_loss": 0.007023417670279741 + }, + { + "epoch": 8.62089031131801, + "flow_matching_loss": 0.0177566877566278, + "grad_norm": 0.15058483183383942, + "learning_rate": 1.654408098326134e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005211060913279652, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002243866561912, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012665988225489856, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5098170101642608, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29630, + "weighted_flow_matching_loss": 0.0177566877566278, + "weighted_rkd_loss": 0.006457232590764761 + }, + { + "epoch": 8.623799825429153, + "flow_matching_loss": 0.017850804887712002, + "grad_norm": 0.14011134207248688, + "learning_rate": 1.6535857318715307e-05, + "loss": 0.0238, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004806125350296497, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002010239928495139, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011622490733861924, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5095552563667297, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29640, + "weighted_flow_matching_loss": 0.017850804887712002, + "weighted_rkd_loss": 0.005922340322285891 + }, + { + "epoch": 8.626709339540296, + "flow_matching_loss": 0.018368739634752274, + "grad_norm": 0.13669832050800323, + "learning_rate": 1.6527633187616376e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004676266643218696, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019303803681395947, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01128291366621852, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5092934966087341, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29650, + "weighted_flow_matching_loss": 0.018368739634752274, + "weighted_rkd_loss": 0.005746327899396419 + }, + { + "epoch": 8.62961885365144, + "flow_matching_loss": 0.016190203465521337, + "grad_norm": 0.1298188865184784, + "learning_rate": 1.6519408592462818e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00542693673633039, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002301130455452949, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013155004009604454, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5090317368507385, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29660, + "weighted_flow_matching_loss": 0.016190203465521337, + "weighted_rkd_loss": 0.006696316506713628 + }, + { + "epoch": 8.632528367762584, + "flow_matching_loss": 0.01663348264992237, + "grad_norm": 0.14892300963401794, + "learning_rate": 1.6511183535753044e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005212189443409443, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021400668076239525, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012564445845782757, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5087699830532074, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29670, + "weighted_flow_matching_loss": 0.01663348264992237, + "weighted_rkd_loss": 0.006392405647784471 + }, + { + "epoch": 8.635437881873727, + "flow_matching_loss": 0.019615863263607026, + "grad_norm": 0.14832131564617157, + "learning_rate": 1.650295801998561e-05, + "loss": 0.0261, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005240454711019993, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002172864566091448, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01265377402305603, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5085082113742828, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29680, + "weighted_flow_matching_loss": 0.019615863263607026, + "weighted_rkd_loss": 0.006434532860293984 + }, + { + "epoch": 8.638347395984871, + "flow_matching_loss": 0.016641623806208372, + "grad_norm": 0.1377168744802475, + "learning_rate": 1.6494732047659217e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005098388995975256, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002236675680615008, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012433453556150198, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5082464396953583, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29690, + "weighted_flow_matching_loss": 0.016641623806208372, + "weighted_rkd_loss": 0.006319225672632456 + }, + { + "epoch": 8.641256910096015, + "flow_matching_loss": 0.01652145469561219, + "grad_norm": 0.13705256581306458, + "learning_rate": 1.6486505621272693e-05, + "loss": 0.0229, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005213264701887965, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021641890052706003, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01259071845561266, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5079846739768982, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29700, + "weighted_flow_matching_loss": 0.01652145469561219, + "weighted_rkd_loss": 0.0063959195278584955 + }, + { + "epoch": 8.644166424207157, + "flow_matching_loss": 0.020108638517558574, + "grad_norm": 0.14119233191013336, + "learning_rate": 1.647827874332501e-05, + "loss": 0.0265, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005153431929647923, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002273260988295078, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012580124661326408, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5077229082584381, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29710, + "weighted_flow_matching_loss": 0.020108638517558574, + "weighted_rkd_loss": 0.006387287983670831 + }, + { + "epoch": 8.6470759383183, + "flow_matching_loss": 0.017926150001585484, + "grad_norm": 0.1389397382736206, + "learning_rate": 1.6470051416315277e-05, + "loss": 0.0242, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005164644913747907, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021219834568910302, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012451273389160633, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5074611306190491, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29720, + "weighted_flow_matching_loss": 0.017926150001585484, + "weighted_rkd_loss": 0.006318543711677194 + }, + { + "epoch": 8.649985452429444, + "flow_matching_loss": 0.017175946943461896, + "grad_norm": 0.1360991895198822, + "learning_rate": 1.6461823642742734e-05, + "loss": 0.0231, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004766289214603603, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002109058806672692, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011641637049615383, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.50719935297966, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29730, + "weighted_flow_matching_loss": 0.017175946943461896, + "weighted_rkd_loss": 0.005904712062329054 + }, + { + "epoch": 8.652894966540588, + "flow_matching_loss": 0.018019785545766352, + "grad_norm": 0.12181439995765686, + "learning_rate": 1.645359542510677e-05, + "loss": 0.0255, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005862269317731261, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.003009387874044478, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014733926765620709, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.506937575340271, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29740, + "weighted_flow_matching_loss": 0.018019785545766352, + "weighted_rkd_loss": 0.007469208072870969 + }, + { + "epoch": 8.655804480651732, + "flow_matching_loss": 0.018906512577086688, + "grad_norm": 0.1644303947687149, + "learning_rate": 1.6445366765906897e-05, + "loss": 0.0254, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0052561859600245954, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022920659626834094, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012804437894374132, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.506675797700882, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29750, + "weighted_flow_matching_loss": 0.018906512577086688, + "weighted_rkd_loss": 0.006487598689273 + }, + { + "epoch": 8.658713994762875, + "flow_matching_loss": 0.01778602758422494, + "grad_norm": 0.1513386219739914, + "learning_rate": 1.643713766764276e-05, + "loss": 0.0251, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00599814944434911, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024311058688908814, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014427404943853616, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5064140200614929, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29760, + "weighted_flow_matching_loss": 0.01778602758422494, + "weighted_rkd_loss": 0.007306154305115342 + }, + { + "epoch": 8.661623508874017, + "flow_matching_loss": 0.017941666208207606, + "grad_norm": 0.15074308216571808, + "learning_rate": 1.6428908132814146e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005551006738096475, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0024457071674987675, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013547720480710268, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5061522364616394, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29770, + "weighted_flow_matching_loss": 0.017941666208207606, + "weighted_rkd_loss": 0.00685717687010765 + }, + { + "epoch": 8.664533022985161, + "flow_matching_loss": 0.017350023612380028, + "grad_norm": 0.1330665946006775, + "learning_rate": 1.642067816392097e-05, + "loss": 0.0245, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0057555762585252525, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002623905939981341, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.014135058503597975, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5058904528617859, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29780, + "weighted_flow_matching_loss": 0.017350023612380028, + "weighted_rkd_loss": 0.00715072681196034 + }, + { + "epoch": 8.667442537096305, + "flow_matching_loss": 0.017107105627655984, + "grad_norm": 0.1405618041753769, + "learning_rate": 1.6412447763463275e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005499140056781471, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002366777346469462, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013365057483315467, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5056286573410034, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29790, + "weighted_flow_matching_loss": 0.017107105627655984, + "weighted_rkd_loss": 0.006757782818749547 + }, + { + "epoch": 8.670352051207448, + "flow_matching_loss": 0.016290596686303616, + "grad_norm": 0.13471858203411102, + "learning_rate": 1.6404216933941247e-05, + "loss": 0.0228, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005360636883415282, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021310357027687133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012852309551090003, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5053668737411499, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29800, + "weighted_flow_matching_loss": 0.016290596686303616, + "weighted_rkd_loss": 0.006495111295953393 + }, + { + "epoch": 8.673261565318592, + "flow_matching_loss": 0.017911187652498484, + "grad_norm": 0.1253967434167862, + "learning_rate": 1.639598567785519e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005040649371221661, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021125456201843917, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01219384428113699, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5051050841808319, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29810, + "weighted_flow_matching_loss": 0.017911187652498484, + "weighted_rkd_loss": 0.006159248342737555 + }, + { + "epoch": 8.676171079429736, + "flow_matching_loss": 0.017378879804164173, + "grad_norm": 0.14832539856433868, + "learning_rate": 1.6387753997705547e-05, + "loss": 0.0244, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005700917914509773, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002544662926811725, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01394649874418974, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5048432946205139, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29820, + "weighted_flow_matching_loss": 0.017378879804164173, + "weighted_rkd_loss": 0.007040861435234547 + }, + { + "epoch": 8.67908059354088, + "flow_matching_loss": 0.016522313840687274, + "grad_norm": 0.1310790777206421, + "learning_rate": 1.637952189599287e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005644674994982779, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0025715642841532826, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013860914204269648, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5045814990997315, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29830, + "weighted_flow_matching_loss": 0.016522313840687274, + "weighted_rkd_loss": 0.006993905268609523 + }, + { + "epoch": 8.681990107652021, + "flow_matching_loss": 0.01719173640012741, + "grad_norm": 0.13486695289611816, + "learning_rate": 1.6371289375217873e-05, + "loss": 0.0237, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0053377309115603564, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022001204662956296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012875582370907069, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5043197095394134, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29840, + "weighted_flow_matching_loss": 0.01719173640012741, + "weighted_rkd_loss": 0.006493424903601408 + }, + { + "epoch": 8.684899621763165, + "flow_matching_loss": 0.016017137467861174, + "grad_norm": 0.1983809620141983, + "learning_rate": 1.6363056437881372e-05, + "loss": 0.0227, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005451506143435836, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022864082478918136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.013189420662820339, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.504057914018631, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29850, + "weighted_flow_matching_loss": 0.016017137467861174, + "weighted_rkd_loss": 0.0066481609363108875 + }, + { + "epoch": 8.687809135874309, + "flow_matching_loss": 0.018645842373371125, + "grad_norm": 0.12086745351552963, + "learning_rate": 1.6354823086484315e-05, + "loss": 0.0247, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.00495850844308734, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020864692749455573, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012003486230969428, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5037961184978486, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29860, + "weighted_flow_matching_loss": 0.018645842373371125, + "weighted_rkd_loss": 0.006047300342470407 + }, + { + "epoch": 8.690718649985453, + "flow_matching_loss": 0.015843313559889792, + "grad_norm": 0.1376199573278427, + "learning_rate": 1.634658932352777e-05, + "loss": 0.0217, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.0048790497472509745, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019598883227445185, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011717987805604934, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.503534322977066, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29870, + "weighted_flow_matching_loss": 0.015843313559889792, + "weighted_rkd_loss": 0.005900452379137278 + }, + { + "epoch": 8.693628164096596, + "flow_matching_loss": 0.017647433560341597, + "grad_norm": 0.15373347699642181, + "learning_rate": 1.633835515151295e-05, + "loss": 0.0243, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005453508673235774, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022278503980487584, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01313486760482192, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5032725214958191, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29880, + "weighted_flow_matching_loss": 0.017647433560341597, + "weighted_rkd_loss": 0.006610465934500098 + }, + { + "epoch": 8.69653767820774, + "flow_matching_loss": 0.018332714680582285, + "grad_norm": 0.1312570720911026, + "learning_rate": 1.6330120572941166e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005302869062870741, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021921329898759723, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012797870952636004, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5030107259750366, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29890, + "weighted_flow_matching_loss": 0.018332714680582285, + "weighted_rkd_loss": 0.006437594443559647 + }, + { + "epoch": 8.699447192318882, + "flow_matching_loss": 0.017668481450527906, + "grad_norm": 0.11483374238014221, + "learning_rate": 1.6321885590313876e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005106448917649687, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002223833731841296, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012436731532216073, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5027489244937897, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29900, + "weighted_flow_matching_loss": 0.017668481450527906, + "weighted_rkd_loss": 0.006252560997381806 + }, + { + "epoch": 8.702356706430026, + "flow_matching_loss": 0.01692213276401162, + "grad_norm": 0.1298714429140091, + "learning_rate": 1.6313650206132643e-05, + "loss": 0.0232, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005191482696682214, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002182361262384802, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012565326690673829, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5024871289730072, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29910, + "weighted_flow_matching_loss": 0.01692213276401162, + "weighted_rkd_loss": 0.006313943257555366 + }, + { + "epoch": 8.70526622054117, + "flow_matching_loss": 0.017339573614299296, + "grad_norm": 0.1181827187538147, + "learning_rate": 1.6305414422899163e-05, + "loss": 0.0239, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005421560117974877, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022976038744673133, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01314072422683239, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5022253215312957, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29920, + "weighted_flow_matching_loss": 0.017339573614299296, + "weighted_rkd_loss": 0.006599749205633998 + }, + { + "epoch": 8.708175734652313, + "flow_matching_loss": 0.01742311557754874, + "grad_norm": 0.12555256485939026, + "learning_rate": 1.629717824311525e-05, + "loss": 0.0241, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005501291481778026, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002320782945025712, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01332336589694023, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5019635260105133, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29930, + "weighted_flow_matching_loss": 0.01742311557754874, + "weighted_rkd_loss": 0.0066879056859761475 + }, + { + "epoch": 8.711085248763457, + "flow_matching_loss": 0.017196383606642485, + "grad_norm": 0.12426957488059998, + "learning_rate": 1.6288941669282834e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005237928777933121, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.002061658026650548, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012537515535950661, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5017017245292663, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29940, + "weighted_flow_matching_loss": 0.017196383606642485, + "weighted_rkd_loss": 0.006290188850834965 + }, + { + "epoch": 8.7139947628746, + "flow_matching_loss": 0.01699206754565239, + "grad_norm": 0.13117314875125885, + "learning_rate": 1.628070470390398e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004867416177876294, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021570858545601366, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.011891918163746596, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.501439917087555, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29950, + "weighted_flow_matching_loss": 0.01699206754565239, + "weighted_rkd_loss": 0.005963043868541717 + }, + { + "epoch": 8.716904276985744, + "flow_matching_loss": 0.017336604371666908, + "grad_norm": 0.11928527057170868, + "learning_rate": 1.627246734948084e-05, + "loss": 0.0236, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005184647766873241, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0020896838745102287, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012458979617804289, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.501178115606308, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29960, + "weighted_flow_matching_loss": 0.017336604371666908, + "weighted_rkd_loss": 0.006244175508618355 + }, + { + "epoch": 8.719813791096886, + "flow_matching_loss": 0.018437707237899302, + "grad_norm": 0.1504920870065689, + "learning_rate": 1.6264229608515725e-05, + "loss": 0.0248, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005275366478599608, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022420782246626914, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012792811077088118, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.500916314125061, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29970, + "weighted_flow_matching_loss": 0.018437707237899302, + "weighted_rkd_loss": 0.006408122880384326 + }, + { + "epoch": 8.72272330520803, + "flow_matching_loss": 0.016581560019403696, + "grad_norm": 0.11790060251951218, + "learning_rate": 1.6255991483511035e-05, + "loss": 0.023, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005312191578559578, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0022407250595279037, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012865108344703913, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5006545066833497, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29980, + "weighted_flow_matching_loss": 0.016581560019403696, + "weighted_rkd_loss": 0.00644097519107163 + }, + { + "epoch": 8.725632819319173, + "flow_matching_loss": 0.017109373398125172, + "grad_norm": 0.1380632519721985, + "learning_rate": 1.6247752976969288e-05, + "loss": 0.0235, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.005314710177481174, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0021632587420754136, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.012792679108679294, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5003927052021027, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 29990, + "weighted_flow_matching_loss": 0.017109373398125172, + "weighted_rkd_loss": 0.006401348579674959 + }, + { + "epoch": 8.728542333430317, + "flow_matching_loss": 0.01589776687324047, + "grad_norm": 0.1293402761220932, + "learning_rate": 1.6239514091393138e-05, + "loss": 0.0216, + "rkd_action_encoder_cls_ffn_hidden_dim": 2048.0, + "rkd_action_encoder_cls_num_heads": 8.0, + "rkd_action_encoder_cls_num_layers": 1.0, + "rkd_action_encoder_projector_dim": 512.0, + "rkd_action_encoder_projector_enabled": 0.0, + "rkd_action_encoder_projector_pooling_id": 0.0, + "rkd_action_temp": 0.10000000149011612, + "rkd_angle_loss": 0.004751146538183093, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss": 0.0019685216597281397, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05000000074505806, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode_id": 0.0, + "rkd_dtw_normalize_components": 1.0, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_exclude_diagonal": 1.0, + "rkd_fm_loss_weight": 1.0, + "rkd_loss": 0.01147081470116973, + "rkd_loss_type_id": 3.0, + "rkd_loss_weight": 0.5001309037208557, + "rkd_raw_action_projector_dim": 512.0, + "rkd_raw_action_projector_enabled": 0.0, + "rkd_raw_action_projector_pooling_id": 0.0, + "rkd_relation_mode_id": 0.0, + "rkd_student_reconstruction_head_dim": 512.0, + "rkd_student_reconstruction_head_enabled": 1.0, + "rkd_student_reconstruction_head_pooling_id": 0.0, + "rkd_student_similarity_mean": 0.0, + "rkd_teacher_entropy": 0.0, + "rkd_teacher_similarity_mean": 0.0, + "rkd_teacher_source_id": 3.0, + "rkd_vlm_temp": 0.10000000149011612, + "step": 30000, + "weighted_flow_matching_loss": 0.01589776687324047, + "weighted_rkd_loss": 0.005736908782273531 + } + ], + "logging_steps": 10, + "max_steps": 60000, + "num_input_tokens_seen": 0, + "num_train_epochs": 18, + "save_steps": 30000, + "stateful_callbacks": { + "TrainerControl": { + "args": { + "should_epoch_stop": false, + "should_evaluate": false, + "should_log": false, + "should_save": true, + "should_training_stop": false + }, + "attributes": {} + } + }, + "total_flos": 0.0, + "train_batch_size": 128, + "trial_name": null, + "trial_params": null +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/config.json new file mode 100644 index 0000000000000000000000000000000000000000..064785cc03d1e2223bf2981a1b22674a7e445646 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.0, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/experiment_cfg/metadata.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/experiment_cfg/metadata.json new file mode 100644 index 0000000000000000000000000000000000000000..300eb070963337016e16465f50ba8351ea2eb25c --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/experiment_cfg/metadata.json @@ -0,0 +1,431 @@ +{ + "new_embodiment": { + "statistics": { + "state": { + "base_position": { + "max": [ + 7.3139495849609375, + 0.4876587688922882, + 0.7196521759033203 + ], + "min": [ + -4.94293737411499, + -6.890198230743408, + 0.6996827125549316 + ], + "mean": [ + 2.749288365724937, + -2.0455216239909983, + 0.700532551020533 + ], + "std": [ + 1.6786295897046262, + 1.312897969434244, + 0.00133751758906258 + ], + "q01": [ + -1.4241907881148037, + -5.1716547935950885, + 0.6999670898768614 + ], + "q99": [ + 6.198111545831555, + -0.5873531334104489, + 0.7038833014242587 + ] + }, + "base_rotation": { + "max": [ + 0.0, + 0.0, + 1.0, + 1.0 + ], + "min": [ + 0.0, + 0.0, + -1.0, + 0.0 + ], + "mean": [ + 0.0, + 0.0, + 0.23084999587450247, + 0.6238431088581847 + ], + "std": [ + 0.0, + 0.0, + 0.6556404371644047, + 0.3572252105732042 + ], + "q01": [ + 0.0, + 0.0, + -0.9999999999999999, + 1.7482354593273349e-06 + ], + "q99": [ + 0.0, + 0.0, + 0.9999999999999999, + 0.9999999999999999 + ] + }, + "end_effector_position_relative": { + "max": [ + 0.9014528393745422, + 0.8003877401351929, + 0.9829942584037781 + ], + "min": [ + -0.37471598386764526, + -0.8472502827644348, + -0.25070279836654663 + ], + "mean": [ + 0.28667332601863704, + -0.038315473170876704, + 0.45974658906777444 + ], + "std": [ + 0.17067058828305154, + 0.23569952037784195, + 0.21950702354181487 + ], + "q01": [ + 0.015172948963784228, + -0.44450502063220615, + 0.23314444103329338 + ], + "q99": [ + 0.5609069689572412, + 0.38454179128009547, + 0.7028102736473852 + ] + }, + "end_effector_rotation_relative": { + "max": [ + 0.9999998807907104, + 0.9984455108642578, + 0.9434727430343628, + 0.9062229990959167 + ], + "min": [ + -0.9999930262565613, + -0.9988337755203247, + -0.9618149995803833, + 2.1872274658107926e-07 + ], + "mean": [ + -0.25672297401336, + 0.024425335806687216, + -0.08740977164483847, + 0.16608739644018397 + ], + "std": [ + 0.7932585208392383, + 0.3102260195580416, + 0.3772472576315148, + 0.17451959300158773 + ], + "q01": [ + -0.99379006987882, + -0.5478102050038828, + -0.6101777952971636, + 0.002274085014134748 + ], + "q99": [ + 0.8804856038896288, + 0.5910611092873037, + 0.5216058042993562, + 0.5231965974012255 + ] + }, + "gripper_qpos": { + "max": [ + 0.055869169533252716, + 0.010916369967162609 + ], + "min": [ + -0.011436971835792065, + -0.05664053186774254 + ], + "mean": [ + 0.031651551516052555, + -0.03161481387920421 + ], + "std": [ + 0.013119769222217526, + 0.01306078225192402 + ], + "q01": [ + 0.0060999246446777336, + -0.04062601653964198 + ], + "q99": [ + 0.04054891621517283, + -0.0059163567113456345 + ] + } + }, + "action": { + "base_motion": { + "max": [ + 1.0, + 1.0, + 1.0, + 0.0 + ], + "min": [ + -1.0, + -1.0, + -1.0, + 0.0 + ], + "mean": [ + 0.008144896753718373, + -0.00018893135146078263, + -0.0008739062727221845, + 0.0 + ], + "std": [ + 0.11030045424364411, + 0.10148082570313594, + 0.08975698373467506, + 0.0 + ], + "q01": [ + -0.07287373067581518, + -0.09446994199569948, + -0.07702818259249572, + 0.0 + ], + "q99": [ + 0.04485485414407898, + 0.09626941605965177, + 0.07610665906122742, + 0.0 + ] + }, + "control_mode": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.9221974313425939 + ], + "std": [ + 0.38670194342612674 + ], + "q01": [ + -1.0 + ], + "q99": [ + -0.384693946883099 + ] + }, + "end_effector_position": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.01976129923017491, + -0.020870631344490034, + -0.05993476517349181 + ], + "std": [ + 0.4347024990804053, + 0.4221662976569997, + 0.3845506843044066 + ], + "q01": [ + -0.8835792939767807, + -0.9280919251713778, + -0.783361071470956 + ], + "q99": [ + 0.7622084438449307, + 0.8625125252609077, + 0.7470974753250706 + ] + }, + "end_effector_rotation": { + "max": [ + 1.0, + 1.0, + 1.0 + ], + "min": [ + -1.0, + -1.0, + -1.0 + ], + "mean": [ + 0.008089673068544335, + -0.026498254627927348, + 0.0029083551743059005 + ], + "std": [ + 0.11216734503733773, + 0.13206002613798629, + 0.12615870412692864 + ], + "q01": [ + -0.2814627972846672, + -0.4342386516115439, + -0.34489755628340574 + ], + "q99": [ + 0.31879353266939386, + 0.28411797683220563, + 0.3597718671669894 + ] + }, + "gripper_close": { + "max": [ + 1.0 + ], + "min": [ + -1.0 + ], + "mean": [ + -0.3824228874769045 + ], + "std": [ + 0.9240015096469786 + ], + "q01": [ + -1.0 + ], + "q99": [ + 0.5597272305813592 + ] + } + } + }, + "modalities": { + "video": { + "robot0_eye_in_hand": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_left": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + }, + "robot0_agentview_right": { + "resolution": [ + 256, + 256 + ], + "channels": 3, + "fps": 20.0 + } + }, + "state": { + "base_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "base_rotation": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "end_effector_position_relative": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation_relative": { + "absolute": true, + "rotation_type": "quaternion", + "shape": [ + 4 + ], + "continuous": true + }, + "gripper_qpos": { + "absolute": true, + "rotation_type": null, + "shape": [ + 2 + ], + "continuous": true + } + }, + "action": { + "base_motion": { + "absolute": true, + "rotation_type": null, + "shape": [ + 4 + ], + "continuous": true + }, + "control_mode": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + }, + "end_effector_position": { + "absolute": true, + "rotation_type": null, + "shape": [ + 3 + ], + "continuous": true + }, + "end_effector_rotation": { + "absolute": true, + "rotation_type": "axis_angle", + "shape": [ + 3 + ], + "continuous": true + }, + "gripper_close": { + "absolute": true, + "rotation_type": null, + "shape": [ + 1 + ], + "continuous": true + } + } + }, + "embodiment_tag": "new_embodiment" + } +} \ No newline at end of file diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/scheduler.pt b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/scheduler.pt new file mode 100644 index 0000000000000000000000000000000000000000..e0c04a5e834651dd0b7d7d785e6b6a92d586fc38 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/checkpoint-60000/scheduler.pt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:7ae8d5d8de3b387871d02c6360a92a63b63211c9bc3fc033261259c682861fdd +size 1465 diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/config.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/config.json new file mode 100644 index 0000000000000000000000000000000000000000..064785cc03d1e2223bf2981a1b22674a7e445646 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/config.json @@ -0,0 +1,101 @@ +{ + "action_dim": 32, + "action_head_cfg": { + "action_dim": 32, + "action_horizon": 16, + "add_pos_embed": true, + "backbone_embedding_dim": 2048, + "diffusion_model_cfg": { + "attention_head_dim": 48, + "cross_attention_dim": 2048, + "dropout": 0.2, + "final_dropout": true, + "interleave_self_attention": true, + "norm_type": "ada_norm", + "num_attention_heads": 32, + "num_layers": 16, + "output_dim": 1024, + "positional_embeddings": null + }, + "hidden_size": 1024, + "input_embedding_dim": 1536, + "max_action_dim": 32, + "max_state_dim": 64, + "model_dtype": "float32", + "noise_beta_alpha": 1.5, + "noise_beta_beta": 1.0, + "noise_s": 0.999, + "num_inference_timesteps": 4, + "num_target_vision_tokens": 32, + "num_timestep_buckets": 1000, + "tune_diffusion_model": true, + "tune_projector": true, + "use_vlln": true, + "vl_self_attention_cfg": { + "attention_head_dim": 64, + "dropout": 0.2, + "final_dropout": true, + "num_attention_heads": 32, + "num_layers": 4, + "positional_embeddings": null + } + }, + "action_horizon": 16, + "architectures": [ + "GR00T_N1_5_RKD" + ], + "attn_implementation": null, + "backbone_cfg": { + "eagle_path": "NVEagle/eagle_er-qwen3_1_7B-Siglip2_400M_stage1_5_128gpu_er_v7_1mlp_nops", + "load_bf16": false, + "project_to_dim": null, + "reproject_vision": false, + "select_layer": 12, + "tune_llm": false, + "tune_visual": true, + "use_flash_attention": true + }, + "compute_dtype": "bfloat16", + "hidden_size": 2048, + "model_dtype": "float32", + "model_type": "gr00t_n1_5", + "rkd_action_encoder_cls_ffn_hidden_dim": 2048, + "rkd_action_encoder_cls_num_heads": 8, + "rkd_action_encoder_cls_num_layers": 1, + "rkd_action_encoder_projector_dim": 512, + "rkd_action_encoder_projector_enabled": false, + "rkd_action_encoder_projector_pooling": "flatten", + "rkd_action_temp": 0.1, + "rkd_angle_loss_weight": 2.0, + "rkd_distance_loss_weight": 1.0, + "rkd_dtw_adaptive_floor": 0.05, + "rkd_dtw_adaptive_mix_lambda": 0.5, + "rkd_dtw_base_weight": 1.0, + "rkd_dtw_gripper_weight": 0.5, + "rkd_dtw_mix_mode": "distance_mix", + "rkd_dtw_normalize_components": true, + "rkd_dtw_pos_weight": 1.0, + "rkd_dtw_rot_weight": 1.0, + "rkd_dtw_timing_log_interval": 1000, + "rkd_enabled": true, + "rkd_exclude_diagonal": true, + "rkd_fm_loss_weight": 1.0, + "rkd_loss_type": "distance_angle", + "rkd_loss_weight": 0.0, + "rkd_loss_weight_end": 0.0, + "rkd_loss_weight_schedule": "cosine_decay", + "rkd_loss_weight_start": 1.0, + "rkd_raw_action_projector_dim": 512, + "rkd_raw_action_projector_enabled": false, + "rkd_raw_action_projector_pooling": "flatten", + "rkd_relation_mode": "token_pair_mean", + "rkd_student_reconstruction_head_dim": 512, + "rkd_student_reconstruction_head_enabled": true, + "rkd_student_reconstruction_head_hidden_dim": 512, + "rkd_student_reconstruction_head_pooling": "attention", + "rkd_teacher_source": "timewarp_action", + "rkd_timewarp_vae_checkpoint": "${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46", + "rkd_vlm_temp": 0.1, + "torch_dtype": "bfloat16", + "transformers_version": "4.51.3" +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/model.safetensors.index.json b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/model.safetensors.index.json new file mode 100644 index 0000000000000000000000000000000000000000..b20baed71f4d16a411223b70d45d934c6cbbfe97 --- /dev/null +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k/model.safetensors.index.json @@ -0,0 +1,963 @@ +{ + "metadata": { + "total_size": 7663767476 + }, + "weight_map": { + "action_head.action_decoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.action_decoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W1.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W2.b": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.W": "model-00002-of-00002.safetensors", + "action_head.action_encoder.W3.b": "model-00002-of-00002.safetensors", + "action_head.future_tokens.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_1.weight": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.bias": "model-00002-of-00002.safetensors", + "action_head.model.proj_out_2.weight": "model-00002-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_1.weight": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.bias": "model-00001-of-00002.safetensors", + "action_head.model.timestep_encoder.timestep_embedder.linear_2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.0.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.1.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.10.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.11.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.12.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.13.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.14.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.bias": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.15.norm1.linear.weight": "model-00002-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.2.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.3.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.4.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.5.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.6.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.7.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.8.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_k.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_out.0.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_q.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.attn1.to_v.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.0.proj.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.ff.net.2.weight": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.bias": "model-00001-of-00002.safetensors", + "action_head.model.transformer_blocks.9.norm1.linear.weight": "model-00001-of-00002.safetensors", + "action_head.position_embedding.weight": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer1.b": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.W": "model-00002-of-00002.safetensors", + "action_head.state_encoder.layer2.b": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.0.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.1.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.2.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_k.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_out.0.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_q.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.attn1.to_v.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.0.proj.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.ff.net.2.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm1.weight": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.bias": "model-00002-of-00002.safetensors", + "action_head.vl_self_attention.transformer_blocks.3.norm3.weight": "model-00002-of-00002.safetensors", + "action_head.vlln.bias": "model-00002-of-00002.safetensors", + "action_head.vlln.weight": "model-00002-of-00002.safetensors", + "backbone.eagle_model.language_model.lm_head.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.embed_tokens.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.gate_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.language_model.model.norm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.mlp1.0.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.patch_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.embeddings.position_embedding.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.20.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.21.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.22.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.23.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.24.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.25.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.26.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.layer_norm2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.encoder.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.in_proj_weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.attention.out_proj.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.layernorm.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc1.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.mlp.fc2.weight": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.head.probe": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.bias": "model-00001-of-00002.safetensors", + "backbone.eagle_model.vision_model.vision_model.post_layernorm.weight": "model-00001-of-00002.safetensors", + "rkd_action_encoder_cls_projector.cls_token": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.linear2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm1.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.norm2.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.in_proj_weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.encoder.layers.0.self_attn.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.in_proj.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.norm.weight": "model-00002-of-00002.safetensors", + "rkd_action_encoder_cls_projector.position_embedding": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.bias": "model-00002-of-00002.safetensors", + "rkd_action_encoder_projector.weight": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.bias": "model-00002-of-00002.safetensors", + "rkd_raw_action_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.bias": "model-00002-of-00002.safetensors", + "rkd_student_mean_projector.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.attn_score.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.out_proj.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.0.weight": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.bias": "model-00002-of-00002.safetensors", + "rkd_student_reconstruction_head.token_mlp.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.decoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.logvar.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.mu.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.monotonic.knots": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.0.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.2.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.net.4.weight": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.bias": "model-00002-of-00002.safetensors", + "rkd_timewarp_vae.timewarper.parameter_encoder.out.weight": "model-00002-of-00002.safetensors" + } +} diff --git a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/resolved_config.yaml b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/resolved_config.yaml index 792d759e3729b53b31cac567a4cb7c154ba7c256..1a06e1ac189142984b47554c3803d013cb565a87 100644 --- a/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/resolved_config.yaml +++ b/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/resolved_config.yaml @@ -9,6 +9,7 @@ training: num_gpus: 1 batch_size: 128 seed: 42 + report_to: wandb model: rkd_teacher_source: timewarp_action rkd_timewarp_vae_checkpoint: ${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46 diff --git a/TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1_20260702_121805.log b/TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1_20260702_121805.log new file mode 100644 index 0000000000000000000000000000000000000000..85ea8e05f8abdea4a358ba2d956174961326c6b5 --- /dev/null +++ b/TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1_20260702_121805.log @@ -0,0 +1,195 @@ +[robosuite WARNING] Could not import robosuite_models. Some robots may not be available. If you want to use these robots, please install robosuite_models from source (https://github.com/ARISE-Initiative/robosuite_models) or through pip install. (__init__.py:30) +[robosuite WARNING] Could not load the mink-based whole-body IK. Make sure you install related import properly (e.g. pip install mink==0.0.5), otherwise you will not be able to use the default IK controller setting for GR1 robot. (__init__.py:40) +WARNING: mimicgen environments not imported since mimicgen is not installed! +/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/albumentations/__init__.py:13: UserWarning: A new version of Albumentations is available: 2.0.8 (you have 1.4.18). Upgrade using: pip install -U albumentations. To disable automatic update checks, set the environment variable NO_ALBUMENTATIONS_UPDATE to 1. + check_for_updates() +`use_fast` is set to `True` but the image processor class does not have a fast version. Falling back to the slow version. + +================================================== +GR00T FINE-TUNING CONFIGURATION: +================================================== +config: /home/ext_minje/clvla/benchmarks/robocasa_v2/experiment_cfg/Experiments_v2/DTW/TimewarpVAE/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k.yaml +dataset_soup: None +output_dir: /tmp/gr00t +output_root: None +data_config: panda_omron +batch_size: 128 +max_steps: 300000 +num_gpus: 1 +save_steps: 20000 +run_name: None +save_total_limit: 100 +seed: 42 +base_model_path: nvidia/GR00T-N1.5-3B +tune_llm: False +tune_visual: False +tune_projector: True +tune_diffusion_model: True +resume: False +learning_rate: 3e-05 +weight_decay: 1e-05 +warmup_ratio: 0.05 +lora_rank: 0 +lora_alpha: 16 +lora_dropout: 0.1 +lora_full_model: False +dataloader_num_workers: 8 +report_to: wandb +embodiment_tag: new_embodiment +video_backend: opencv +balance_dataset_weights: True +balance_trajectory_weights: True +ds_weights_alpha: 0.4 +================================================== + +Using 1 GPUs + +================================================================================ +Starting sweep branch: default +Sweep vars: {} +================================================================================ + +-------------------------------------------------------------------------------- +Running phase 1: fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k +Policy type: groot_RKDv3 +Base model path: nvidia/GR00T-N1.5-3B +Output dir: /home/ext_minje/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k +Trainable preset: None +Policy overrides: {'rkd_teacher_source': 'timewarp_action', 'rkd_timewarp_vae_checkpoint': '${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46', 'rkd_relation_mode': 'token_pair_mean', 'rkd_loss_type': 'distance_angle', 'rkd_student_reconstruction_head_pooling': 'attention', 'rkd_student_reconstruction_head_enabled': True, 'rkd_student_reconstruction_head_dim': 512, 'rkd_student_reconstruction_head_hidden_dim': 512, 'rkd_distance_loss_weight': 1.0, 'rkd_angle_loss_weight': 2.0, 'rkd_exclude_diagonal': True, 'rkd_enabled': True, 'rkd_fm_loss_weight': 1.0, 'rkd_loss_weight': 1.0, 'rkd_loss_weight_schedule': 'cosine_decay', 'rkd_loss_weight_start': 1.0, 'rkd_loss_weight_end': 0.0} +-------------------------------------------------------------------------------- + +[{'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseBlenderLid/20250822/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'CloseBlenderLid', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseFridge/20250819/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'CloseFridge', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseToasterOvenDoor/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'CloseToasterOvenDoor', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CoffeeSetupMug/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'CoffeeSetupMug', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/NavigateKitchen/20250821/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'NavigateKitchen', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenCabinet/20250819/lerobot', 'horizon': 1050, 'filter_key': '100_demos', 'task': 'OpenCabinet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenDrawer/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'OpenDrawer', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenStandMixerHead/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'OpenStandMixerHead', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCounterToCabinet/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'PickPlaceCounterToCabinet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCounterToStove/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'PickPlaceCounterToStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceDrawerToCounter/20250820/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'PickPlaceDrawerToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceSinkToCounter/20250819/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'PickPlaceSinkToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceToasterToCounter/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'PickPlaceToasterToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCabinetToCounter/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'PickPlaceCabinetToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnElectricKettle/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnElectricKettle', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnMicrowave/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnMicrowave', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffMicrowave/20250819/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnOffMicrowave', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenElectricKettleLid/20250820/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'OpenElectricKettleLid', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/SlideDishwasherRack/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'SlideDishwasherRack', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnSinkFaucet/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'TurnOnSinkFaucet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnSinkSpout/20250820/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnSinkSpout', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffSinkFaucet/20250819/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnOffSinkFaucet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/AdjustWaterTemperature/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'AdjustWaterTemperature', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffStove/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'TurnOffStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnStove/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/AdjustToasterOvenTemperature/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'AdjustToasterOvenTemperature', 'split': 'pretrain', 'source': 'human'}] +Using 100 subset demos for filter_key: 100_demos +/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/data/transform/state_action.py:257: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.detach().clone() or sourceTensor.detach().clone().requires_grad_(True), rather than torch.tensor(sourceTensor). + self.statistics[key] = torch.tensor(value) +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +dataset weights: [1. 0.88494669 0.76443286 0.83888607 0.73785464 1.00501644 + 0.80029956 0.66065525 0.83967491 0.83515002 0.95047759 0.86791692 + 0.88329477 0.78510653 0.64381843 0.67404766 0.69697218 0.60443684 + 0.78448103 0.83447486 0.61146215 0.64384058 0.79545025 0.93845856 + 0.75517122 0.7973985 ] +Loaded 26 datasets +Loading pretrained dual brain from nvidia/GR00T-N1.5-3B +Tune backbone vision tower: False +Tune backbone LLM: False +Tune action head projector: True +Tune action head DiT: True + Fetching 13 files: 0%| | 0/13 [00:00 0.0 (cosine_decay) +Traceback (most recent call last): + File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 1158, in + run_yaml_experiment( + File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 1117, in run_yaml_experiment + main(args, policy_type=policy_type, policy_overrides=policy_overrides, trainable_preset=preset) + File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 847, in main + experiment.train() + File "/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/experiment/runner.py", line 172, in train + self.trainer.train(resume_from_checkpoint=self.resume_from_checkpoint) + File "/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/experiment/trainer.py", line 176, in train + return super().train(resume_from_checkpoint, trial, ignore_keys_for_eval, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer.py", line 2245, in train + return inner_training_loop( + ^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer.py", line 2469, in _inner_training_loop + self.control = self.callback_handler.on_train_begin(args, self.state, self.control) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer_callback.py", line 506, in on_train_begin + return self.call_event("on_train_begin", args, state, control) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer_callback.py", line 556, in call_event + result = getattr(callback, event)( + ^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/integrations/integration_utils.py", line 930, in on_train_begin + self.setup(args, state, model, **kwargs) + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/integrations/integration_utils.py", line 857, in setup + self._wandb.init( + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 1595, in init + get_sentry().reraise(e) + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/analytics/sentry.py", line 190, in reraise + raise exc.with_traceback(tb) + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 1516, in init + wi.maybe_login(init_settings) + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 193, in maybe_login + wandb_login._login( + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_login.py", line 190, in _login + auth = _find_or_prompt_for_key( + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_login.py", line 266, in _find_or_prompt_for_key + auth = wbauth.authenticate_session( + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/lib/wbauth/authenticate.py", line 139, in authenticate_session + raise UsageError( +wandb.errors.errors.UsageError: No API key configured. Use `wandb login` to log in. diff --git a/TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1_tensorboard_20260702_122542.log b/TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1_tensorboard_20260702_122542.log new file mode 100644 index 0000000000000000000000000000000000000000..f11994f2f6973cd11aef5b4629863ce1e4424ced --- /dev/null +++ b/TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k_gpu7_bs128_numgpu1_tensorboard_20260702_122542.log @@ -0,0 +1,226 @@ +[robosuite WARNING] Could not import robosuite_models. Some robots may not be available. If you want to use these robots, please install robosuite_models from source (https://github.com/ARISE-Initiative/robosuite_models) or through pip install. (__init__.py:30) +[robosuite WARNING] Could not load the mink-based whole-body IK. Make sure you install related import properly (e.g. pip install mink==0.0.5), otherwise you will not be able to use the default IK controller setting for GR1 robot. (__init__.py:40) +WARNING: mimicgen environments not imported since mimicgen is not installed! +/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/albumentations/__init__.py:13: UserWarning: A new version of Albumentations is available: 2.0.8 (you have 1.4.18). Upgrade using: pip install -U albumentations. To disable automatic update checks, set the environment variable NO_ALBUMENTATIONS_UPDATE to 1. + check_for_updates() +`use_fast` is set to `True` but the image processor class does not have a fast version. Falling back to the slow version. + +================================================== +GR00T FINE-TUNING CONFIGURATION: +================================================== +config: /home/ext_minje/clvla/benchmarks/robocasa_v2/experiment_cfg/Experiments_v2/DTW/TimewarpVAE/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k.yaml +dataset_soup: None +output_dir: /tmp/gr00t +output_root: None +data_config: panda_omron +batch_size: 128 +max_steps: 300000 +num_gpus: 1 +save_steps: 20000 +run_name: None +save_total_limit: 100 +seed: 42 +base_model_path: nvidia/GR00T-N1.5-3B +tune_llm: False +tune_visual: False +tune_projector: True +tune_diffusion_model: True +resume: False +learning_rate: 3e-05 +weight_decay: 1e-05 +warmup_ratio: 0.05 +lora_rank: 0 +lora_alpha: 16 +lora_dropout: 0.1 +lora_full_model: False +dataloader_num_workers: 8 +report_to: tensorboard +embodiment_tag: new_embodiment +video_backend: opencv +balance_dataset_weights: True +balance_trajectory_weights: True +ds_weights_alpha: 0.4 +================================================== + +Using 1 GPUs + +================================================================================ +Starting sweep branch: default +Sweep vars: {} +================================================================================ + +-------------------------------------------------------------------------------- +Running phase 1: fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k +Policy type: groot_RKDv3 +Base model path: nvidia/GR00T-N1.5-3B +Output dir: /home/ext_minje/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_120k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_120k +Trainable preset: None +Policy overrides: {'rkd_teacher_source': 'timewarp_action', 'rkd_timewarp_vae_checkpoint': '${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46', 'rkd_relation_mode': 'token_pair_mean', 'rkd_loss_type': 'distance_angle', 'rkd_student_reconstruction_head_pooling': 'attention', 'rkd_student_reconstruction_head_enabled': True, 'rkd_student_reconstruction_head_dim': 512, 'rkd_student_reconstruction_head_hidden_dim': 512, 'rkd_distance_loss_weight': 1.0, 'rkd_angle_loss_weight': 2.0, 'rkd_exclude_diagonal': True, 'rkd_enabled': True, 'rkd_fm_loss_weight': 1.0, 'rkd_loss_weight': 1.0, 'rkd_loss_weight_schedule': 'cosine_decay', 'rkd_loss_weight_start': 1.0, 'rkd_loss_weight_end': 0.0} +-------------------------------------------------------------------------------- + +[{'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseBlenderLid/20250822/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'CloseBlenderLid', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseFridge/20250819/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'CloseFridge', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseToasterOvenDoor/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'CloseToasterOvenDoor', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CoffeeSetupMug/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'CoffeeSetupMug', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/NavigateKitchen/20250821/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'NavigateKitchen', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenCabinet/20250819/lerobot', 'horizon': 1050, 'filter_key': '100_demos', 'task': 'OpenCabinet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenDrawer/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'OpenDrawer', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenStandMixerHead/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'OpenStandMixerHead', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCounterToCabinet/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'PickPlaceCounterToCabinet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCounterToStove/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'PickPlaceCounterToStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceDrawerToCounter/20250820/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'PickPlaceDrawerToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceSinkToCounter/20250819/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'PickPlaceSinkToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceToasterToCounter/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'PickPlaceToasterToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCabinetToCounter/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'PickPlaceCabinetToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnElectricKettle/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnElectricKettle', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnMicrowave/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnMicrowave', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffMicrowave/20250819/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnOffMicrowave', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenElectricKettleLid/20250820/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'OpenElectricKettleLid', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/SlideDishwasherRack/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'SlideDishwasherRack', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnSinkFaucet/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'TurnOnSinkFaucet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnSinkSpout/20250820/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnSinkSpout', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffSinkFaucet/20250819/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnOffSinkFaucet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/AdjustWaterTemperature/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'AdjustWaterTemperature', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffStove/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'TurnOffStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnStove/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/AdjustToasterOvenTemperature/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'AdjustToasterOvenTemperature', 'split': 'pretrain', 'source': 'human'}] +Using 100 subset demos for filter_key: 100_demos +/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/data/transform/state_action.py:257: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.detach().clone() or sourceTensor.detach().clone().requires_grad_(True), rather than torch.tensor(sourceTensor). + self.statistics[key] = torch.tensor(value) +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +dataset weights: [1. 0.88494669 0.76443286 0.83888607 0.73785464 1.00501644 + 0.80029956 0.66065525 0.83967491 0.83515002 0.95047759 0.86791692 + 0.88329477 0.78510653 0.64381843 0.67404766 0.69697218 0.60443684 + 0.78448103 0.83447486 0.61146215 0.64384058 0.79545025 0.93845856 + 0.75517122 0.7973985 ] +Loaded 26 datasets +Loading pretrained dual brain from nvidia/GR00T-N1.5-3B +Tune backbone vision tower: False +Tune backbone LLM: False +Tune action head projector: True +Tune action head DiT: True + Fetching 13 files: 0%| | 0/13 [00:00 0.0 (cosine_decay) + 0%| | 0/120000 [00:00 0.0 (cosine_decay) +Traceback (most recent call last): + File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 1158, in + run_yaml_experiment( + File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 1117, in run_yaml_experiment + main(args, policy_type=policy_type, policy_overrides=policy_overrides, trainable_preset=preset) + File "/home/ext_minje/clvla/benchmarks/robocasa_v2/my_scripts/gr00t_finetune.py", line 847, in main + experiment.train() + File "/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/experiment/runner.py", line 172, in train + self.trainer.train(resume_from_checkpoint=self.resume_from_checkpoint) + File "/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/experiment/trainer.py", line 176, in train + return super().train(resume_from_checkpoint, trial, ignore_keys_for_eval, **kwargs) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer.py", line 2245, in train + return inner_training_loop( + ^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer.py", line 2469, in _inner_training_loop + self.control = self.callback_handler.on_train_begin(args, self.state, self.control) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer_callback.py", line 506, in on_train_begin + return self.call_event("on_train_begin", args, state, control) + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/trainer_callback.py", line 556, in call_event + result = getattr(callback, event)( + ^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/integrations/integration_utils.py", line 930, in on_train_begin + self.setup(args, state, model, **kwargs) + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/transformers/integrations/integration_utils.py", line 857, in setup + self._wandb.init( + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 1595, in init + get_sentry().reraise(e) + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/analytics/sentry.py", line 190, in reraise + raise exc.with_traceback(tb) + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 1516, in init + wi.maybe_login(init_settings) + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_init.py", line 193, in maybe_login + wandb_login._login( + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_login.py", line 190, in _login + auth = _find_or_prompt_for_key( + ^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/wandb_login.py", line 266, in _find_or_prompt_for_key + auth = wbauth.authenticate_session( + ^^^^^^^^^^^^^^^^^^^^^^^^^^^^ + File "/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/wandb/sdk/lib/wbauth/authenticate.py", line 139, in authenticate_session + raise UsageError( +wandb.errors.errors.UsageError: No API key configured. Use `wandb login` to log in. diff --git a/TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k_gpu0_bs128_numgpu1_tensorboard_20260702_122542.log b/TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k_gpu0_bs128_numgpu1_tensorboard_20260702_122542.log new file mode 100644 index 0000000000000000000000000000000000000000..663a957f29d50ae0d3468057620c961098b6632e --- /dev/null +++ b/TimewarpVAE/logs/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k_gpu0_bs128_numgpu1_tensorboard_20260702_122542.log @@ -0,0 +1,226 @@ +[robosuite WARNING] Could not import robosuite_models. Some robots may not be available. If you want to use these robots, please install robosuite_models from source (https://github.com/ARISE-Initiative/robosuite_models) or through pip install. (__init__.py:30) +[robosuite WARNING] Could not load the mink-based whole-body IK. Make sure you install related import properly (e.g. pip install mink==0.0.5), otherwise you will not be able to use the default IK controller setting for GR1 robot. (__init__.py:40) +WARNING: mimicgen environments not imported since mimicgen is not installed! +/home/ext_minje/miniconda3/envs/robocasa/lib/python3.11/site-packages/albumentations/__init__.py:13: UserWarning: A new version of Albumentations is available: 2.0.8 (you have 1.4.18). Upgrade using: pip install -U albumentations. To disable automatic update checks, set the environment variable NO_ALBUMENTATIONS_UPDATE to 1. + check_for_updates() +`use_fast` is set to `True` but the image processor class does not have a fast version. Falling back to the slow version. + +================================================== +GR00T FINE-TUNING CONFIGURATION: +================================================== +config: /home/ext_minje/clvla/benchmarks/robocasa_v2/experiment_cfg/Experiments_v2/DTW/TimewarpVAE/RKDv3_TimewarpVAE_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k.yaml +dataset_soup: None +output_dir: /tmp/gr00t +output_root: None +data_config: panda_omron +batch_size: 128 +max_steps: 300000 +num_gpus: 1 +save_steps: 20000 +run_name: None +save_total_limit: 100 +seed: 42 +base_model_path: nvidia/GR00T-N1.5-3B +tune_llm: False +tune_visual: False +tune_projector: True +tune_diffusion_model: True +resume: False +learning_rate: 3e-05 +weight_decay: 1e-05 +warmup_ratio: 0.05 +lora_rank: 0 +lora_alpha: 16 +lora_dropout: 0.1 +lora_full_model: False +dataloader_num_workers: 8 +report_to: tensorboard +embodiment_tag: new_embodiment +video_backend: opencv +balance_dataset_weights: True +balance_trajectory_weights: True +ds_weights_alpha: 0.4 +================================================== + +Using 1 GPUs + +================================================================================ +Starting sweep branch: default +Sweep vars: {} +================================================================================ + +-------------------------------------------------------------------------------- +Running phase 1: fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k +Policy type: groot_RKDv3 +Base model path: nvidia/GR00T-N1.5-3B +Output dir: /home/ext_minje/groot_robocasa/robocasa_v2/TimewarpVAE/RKDv3_DA_APHead_FM_RKD_CosineDecay_from_GR00TBase_to_60k/default/fm_rkdv3_timewarp_vae_da_aphead_cosine_decay_60k +Trainable preset: None +Policy overrides: {'rkd_teacher_source': 'timewarp_action', 'rkd_timewarp_vae_checkpoint': '${HOME}/groot_robocasa/TimeWarping/TimewarpVAE/robocasa_vanilla_t16_z8_g8_h32_reg0p01_beta0p01/checkpoint-step-60000-epoch-17.46', 'rkd_relation_mode': 'token_pair_mean', 'rkd_loss_type': 'distance_angle', 'rkd_student_reconstruction_head_pooling': 'attention', 'rkd_student_reconstruction_head_enabled': True, 'rkd_student_reconstruction_head_dim': 512, 'rkd_student_reconstruction_head_hidden_dim': 512, 'rkd_distance_loss_weight': 1.0, 'rkd_angle_loss_weight': 2.0, 'rkd_exclude_diagonal': True, 'rkd_enabled': True, 'rkd_fm_loss_weight': 1.0, 'rkd_loss_weight': 1.0, 'rkd_loss_weight_schedule': 'cosine_decay', 'rkd_loss_weight_start': 1.0, 'rkd_loss_weight_end': 0.0} +-------------------------------------------------------------------------------- + +[{'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseBlenderLid/20250822/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'CloseBlenderLid', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseFridge/20250819/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'CloseFridge', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CloseToasterOvenDoor/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'CloseToasterOvenDoor', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/CoffeeSetupMug/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'CoffeeSetupMug', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/NavigateKitchen/20250821/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'NavigateKitchen', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenCabinet/20250819/lerobot', 'horizon': 1050, 'filter_key': '100_demos', 'task': 'OpenCabinet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenDrawer/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'OpenDrawer', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenStandMixerHead/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'OpenStandMixerHead', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCounterToCabinet/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'PickPlaceCounterToCabinet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCounterToStove/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'PickPlaceCounterToStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceDrawerToCounter/20250820/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'PickPlaceDrawerToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceSinkToCounter/20250819/lerobot', 'horizon': 900, 'filter_key': '100_demos', 'task': 'PickPlaceSinkToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceToasterToCounter/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'PickPlaceToasterToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/PickPlaceCabinetToCounter/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'PickPlaceCabinetToCounter', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnElectricKettle/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnElectricKettle', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnMicrowave/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnMicrowave', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffMicrowave/20250819/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnOffMicrowave', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/OpenElectricKettleLid/20250820/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'OpenElectricKettleLid', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/SlideDishwasherRack/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'SlideDishwasherRack', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnSinkFaucet/20250819/lerobot', 'horizon': 600, 'filter_key': '100_demos', 'task': 'TurnOnSinkFaucet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnSinkSpout/20250820/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnSinkSpout', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffSinkFaucet/20250819/lerobot', 'horizon': 300, 'filter_key': '100_demos', 'task': 'TurnOffSinkFaucet', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/AdjustWaterTemperature/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'AdjustWaterTemperature', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOffStove/20250819/lerobot', 'horizon': 750, 'filter_key': '100_demos', 'task': 'TurnOffStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/TurnOnStove/20250819/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'TurnOnStove', 'split': 'pretrain', 'source': 'human'}, {'path': '/home/ext_minje/clvla/benchmarks/robocasa_v2/robocasa/datasets/v1.0/pretrain/atomic/AdjustToasterOvenTemperature/20250820/lerobot', 'horizon': 450, 'filter_key': '100_demos', 'task': 'AdjustToasterOvenTemperature', 'split': 'pretrain', 'source': 'human'}] +Using 100 subset demos for filter_key: 100_demos +/home/ext_minje/clvla/benchmarks/robocasa_v2/Isaac-GR00T/gr00t/data/transform/state_action.py:257: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.detach().clone() or sourceTensor.detach().clone().requires_grad_(True), rather than torch.tensor(sourceTensor). + self.statistics[key] = torch.tensor(value) +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +Using 100 subset demos for filter_key: 100_demos +Initialized dataset lerobot with EmbodimentTag.NEW_EMBODIMENT +dataset weights: [1. 0.88494669 0.76443286 0.83888607 0.73785464 1.00501644 + 0.80029956 0.66065525 0.83967491 0.83515002 0.95047759 0.86791692 + 0.88329477 0.78510653 0.64381843 0.67404766 0.69697218 0.60443684 + 0.78448103 0.83447486 0.61146215 0.64384058 0.79545025 0.93845856 + 0.75517122 0.7973985 ] +Loaded 26 datasets +Loading pretrained dual brain from nvidia/GR00T-N1.5-3B +Tune backbone vision tower: False +Tune backbone LLM: False +Tune action head projector: True +Tune action head DiT: True + Fetching 13 files: 0%| | 0/13 [00:00 0.0 (cosine_decay) + 0%| | 0/60000 [00:00